Syllables

[chris_notts]

A lot of languages allow bigger clusters at word edges than in the middle of (mono-morphemic) words. This falls out simply from various common sound changes such as the deletion of final vowels, or syncope of vowels between consonants occurring before or after a more prominent (e.g. stressed) syllable. Similarly, deletion of some initial consonants or simplification of internal vowel sequences can produce a language which permits onset-less syllables word initially but nowhere else (this is quite common).

There are often questions, especially when the clusters are large, about whether the clusters are really tautosyllabic. For a lot of languages with large initial or final clusters, people talk about appendices, by which they mean consonants which aren't really attached to a proper syllable. IIRC, 'Appendices' are also sometimes offered as an explanation when a coda in a final syllable doesn't produce some expected effect. For example, in many languages there are weight restrictions such that the syllable types CVV and CVC are both possible, but CVVC (super heavy) isn't. In a number of such languages, the final syllable is an exception to this rule.

Some languages also have strongly preferred templates for word or root shapes, which may force different syllable shapes in different locations. PIE seems to have been an example of this, since there are numerous restrictions on root shape that most roots obey, although there are a few exceptions.

[finlay]

German: Herkunftswörterbuch (4C-1C; all-native for 'etymologic dictionary')

Edit: herkunftsgleich (4C-2C; 'being of the same origin') has even one more German allows up to 4 consonants in codas, and up to 3 in onsets.

whenever i try to say that i go 'nfst'...

[TomHChappell]

Polish words can easily start with 4 consonants, e.g. pstry and there can be 4 consonants between vowels, e.g. abstrakt or even 5, e.g. skąpstwo. There is actually a 5-consonant onset, as mpstw-, but I can't lay my hands on an example.

I guess it wasn't Polish I read about. I wonder what it was?

... (other good stuff snipped out).... For example, in many languages there are weight restrictions such that the syllable types CVV and CVC are both possible, but CVVC (super heavy) isn't. In a number of such languages, the final syllable is an exception to this rule.
Some languages also have strongly preferred templates for word or root shapes, which may force different syllable shapes in different locations. PIE seems to have been an example of this, since there are numerous restrictions on root shape that most roots obey, although there are a few exceptions.

What Chris (and others) said, and:
According to the STRESSTYP database, languages with weight-sensitive stress that have three weight-classes -- namely, light, heavy, and superheavy -- typically have a rule that in any polysyllabic word with a superheavy syllable the primary stress always goes on a superheavy syllable. Since no word has more than one primarily-stressed syllable, a corollary is that no word has more than one superheavy syllable.

Note, though, that what constitutes "heavy" varies from language to language. For instance in at least one language in their database a syllable is heavy only if it is doubly-closed, that is, if its coda contains a cluster.

[Grigor]

Does anybody know if any natural language disallows onset clusters but allow coda clusters?

[chris_notts]

According to the STRESSTYP database, languages with weight-sensitive stress that have three weight-classes -- namely, light, heavy, and superheavy -- typically have a rule that in any polysyllabic word with a superheavy syllable the primary stress always goes on a superheavy syllable. Since no word has more than one primarily-stressed syllable, a corollary is that no word has more than one superheavy syllable.

What happens in these languages when compounding or inflection produces a word that should have two superheavy syllables? I would assume that the vowel in one of them is shortened, since that seems like the most likely strategy to produce a valid word.

Note, though, that what constitutes "heavy" varies from language to language. For instance in at least one language in their database a syllable is heavy only if it is doubly-closed, that is, if its coda contains a cluster.

Not just that. Languages often disagree about whether only CVV is heavy, or whether CVC is heavy, or both. I read an interesting paper that claimed that this correlated (imperfectly) with the most common coda consonants - it claimed that if CVC is treated as heavy, then the language tends to have mostly sonorant codas, whereas if CVC is light the language tends to have lots of obstruent codas. If true, that would mean that languages tend to prefer that heavy syllables contain long sonority peaks.

Also, different weight sensitive phenomena in the language don't always agree about what is heavy and what is light.

EDIT: it also raises an interesting question - are there any languages for which the coda consonant determines the syllable weight? E.g. syllables with coda nasals are heavy, but with coda stops are light. IIRC, in Piraha the onset consonant type contributes to the syllable weight.

[TomHChappell]

Does anybody know if any natural language disallows onset clusters but allow coda clusters?

