Why are clicks so rare?

[Otto Kretschmer]

Why are they?

Only the languages of Kalahari Bushmen have them + Bantu languages that borrowed clicks from them.

Is this because they are inherently difficult to pronounce?

[bradrn]

It’s difficult to say for sure, but I strongly suspect that it is simply very difficult to get clicks from any reasonable sound change. I found a document which links the presence of clicks in Africa in particular to the presence of labiovelars: it’s easier to develop clicks from labiovelars than it is from most other consonants. (The only other area with any labiovelars at all is New Guinea — and indeed near there we find Damin.) Of course, this just pushes back the question to that of why labiovelars have such a skewed distribution — which is a mystery indeed, given that they’re very easy to pronounce, but are possibly even more skewed than clicks are, as shown by PHOIBLE.

On the other hand, while clicks might be very difficult to innovate, they are very easy to borrow. Even Europeans start to use them after a while:

[Meinhof] cites [Lucy] Lloyd’s statement to the effect that it was difficult for her, after long contact with the Bushmen and Hottentots, to refrain from using clicks when speaking English. Certainly clicking is to some extent “contagious,” and I have often heard Europeans in South Africa sportively substituting clicks for other consonants when speaking English, in imitation of the Hottentot or Zulu.

(By the way, I should note that ‘Bushman’ can be considered somewhat offensive today, though it’s at least better than ‘Hottentot’.)

[Pabappa]

That's a big leap of judgment though ...... I'll just say I don't buy it. Northern Australia and Papua New Guinea are not that close from the standpoint of premodern transportation, and evidence of that is that there aren't many isoglosses that cover both regions .... really, comparisons tend to focus on a single language spoken in the Torres Strait, where there has been definite contact, but even at the northernmost tip of the Australian landmass there is little or no evidence of Papuan influence.

Likewise, before the Bantu expansion, the various African languages with the coarticulated labial-velar consonants may have been separated by thousands of miles from the Khoisan languages with clicks, and we're just filling in blank spots on the map at this point because we don't know what was spoken in between. It could happen, but there's no evidence that it did .... so far as I know, we have no known instances of labial-velar consonants developing into clicks, whether in Bantu languages, Khoisan languages, or in Nilo-Saharan.

This theory would only explain the bilabial click, anyway .... and from what I can see on Wikipedia, not even one Bantu language has a bilabial click.

Lastly, while this may seem petty, the proper term is labial-velar, not labiovelar .... I'd expect a linguist to know that, and it may be a sign of sloppy research in general, as I've raised objections like this to other publications which seem poorly researched in general.

[Richard W]

Why are they?

Only the languages of Kalahari Bushmen have them + Bantu languages that borrowed clicks from them.

Is this because they are inherently difficult to pronounce?

There's another cluster in East Africa - Dahalo (Cushitic) , Sandawe and Hadza (isolate), though again that may be regional. Sandawe may be related to the southern African Khoisan. As far as I am aware, only Central Khoisan, Sandawe and Dahalo have sex-based gender.

[Otto Kretschmer]

A question - are the click consonants ancient or are they a recent invention?

[Rounin Ryuuji]

I imagine the ability of the human mouth to click has existed for quite a long time.

[Raphael]

I imagine the ability of the human mouth to click has existed for quite a long time.

Which arguably adds to the mystery.

[KathTheDragon]

I read a while ago (I don't recall which paper, so I can't find it again) that some German speakers occasionally realise certain consonant clusters with clicks, so it seems to me more that the rarity of clicks is due in large part to the highly specific conditions necessary for clicks to develop in the first place - heavy word-initial clusters of a certain type. Areal affects are possibly also at play. Languages seem more likely to develop clicks when there are nearby languages with them, and less likely when the opposite is true, so that would concentrate clicks into pockets that can only appear if they appear virtually all at once.

[anteallach]

I read a while ago (I don't recall which paper, so I can't find it again) that some German speakers occasionally realise certain consonant clusters with clicks, so it seems to me more that the rarity of clicks is due in large part to the highly specific conditions necessary for clicks to develop in the first place - heavy word-initial clusters of a certain type. Areal affects are possibly also at play. Languages seem more likely to develop clicks when there are nearby languages with them, and less likely when the opposite is true, so that would concentrate clicks into pockets that can only appear if they appear virtually all at once.

