zompist bboard

I'm pretty sure this has been answered before but I can't seem to find anything:

Is there an explanation for why languages with small phoneme inventories seem to also have simple syllable structures?

If I were made to venture a guess, however stupid, I'd have to say that it's probably something to do with the rather overwhelming predominance of occlusives among the few consonants that a phonologically simple language is likely to possess.

In Central Rotokas, for an extreme example, all consonants are plosives in only three POAs with only voicing to distinguish /b d g/ from /p t k/. Plosive clusters are hard enough to pronounce, and I think there is a tendency for one of the elements to become a fricative (if I remember correctly, one of the sound changes between Ancient and Modern Greek is p > f / _t). Also, it's difficult to maintain voicing in one element of a cluster when the other doesn't have it, and vice versa.

That seems like a reasonable explaination.

Also, can anyone think of any counterexamples?

You do find at least CV(C) (with restricted codas) in many small phonologies, like in the Algonquian langs. I don't offhand know of any truly small phonologies (<14 consonants) where clusters of more than two consonants are permitted.

If you mean overall complexity of syllables, rather than just clusters specifically, you do get various langs (e.g. Polynesian, Pirahã) with complex structures for how vowels and/or tones and/or lengths can be arranged in multimoraic syllables.

....

Thinking more on the topic, I notice also that the languages with perhaps the most complex syllables in the world, the Pacific Northwest langs, are also notable for having some of the largest consonant inventories. None beat Ubykh or anything, but most of the PNW langs are still around the 40-50 range, which as an areal feature is behind only the Caucasian and the Khoisan areas.

So it seems likely there is at least a loose correlation between inventory size and potential cluster permissivity.

This points to a possible explanation in terms of phonetic feature sensitivity. Consider this: for a given syllable structure and speed of speech, the more consonants a language distinguishes, the higher the rate at which phonetic consonant features are produced and listened for. That is, in a lang like Rotokas, speakers need only listen to two consonant features per syllable - POA and voicing - with all others irrelevant. Add in just one more consonant series, say a nasal series, and instead of two features per syllable speakers now have to maintain three: the "feature production" rate has gone up by 50%.

So my hypothesis is that the size-complexity correlation (to the extent there is one) may be interpretable in terms of languages differing in their tolerance for high feature-production rates. Those with high tolerance may end up with simple or complex syllable structures as the fortunes of diachronic accident decree, while those with low tolerances may thusly possess motivation against sound changes that would increase the feature production rate, including clusters but also phonemic splits.

We could test this hypothesis by looking up diachronic histories of small-inventory languages and looking for phonemic splits... though it'd be a lot of work.

You know, I think partly it's a matter of how linguists go about counting phonemes.

Suppose you're doing field work on a language with /p t k s x m n/ or some such, and then you hear someone say [tsa]. You've got a choice: do you analyze [ts] as a new consonant phoneme, /ts)/ or as a /t/+/s/ cluster? Most linguists seem to go for /ts)/, but it's not clear to me that there's always a strong empirical motivation for it; it's just the conventional thing to do under the circumstances.

Similarly, [kx] or [mp] or [nt] are likely, in such a language, to be analyzed as new complex consonant phonemes rather than as /k/+/x/, /m/+/p/, /n/+/t/. [hp ht hk] are likely to be analyzed as a new series of preaspirated consonant phonemes rather than as /h/+C clusters. The result is a "larger" consonant inventory and a "simpler" syllable structure.

Now, there are some combinations of phones — things like [kt] or [mk] or [sm] — that linguists will unanimously treat as clusters. But these combinations are typologically rare, and they only tend to show up in a language that also includes typologically more common combinations. That is, if a language allows [tk] it probably also allows [ts]; if it allows [mk] it probably also allows [mp]. And remember, [ts mp] etc. tend to get analyzed as single phonemes, meaning that a language which contains them is no longer considered to have a "small" consonant inventory.

(All that said, check out the Iroquoian languages. Here's Onondaga, f'rinstance, which has a mere fifteen segmental phonemes plus vowel length, and has some pretty fearsome consonant clusters. Some of those clusters could have been analyzed as phonemes in their own right — /tj/, for instance, or /kw/, or /tsh/, or /nt/ — but it would have made the morphology of the language hellishly difficult to describe, so the small-inventory-lotsa-clusters analysis won out.)