I went to http://www.unileiden.net/stresstyp/form2b.htm and entered "Syllable_Template Contains (C)V(C)(C) and Syllable_Template Is not equal to (C)(C)V(C)(C)" and found a few.
According to STRESSTYP
78. Burarra; Barera Australian, Burarran.
320. Ngalakan Australian, Gunwingguan, Ngalakanic.
365. Quechan; Yuma; Kwitchyan Hokan, Esselen-Yuman, Yuman, River Yuman.
482. Weri; Were Trans-New Guinea, Main Section, Eastern, Central and Southeastern, Goilalan, Kunimaipa.

and maybe 287. Menomini Algic, Algonquian, Central.

Now since I left out V(V) and V(:) nuclei, and left out mandatory onsets, so stuff like CV(V)(C)(C) and CV(:)(C)(C), that can't be all.

So you should also look for
136. Gaalpu Dhangu Australian, Pama-Nyungan, Yuulngu.
181. Jacalteco Mayan, Kanjobalan-Chujean, Kanjobalan, Kanjobal-Jacaltec.
318. Nez Perce Penutian, Plateau Penutian, Sahaptin.

and

186. Juwalarai; Yuwaalaraay (*) Yualyai Australian - Pama-Nyungan - Wiradjuric
234. Laragia; Larakia Australian, Laragiyan.

and that still, probably, won't be all.

... languages ... that have three weight-classes -- namely, light, heavy, and superheavy -- typically have a rule that ... no word has more than one superheavy syllable.

What happens in these languages when compounding or inflection produces a word that should have two superheavy syllables? I would assume that the vowel in one of them is shortened, since that seems like the most likely strategy to produce a valid word.

That or something similar would be bound to be true, or at least so it seems to me. At any rate, one way or another, one of the superheavy syllables would have to be reduced to merely "heavy", or the "compound word" would have to remain a phrase instead of a "word". The way you have mentioned is certainly plausible; at first thought it's not clear to me that it's the only plausible way, but I haven't thought too hard.

Not just that. Languages often disagree about whether only CVV is heavy, or whether CVC is heavy, or both.

Yes.
If a natlang has weight-sensitive stress and/or weight-sensitive rhythm, and has no closed syllables (i.e. no codas) but doesn have diphthongal nuclei or phonemic length for vowels, then it won't have to worry about whether CVC syllables are light or heavy because it won't have any CVC syllables; but it will make CV syllables light and make CV: and/or CVV syllables heavy.
If a natlang has weight-sensitive stress and/or weight-sensitive rhythm, and has no phonemic length for vowels and no diphthongal nuclei but does have closed syllables as well as open syllables, then it won't have to worry about whether CV: or CVV syllables are light or heavy because it won't have any CV: or CVV syllables; but it will make CV syllables light and make at least some CVC syllables heavy.
If such a language allows non-sonorant codas as well as sonorant codas, it will make CVR syllables heavy (R stands for a Sonorant), but some of them make CVB syllables (B stands for a non-Sonorant -- e.g. in some languages some obstruents) light.

I read an interesting paper that claimed that this correlated (imperfectly) with the most common coda consonants - it claimed that if CVC is treated as heavy, then the language tends to have mostly sonorant codas, whereas if CVC is light the language tends to have lots of obstruent codas.

I don't know. But, if the language has no V: nor VV nuclei, and if all of the possible codas are sonorant, then all of the CVC syllables will be heavy; while if the language has no V: nor VV nuclei and all of the possible codas are obstruents, then still all of the CVC syllables will be heavy. If it has no V: nor VV nuclei and some CVC syllables are heavy but some CVC syllables are light, then the heavy CVC syllables will have sonorant codas and the light CVC syllables' codas won't be sonorant.

If true, that would mean that languages tend to prefer that heavy syllables contain long sonority peaks.

That is true. In languages that make the light vs heavy distinction, light syllables take less time to say than heavy syllables. That doesn't mean every heavy syllable takes exactly twice as long to say as every light syllable; just that, in general, when two syllables are near each other and one is light and one is heavy, then, ignoring any difference caused by differing stress or differing lexical tone, odds are very high the heavy one is going to take longer to say than the light one.

Also, different weight sensitive phenomena in the language don't always agree about what is heavy and what is light.