Possibly Weak Clicks in German? by Fuchs, Koenig and Winkler?

[Travis B.]

I read a while ago (I don't recall which paper, so I can't find it again) that some German speakers occasionally realise certain consonant clusters with clicks, so it seems to me more that the rarity of clicks is due in large part to the highly specific conditions necessary for clicks to develop in the first place - heavy word-initial clusters of a certain type. Areal affects are possibly also at play. Languages seem more likely to develop clicks when there are nearby languages with them, and less likely when the opposite is true, so that would concentrate clicks into pockets that can only appear if they appear virtually all at once.

Possibly Weak Clicks in German? by Fuchs, Koenig and Winkler?

That's a broken link for me.

[KathTheDragon]

That looks like it might be the one.

[Nortaneous]

It’s difficult to say for sure, but I strongly suspect that it is simply very difficult to get clicks from any reasonable sound change. I found a document which links the presence of clicks in Africa in particular to the presence of labiovelars: it’s easier to develop clicks from labiovelars than it is from most other consonants. (The only other area with any labiovelars at all is New Guinea — and indeed near there we find Damin.) Of course, this just pushes back the question to that of why labiovelars have such a skewed distribution — which is a mystery indeed, given that they’re very easy to pronounce, but are possibly even more skewed than clicks are, as shown by PHOIBLE.

But bilabial clicks are rare even for clicks! There was a paper about weak bilabial clicks as allophones of /mw/ in some West African language or other, though, and you occasionally get this even in English.

Most innovative lexical material in Damin probably developed out of paralinguistic accompaniments to the sign language of initiates. Have any similar registers - initiate sign, hunting jargons, etc. - been attested in click languages? But multiple paths of origin are possible, and it seems likely that they came from complex initial clusters at least in the Tuu languages. !Xoon seems to have a Kartveloid 'harmonic cluster' rule, so maybe *tk- *kt- > tx- k|-?

Clicks were probably much more widespread across the southern half of Africa before the Bantu expansion.

[bradrn]

That's a big leap of judgment though ...... I'll just say I don't buy it. Northern Australia and Papua New Guinea are not that close from the standpoint of premodern transportation, and evidence of that is that there aren't many isoglosses that cover both regions .... really, comparisons tend to focus on a single language spoken in the Torres Strait, where there has been definite contact, but even at the northernmost tip of the Australian landmass there is little or no evidence of Papuan influence.

Yes, this is definitely the weakest part of my post. I’m not sure I buy it myself; it’s just an interesting connection.

This theory would only explain the bilabial click, anyway .... and from what I can see on Wikipedia, not even one Bantu language has a bilabial click.

And I agree this is a problem as well.

Likewise, before the Bantu expansion, the various African languages with the coarticulated labial-velar consonants may have been separated by thousands of miles from the Khoisan languages with clicks, and we're just filling in blank spots on the map at this point because we don't know what was spoken in between. It could happen, but there's no evidence that it did .... so far as I know, we have no known instances of labial-velar consonants developing into clicks, whether in Bantu languages, Khoisan languages, or in Nilo-Saharan.

Well, we have very few known cases of any consonants developing into clicks, so anything we say is necessarily speculative.

Lastly, while this may seem petty, the proper term is labial-velar, not labiovelar .... I'd expect a linguist to know that, and it may be a sign of sloppy research in general, as I've raised objections like this to other publications which seem poorly researched in general.

Wikipedia says that ‘labiovelar’ can be used for both ‘labial–velar’ and ‘labialised velar’, and this is what I am used to. Generally I am a stickler for precise terminology, but here it doesn’t matter so much since it’s clear from context.

Why are they?

Only the languages of Kalahari Bushmen have them + Bantu languages that borrowed clicks from them.

Is this because they are inherently difficult to pronounce?