(BTW, I found a thread on a different forum addressing the same question: here. It doesn't actually solve the problem though...)

That leads to the exact opposite situation than what weldingfish and others have been talking about, though. That is, what you're saying is that because of the way linguists analyze what constitute phonemes, the more "phonemes" a language is considered to have, the simpler syllable its syllable structure will tend to be. But what weldingfish is asking about is the opposite observation (which, in my personal experience, tends to hold pretty well), where languages with fewer phonemes tend to have simpler syllable structures (or at least fewer consonant clusters), and those with more phonemes tend to have more complex syllable structures (or at least more consonant clusters).

Let's take some examples (based partly on previous posts in the thread), and choose three languages to look at: Rotokas (very few consonant phonemes), English (moderate number of consonant phonemes), and Columbia Salish (very large number of consonant phonemes). (Yes, I purposely choose them to illustrate my point, but too bad):

*Central Rotokas: six consonant phonemes, /p t k b d g/. There are no consonant clusters (though there are plenty of vowel sequences and other suprasegmental complexities).

*English: ~24 consonant phonemes (in most dialects), /p t k b d g/ /f v s z θ ð ʃ ʒ h/ /ʧ ʤ/ /m n ŋ/ /l ɹ w j/. Relatively complex syllable structures (details here).

*Columbia Salish [aka Nxaʔamxcín]: 45 consonant phonemes, /p t k kʷ q qʷ ʔ pʼ tʼ kʼ kʷʼ qʼ qʷʼ/ /s sˠ ɬ x xʷ χ χʷ ʕ ʕʼ ʕʷ ʕʷʼ ħ ħʷ h/ /ʦ ʦˠ ʦʼ tɬʼ/ /m mʼ n nʼ/ /w wʼ l lʼ lˠ lˠʼ r rʼ j jʼ/ (the consonants marked as "velarized" are apparently pronounced with a retracted tongue root; I've added /w wʼ/ to the list even though they're not listed in the source I'm using because they're found in plenty of words given in the same source, so I'm assuming an error). Batshit insane syllable structure, like all Salishan languages--to the extent that there's a huge amount of disagreement over what constitutes a "syllable" in Salishan languages, how many consonants can be "extrasyllabic", etc. Instead of providing a description of Nxaʔamxcín syllables (since there's no universally-accepted one), I'll provide a few examples of Nxaʔamxcín words, taken from "Simple Syllables in Nxaʔamxcín" by Ewa Czaykowska-Higgins and Marie Louise Willett, International Journal of American Linguistics (1997), pp. 385-411:

/ʦʼqʷʼqʷʼunlʼəx/ "land gets named"
/panʔitqps/ "springtime"
/ʦijatkʷpmtaʔ/ "build a fire"
/sxitlqswil/ "bow of canoe"
/snkɬxʷpawʼstən/ "clothesline"

Well, no, I'm not saying "Languages with small inventories have more clusters" (which, as you point out, is obviously false). It's a little more complicated than that.

With some languages, you have a choice between two analyses:

1) An analysis with a large phoneme inventory, but a simple syllable structure. Find a syllable that starts with [ts]? Okay, /ts/ is a phoneme. Find one that starts with [nts]? Okay, /nts/ is a phoneme. Find one that starts with [ntsw]? Okay, /nts_w/ is a phoneme.

2) An analysis with a small phoneme inventory, but a complicated syllable structure. Find a syllable that starts with [ts]? That's fine, don't add a /ts/ phoneme to your inventory, just call it a CCV syllable. Find one that starts with [nts]? Okay, call it a CCCV syllable. Find one that starts with [ntsw]? Okay, call it a CCCCV syllable.

If more linguists followed strategy 2, we would see more phonological descriptions that looked like counterexamples to weldingfish's generalization. It's because linguists tend to follow strategy 1 that we see few or none that look like counterexamples.

So part of the answer to weldingfish's question is, "Because of the way we like to do phonology." Rather than a purely naturalistic observation about how languages are ("when they have small consonant inventories, they also have simple syllables") we have a sociological observation about what linguists like ("given the choice, we prefer to claim a language has a big inventory with simple syllables, rather than a small inventory with complex syllables").