Exactly so; that's exactly why the STRESSTYP database has a "Heavy_for_Stress" field and a "Heavy_for_Rhythm" field and they don't have to have the same value.

... an interesting question - are there any languages for which the coda consonant determines the syllable weight? E.g. syllables with coda nasals are heavy, but with coda stops are light.

I believe both WALS and STRESSTYP say that there are such natlangs.

IIRC, in Piraha the onset consonant type contributes to the syllable weight.

305. Múra-Pirahã; Pirahã Mura in STRESSTYP has a hierarchy of about five three weight-classes; KVV > GVV > VV > KV > GV. All syllables are open; there are no onset clusters; nuclei may be monophthongs or diphthongs (or could that be "short or long"?); and a syllable with a monophthongal nucleus must have an onset. "G" represents a voiced onset and "K" represents an unvoiced onset. Primary stress falls on the heaviest of the last three syllables in the word; if two (or all three) of those last three syllables tie for heaviest, primary stress falls on the later (or last) one of those tied-for-heaviest last three syllables. STRESSTYP didn't have any information to record about rhythm or secondary stress, the last time they entered information about Pirahã.

It's rarish for the onset to have any influence on the weight; but it's not at all unknown. According to WALS and STRESSTYP, that is. Naturally I personally haven't yet gotten around to learning all the languages there are. Actually I'm sort of pretty far behind, even if I only count living languages with satisfactory reference-grammars published. "One down, one-thousand-nine-hundred-ninety-nine to go" sort of sums it up.

[Richard W]

Does anybody know if any natural language disallows onset clusters but allow coda clusters?

... and that still, probably, won't be all.

I thought most languages disallowed onset clusters. So, with no onset clusters but coda clusters we also have:

Hungarian
Most (all?) modern Arabic dialects
Biblical Hebrew (except for one word, stáyim f. '2')
Farsi (Daric might retain a few clusters)
Turkish

[Tropylium⁺]

I thought most languages disallowed onset clusters. So, with no onset clusters but coda clusters we also have:

Hungarian
(…)
Turkish

Most Uralic languages from Mari eastwards, really. IIUC also most Turkic and Mongolic languages.

German: Herkunftswörterbuch (4C-1C; all-native for 'etymologic dictionary')

Edit: herkunftsgleich (4C-2C; 'being of the same origin') has even one more German allows up to 4 consonants in codas, and up to 3 in onsets.

Ernststraße (a reasonably common street name, based on the name Ernst, which can be used both as a male personal name or as a surname) has the full set of 4+3 consonants: /ɛʁnst.ʃtʁaːsə/

Speaking of pathological examples, I've been told Swedish allows -CCCCCC, in västkustskst "the most west-coastly".

[Jipí]

German would be westküstischst, and German gets [ʃ] <sch> from earlier [sk]. Why isn't it västkustisk, though?

[TomHChappell]

I thought most languages disallowed onset clusters.

According to WALS.info, the most-common type of syllable template is more like
(C1)(C2)V(C3) with very restricted C2.
So apparently more than half of languages -- at least, about 56% of the languages in their database -- allow two-consonant onset clusters and one-consonant codas, but don't allow coda clusters nor three-or-more-consonant tautosyllabic clusters.

A slightly more elaborate syllable structure would add another consonant, either in the final position of the syllable or at its beginning, giving the structures CVC and CCV; these are both modest expansions of the simple CV syllable type. But it is worthwhile to make a distinction between two types of two-consonant strings. In a very large number of languages, although two consonants are allowed in the onset position of a syllable, there are strict limits on what kinds of combinations are permitted. The second of two consonants is commonly limited to being one of a small set belonging to either the class of “liquids” or the class of “glides”. The liquids are the sounds commonly represented by the letters r and l , while glides are vowel-like consonants such as those at the beginning of the English words wet and yet . Liquids and glides have in common that they are produced with a configuration of the speech organs which permits a relatively unobstructed flow of air out of the mouth. Languages which permit a single consonant after the vowel and/or allow two consonants to occur before the vowel, but obey a limitation to only the common two-consonant patterns described above, are counted as having moderately complex syllable structure . An example is Darai (Indo-Aryan; Nepal). Here the most elaborate syllable permitted is CCVC, as in /bwak/ ‘(his) father’, but the only possible second consonant in a sequence of two is /w/.