There's another cluster in East Africa - Dahalo (Cushitic) , Sandawe and Hadza (isolate), though again that may be regional. Sandawe may be related to the southern African Khoisan. As far as I am aware, only Central Khoisan, Sandawe and Dahalo have sex-based gender.

I’ve heard that the Khoe–Kwadi languages also originated in this cluster, but I can’t remember where I saw it. It would certainly make sense given that Khoe–Kwadi languages have fewer clicks than the southern Khoisan languages do.

!Xoon seems to have a Kartveloid 'harmonic cluster' rule, so maybe *tk- *kt- > tx- k|-?

Sounds interesting. Do you have any links where I can read more about this? (In either ǃXóõ or Kartvelian.)

Clicks were probably much more widespread across the southern half of Africa before the Bantu expansion.

I think this very unlikely. As I said, clicks are extremely easy to borrow; I think that if there were more African languages with clicks, the languages in that region would still have clicks, even after language replacement.

---

One thing that strikes me: it looks like it is possible to develop phonetic clicks from other consonants. I knew about the German paper already (not sure what I was thinking in not including it in my OP); Nortaneous says /mw/ can be pronounced as [ʘ], which sounds reasonable. I suppose the question then becomes: why is it the case that weak clicks are relatively common, but only phonologised and developed into strong clicks in Africa?

[Nortaneous]

Wikipedia says that ‘labiovelar’ can be used for both ‘labial–velar’ and ‘labialised velar’, and this is what I am used to. Generally I am a stickler for precise terminology, but here it doesn’t matter so much since it’s clear from context.

This is technically true but I think outdated.

!Xoon seems to have a Kartveloid 'harmonic cluster' rule, so maybe *tk- *kt- > tx- k|-?

Sounds interesting. Do you have any links where I can read more about this? (In either ǃXóõ or Kartvelian.)

I'm just looking at the phoneme inventory. For whatever reason Wikipedia treats sequences like tx- and pʼkxʼ- as units.

"Harmonic cluster" refers to Kartvelian clusters consisting of a labial or coronal plosive followed by a velar plosive or fricative with the same MOA - e.g. /pk- dg- tsʼkʼ-/ ("Type A") and /px- dɣ- tsʼχʼ-/ ("Type B") - which are often considered to be complex segments.

I think this very unlikely. As I said, clicks are extremely easy to borrow; I think that if there were more African languages with clicks, the languages in that region would still have clicks, even after language replacement.

Extensive borrowing of clicks into the Nguni languages was facilitated by the extensive syllable taboo systems of the speakers. See here. This doesn't explain Yeyi, though.

Archaeological evidence is starting to indicate that the Bantu expansion had some pretty violent parts. No one wants to talk about that too much, but I don't think it can be assumed that the preconditions for extensive borrowing existed throughout. (How well do extant click languages line up with Köppen zones? Or subsistence patterns? For how long have the Yeyi been primarily agriculturalists? Do they have an elevated rate of lactase persistence?)

Nortaneous says /mw/ can be pronounced as [ʘ], which sounds reasonable.

Well, [ŋʘʷ].

I suppose the question then becomes: why is it the case that weak clicks are relatively common, but only phonologised and developed into strong clicks in Africa?

There's also Damin - maybe some Australian languages would've had clicks if it'd developed a thousand years earlier and stuck around? It seems like a safe assumption that modality transfer from a signed language to an oral jargon is pretty rare

As for the phonologization question, maybe having complete series of clicks is important? The allophonic weak clicks that are known thus far are nowhere near as Cartesian as inventories of strong click phonemes almost universally are. A solitary [ŋʘʷ] or [k|] may just not be strong enough to reach fixation or resist erosion, the same way it's rare to have languages with only one front rounded vowel or whatever else.

[bradrn]

Wikipedia says that ‘labiovelar’ can be used for both ‘labial–velar’ and ‘labialised velar’, and this is what I am used to. Generally I am a stickler for precise terminology, but here it doesn’t matter so much since it’s clear from context.

This is technically true but I think outdated.

But ‘labiovelar’ is a whole four bytes shorter! Then again, you’re better at phonetics than me, so I should trust you on this.