The trouble with your contention is that there appear to be few, if any, languages that actually would be notable counterexamples no matter how much you re-analyzed the phonology.

Keep in mind we already have examples of small-inventory CVC, so it needs to be better than that to provide a true counterexample: to count, a language shoulud have less than 15 consonants and permit clusters of at least three consonants in sequence under at least one possible analysis. I know of no such language. If you have any examples, please provide them - and if not, you are going to have a very hard time backing up your contention.

To be fair, he did provide such a language in an earlier post: Onondaga can be argued to permit three-consonant clusters, as can some other Iroquoian languages, although I don't think all three consonants are in a single syllable.

What I was more objecting to was that NWP's assumption suggests that both: languages analyzed as having small inventories will tend to be analyzed as having complex syllable structures and languages analyzed as having large inventories will tend to be analyzed as having simple syllable structures. And both of those are contrary to my experience, and to the examples already given in this thread and in my earlier post.

weldingfish wrote:Is there an explanation for why languages with small phoneme inventories seem to also have simple syllable structures?

WALS allows an even stronger claim: a small phoneme inventory not only statistically implies a simple syllable structure but there is even stronger evidence for simple syllable structure to statistically implies a small phoneme inventory. Thus one could of course also ask why languages with large inventories have complex syllables.

As a guess for the explanation as to why small phoneme inventories have simple syllable structures consider that to transmit the same amount of information in the same time, the language with the simpler syllable structure must be either spoken up to twice as fast or get a more complex phonology. Note that one cannot do both because the finer the distinctions between the phonemes the more time is needed to make it clear. For example, voiced, unvoiced and unmarked consonants differ in terms of voice onset time - thus if that distinction is made there must be enough time for it.

However, if we wanted to extend a phonology of 20 consonants such that a CV(C) syllable would carry as much information as a (C)C(C)V(C(C)) syllable we would need to nearly square the number of phonemes. Thus we would need a phonology with hundreds of consonants. But this would make speaking the language even slower. Therefore extending the phonology is not feasible, thus languages with simpler syllable structures have a smaller phonology.

Of course, it probably makes more sense to say that most consonant sound changes happen in clusters, thus if a language has a simpler syllable structure it will probably keep a smaller phonology because splits happen more rarely.

I have asked WALS, and found a weak but noticeable correlation: languages with small consonant inventories tend to have simple syllable structures, while languages with large consonant inventories tend to have complex syllable structures. Don't ask me why.

Hmm. Having taken some time over the weekend to hunt for those PERFECT EXAMPLES that I was SURE WERE OUT THERE... well, you probably don't need me to tell you I didn't find any. I concede.

In pennance, lemme offer up a link to another language that's got a small inventory and loads of clusters any way you slice it: Palauan. /a e i o u ǝ t k ʔ b d s m ŋ r l/, and you get such lovely words as klimd, dakt, tmuu and /mlai/. I ran across it while digging around for support for my own crackpot theory, but once I saw the light I figured it would be worth pointing out anyway.

Looks like vowel reduction is what's behind it. If there's a solid mechanism behind the small-inventory/simple-syllable correlation, I guess we should expect Palauan to add some phonemes to its inventory or simplify some clusters over the next few centuries. Anyone got a time machine?

Whimemsz wrote:/ʦʼqʷʼqʷʼunlʼəx/ "land gets named"
/panʔitqps/ "springtime"
/ʦijatkʷpmtaʔ/ "build a fire"
/sxitlqswil/ "bow of canoe"
/snkɬxʷpawʼstən/ "clothesline"

God I love Salish...I know that's not productive at all, I just needed to say it. And I want to provide a word that I love from Montana Salish, a language that I've been doing a lot of work on lately:

i cíkʷkʷkʷkʷ 'little shiny things sparkling -- sequins on girls' dresses, stars winking'

And each of those stops is released.

roninbodhisattva wrote:i cíkʷkʷkʷkʷ 'little shiny things sparkling -- sequins on girls' dresses, stars winking'

And each of those stops is released.

fuuuuuuuuuck

that's even more insane than tftktst tfktstt "you sprained it and then gave it" (from some berber lang)

nebula wind phone wrote:In pennance, lemme offer up a link to another language that's got a small inventory and loads of clusters any way you slice it: Palauan. /a e i o u ǝ t k ʔ b d s m ŋ r l/, and you get such lovely words as klimd, dakt, tmuu and /mlai/. I ran across it while digging around for support for my own crackpot theory, but once I saw the light I figured it would be worth pointing out anyway.

AHA! We got one! I'm definitely willing to take that as a counterexample. Peripheral two-consonant clusters - at both ends - is of pretty much equivalent structural complexity to medial three-consonant clusters, in that normally if a language permits /klimd/ it is likely to also permit /imdli/ or similar. And even if that's not the case for Palauan specifically, it's still a very good example.

My guess is stability - consonant clusters tend to undergo assimilation/other changes and develop new phonemes, and languages with few phonemes have lots of empty "consonant space" available for new phonemes to appear in, so if a language with few phonemes but complex syllable structure ever existed, it would have been unstable and likely to develop over time into a lang with more phonemes. Only the ones with simple syllable structure are stable and less likely to develop more phonemes, so they tend to remain. Palauan could be an example of this process happening - we'd have to wait a few hundred years to know.

I am trying to find out if this correlation is predicted by Optimality Theory. If a language has few phonemes this would mean according to OT that it has many markedness constraints forbidding certain phonemes dominating faithfulness constraints. If a language has a simple syllable structure it means that markedness constraints forbidding complicated syllable structures dominate faithfulness constraints.
Now say we have a simple toy grammar with two markedness constraints forbidding morphemes (P1, P2), two markedness constraints forbidding complicated syllable constraints (S1, S2) and one faithfulness constraint (F). Lets say this language has the order P1 > P2 > F which would mean that it has very little phonemes. Now there are only 2 orders in which F dominates both S's (P1>P2>F>S1>S2 and P1>P2>F>S2>S1). But there are 12 orders in which S1 and S2 both dominate F. (S1>S2>P1>P2>F, S1>P1>S2>P2>F, S1>P1>P2>S2>F, P1>S1>S2>P2>F, P1>S1>P2>S2>F, P1>P2>S1>S2>F, S2>S1>P1>P2>F, S2>P1>S1>P2>F, S2>P1>P2>S1>F, P1>S2>S1>P2>F, P1>S2>P2>S1>F, P1>P2>S2>S1>F). So it is much more likely that the language has a simple syllable structure.

It is an interesting question. Here's the best answers I can suggest:

1.
- consonant clusters produce more allophony, both of the clusters and of the vowels around them

- we instinctive think "oh, but those allophones don't become phonetic until the cluster simplifies". But that's not true. The cluster doesn't have to simplify COMPLETELY - it only has to change in some way. So yes, some of the new phonemes will compensate for some losses in complexity, but sometimes you'll get new phonemes without losing the clusters. Let's say that alveolars gain rounding after labiovelars but not after velars, and that labiovelars become velars when followed by a consonant, unless they follow /w/. Those rules get us from /awkt/, /akt/, /awk_wt/ and /ak_wt/ (three stop series) to /awkt/, /akt/, /awk_wt_w/ and /akt_w/ - thow in, say, some more vowel reduction between glides and consonants, and /awakt/ becomes /awkt/, which means we've ended up with four stop series fully contrastive without reducing syllable complexity.

2.
- complex consonants can derive from clusters, but the same clusters may or may not turn into complex clusters depending on context. So, for instance, /kp/ may become /kp)/ initially or as codas, but remain /kp/ intervocalically - and then say that final vowels are lost. That moves us from /akp/, /akpka/ and /akpa/ to /akp)/, /akp)k/ and /akp/ all distinguished.

3. complex consonants can derive from clusters, which can result from vowel reduction. Vowel reduction can increase the ambiguity of the word. Ambiguity can be decreased through the creation of compounds. The creation of compounds can re-introduce clusters. So a process that produces complex consonants will tend also to produce clusters down the line.
Eg: unstressed vowels drop. This turns both "kapata" and "kopato" into "kpat". This cluster 'simplifies' into /kp)at/. To disambiguate the two meanings, speakers add the clarifying adjectives "d_wik" and "k'at", giving us words like /kp)atdwik/ and /kp)atk'at/, with both large inventories and complex syllables


Obviously, these three explanations can work in tandem with one another.