I think this very unlikely. As I said, clicks are extremely easy to borrow; I think that if there were more African languages with clicks, the languages in that region would still have clicks, even after language replacement.

Extensive borrowing of clicks into the Nguni languages was facilitated by the extensive syllable taboo systems of the speakers. See here. This doesn't explain Yeyi, though.

Sure, I know about hlonipha and its role in click acquisition. (In fact some searches turned up the same article you linked, though I haven’t gotten around to reading it.) But as you note it doesn’t explain Yeyi. It also doesn’t explain Dahalo, though I hear there’s evidence they may have once spoken a language like Sandawe (attributed to Ten Raa 1969, but I can’t find a copy). I still think ‘very hard to innovate, very easy to borrow’ is the likeliest explanation.

I suppose the question then becomes: why is it the case that weak clicks are relatively common, but only phonologised and developed into strong clicks in Africa?

There's also Damin - maybe some Australian languages would've had clicks if it'd developed a thousand years earlier and stuck around?

Given 1000 years, who knows? But remember that Damin is restricted to warama-initiated men — as such that exerts a pressure against Damin words and sounds being borrowed into everyday language.

As for the phonologization question, maybe having complete series of clicks is important? The allophonic weak clicks that are known thus far are nowhere near as Cartesian as inventories of strong click phonemes almost universally are. A solitary [ŋʘʷ] or [k|] may just not be strong enough to reach fixation or resist erosion, the same way it's rare to have languages with only one front rounded vowel or whatever else.

I think this is the most reasonable suggestion I’ve seen so far.

[WeepingElf]

The term labiovelar is conventionally used, especially in Indo-European linguistics, for a labialized velar, which is something entirely different than a labial-velar.

As for clicks, they probably were an areal feature widespread in sub-Equatorial Africa before most of the languages there were displaced by the Bantu languages. I see no reason to assume that Proto-World (if that ever existed) had clicks and all non-"Khoisan" languages have lost them.

[Nortaneous]

Given 1000 years, who knows? But remember that Damin is restricted to warama-initiated men — as such that exerts a pressure against Damin words and sounds being borrowed into everyday language.

It's only taught to initiates, but my impression is that they didn't try to stop non-initiates from developing at least passive comprehension. So it could happen. (Note that Damin has a series of nasal clicks: ŋʘ ŋǃ ŋ‼ + ŋ|2 ŋǃ2. The dental click doesn't appear plain, but n=1 for the featural significance of click repetition.)

PHOIBLE says there are two languages with only one click: Ronga and Sesotho. Its source says Ronga can have zero, one, or possibly two clicks: there's a dental click that varies with /ndʒ/ and a palatal click that varies with /dl/.

[bradrn]

The term labiovelar is conventionally used, especially in Indo-European linguistics, for a labialized velar, which is something entirely different than a labial-velar.

I wouldn’t say ‘entirely different’. There’s a good few cases of them interconverting: Sougb notably writes ⟨gb⟩ for phonetic [gʷ̚], and there’s some Vanuatuan languages which have developed [k͡pʷ] from [kʷ]. I’m sorely tempted to declare that they are the same consonant, at least on a phonemic level — the only language I can find with both is Maʼdi, and its [k͡p] is obviously the phonetic realisation of underlying /pʷ/.

As for clicks, they probably were an areal feature widespread in sub-Equatorial Africa before most of the languages there were displaced by the Bantu languages. I see no reason to assume that Proto-World (if that ever existed) had clicks and all non-"Khoisan" languages have lost them.

But as I said, this contradicts the evidence that clicks are easily borrowed — to make this hypothesis work you’ll have to assume that not one language between Tanzania and the Kalahari borrowed clicks, yet that area had many languages with clicks before the Bantu expansion.

That being said, there is some evidence to suggest that Khoe–Kwadi originated to the NE of its present position, and there is some poor evidence that it is related to Sandawe. I’m not sure how to reconcile this with the fact that click borrowing seems to be extensive in modern-day Bantu languages.