-----

A good comparison to Palauan is Tai-Kadai. According to the Austro-Tai hypothesis, Tai-Kadai, like Palauan, is derived from Proto-Austronesian (or pre-Proto-Austronesian) via largescale reduction of initial vowels. Most of the resulting languages have both complex syllables and complex consonants, while Austronesian languages, without that reduction, have mostly retained small inventories and simple syllables.

See Naxi for an example of large inventory vs simple syllables:
http://en.wikipedia.org/wiki/Naxi_language#Consonants

There's also probably a couple languages like that in the nilo-saharan family too (the ones with 4 level tones and almost every tone glide combination).

I like Sal's ideas, and your #1 has made me think of the southeastern dialects of Ojibwe (Odawa and Eastern Ojibwe, aka Nishnaabemwin), which have a fairly small inventory but exceedingly complex phonological behaviors (loss of all unstressed vowels, with consequent complex clusters). What specifically I find interesting is that in some subdialects, this creation of new clusters has also led to an increase in phonemes: when /o/ is lost between a velar stop and a following consonant, the velar stop becomes labiovelar, and contrasts in that position with plain velar stops. So hypothetical* aakona / aakana --> aakwna / aakna. So here the creation of new clusters leads to the creation of new phonemic contrasts.


*I can't think of any good real examples off the top of my head...

Whimemsz wrote:What specifically I find interesting is that in some subdialects, this creation of new clusters has also led to an increase in phonemes: when /o/ is lost between a velar stop and a following consonant, the velar stop becomes labiovelar, and contrasts in that position with plain velar stops. So hypothetical* aakona / aakana --> aakwna / aakna. So here the creation of new clusters leads to the creation of new phonemic contrasts.

The same thing happened in Seri (which is why they use C+o-umlaut to write labialized consonants). And it reminds me of the Slavic langs, which went from not allowing closed syllables to (C)(C)(C)(C)V(C)(C)(C)(C) or whatever it is now, and picked up a full series of palatalized consonants along the way.

Hell, I'd guess that might explain some of the correlation: langs with complex syllable structures either pick up phonemes in the process of getting complex syllable structures or simplify the resulting clusters.

Something I thought about recently due to a conlang:

Is it true that CV syllable language change slowly? Ie they undergo fewer soundchanges per unit time.

Example: Polynesian languages (Maori, Samoan, Niuea, Tongan, Fijian, Hawaiian) split 2,000 years ago - before the Romance languages did - but have undergone far fewer sound changes and diverged phonetically much less. Maori has been isolated from the others for about a thousand years, but still hasn't changed very much. Look at the Proto-Polynesian to Hawaiian sound changes in the Correspondences thread, and then compare with the ones from Vulgar Latin to any Modern Romance language. The difference is quite large.

Are there counter examples to this?

It could just be that Polynesian is just a relatively conservative family. (IE has one of those too: Baltic.) There are some pretty interesting Austronesian langs out there: Malagasy, Nias, and Javanese are the three that I can think of, but there are probably a lot more.

Or am I getting the time scales totally wrong here?

Nortaneous wrote:It could just be that Polynesian is just a relatively conservative family. (IE has one of those too: Baltic.) There are some pretty interesting Austronesian langs out there: Malagasy, Nias, and Javanese are the three that I can think of, but there are probably a lot more.

Or am I getting the time scales totally wrong here?

That is possible. I don't know the time scales for the non-polynesian Austronesian langs; I did a quick wikipedia check but can't find it, and I don't have enough time now.

But one big difference with Baltic surely is scale - the entire polynesian language family, covering a lot of ocean, has had very few sound changes. Maori for example has been isolated from the rest for 1,000 years, and in that thousand changed so little it is still intelligible to quite a large extent with other varieties of polynesian. Saying "polynesian is a relatively conservative family" means very little - that's just an observation. We can see it is conservative, yes, the question is why? If it is just coincidence and unrelated to its CV phonology, then why do all of it's languages have this feature, even when separated for long periods of time?