Given 1000 years, who knows? But remember that Damin is restricted to warama-initiated men — as such that exerts a pressure against Damin words and sounds being borrowed into everyday language.

It's only taught to initiates, but my impression is that they didn't try to stop non-initiates from developing at least passive comprehension. So it could happen.

Not just passive; there was active usage in some cases. But the fact remains that it was traditionally taboo for non-warama to use Damin, and as I said that restricts the possibilities for borrowing.

[Nortaneous]

As for clicks, they probably were an areal feature widespread in sub-Equatorial Africa before most of the languages there were displaced by the Bantu languages. I see no reason to assume that Proto-World (if that ever existed) had clicks and all non-"Khoisan" languages have lost them.

But as I said, this contradicts the evidence that clicks are easily borrowed — to make this hypothesis work you’ll have to assume that not one language between Tanzania and the Kalahari borrowed clicks, yet that area had many languages with clicks before the Bantu expansion.

There's evidence and there's counterevidence - as far as I know, none of the Indo-European langauges of South Africa have clicks, and there aren't other click languages near Sandawe, Hadza, and Dahalo. If it's true that the introduction of clicks into Bantu languages have two epicenters, all that's needed is an explanation for the cluster centered on Yeyi, since Dahalo and the Nguni languages both had relevant circumstances.

Then again, how many of those languages acquired clicks due to contact? Maybe there was some sprachbund effect that caused convergence on clicks or the precursors thereof, but it's hard to imagine Hadza participating in that unless it's been substantially phonologically restructured since then. (Which Khoekhoe maybe also was? It has a lot fewer consonants than its relatives, at least. I don't know anything about the historical phonology of these languages and am completely talking out of my ass here - the harmonic clusters in G|ui could be borrowed or something instead.)

edit: apparently Starostin thought Hadza was more likely Afro-Asiatic than 'Khoisan'; Sands 2014 proposes Hadza, Khoe-Kwadi-Sandawe, and Tuu-Kx'a as three distinct phyla

[bradrn]

As for clicks, they probably were an areal feature widespread in sub-Equatorial Africa before most of the languages there were displaced by the Bantu languages. I see no reason to assume that Proto-World (if that ever existed) had clicks and all non-"Khoisan" languages have lost them.

But as I said, this contradicts the evidence that clicks are easily borrowed — to make this hypothesis work you’ll have to assume that not one language between Tanzania and the Kalahari borrowed clicks, yet that area had many languages with clicks before the Bantu expansion.

There's evidence and there's counterevidence - as far as I know, none of the Indo-European langauges of South Africa have clicks

Oorlams reportedly does, but that’s a creole.

and there aren't other click languages near Sandawe, Hadza, and Dahalo.

This is admittedly true.

If it's true that the introduction of clicks into Bantu languages have two epicenters, all that's needed is an explanation for the cluster centered on Yeyi, since Dahalo and the Nguni languages both had relevant circumstances.

What circumstances were there for Dahalo?

Then again, how many of those languages acquired clicks due to contact? Maybe there was some sprachbund effect that caused convergence on clicks or the precursors thereof, but it's hard to imagine Hadza participating in that unless it's been substantially phonologically restructured since then. (Which Khoekhoe maybe also was? It has a lot fewer consonants than its relatives, at least. I don't know anything about the historical phonology of these languages and am completely talking out of my ass here - the harmonic clusters in G|ui could be borrowed or something instead.)

edit: apparently Starostin thought Hadza was more likely Afro-Asiatic than 'Khoisan'; Sands 2014 proposes Hadza, Khoe-Kwadi-Sandawe, and Tuu-Kx'a as three distinct phyla

Haven’t read the paper (link please?), but that last proposal sounds reasonable to me, except I think Tuu–Kxʼa is more likely a Sprachbund than a clade. Khoe-Kwadi–Sandawe is a sensible proposal — the click systems are more similar to each other than either are to Tuu–Kxʼaa, Khoe-Kwadi seems to have come from north-east of the main Khoisan area, and there are similarities in the pronouns (though then again the similarities are no greater than those proposed for ‘Basque–NE Caucasian’). Sarostin I shall ignore as usual.