Yesterday I was looking for examples of synchronic processes (e.g. vowel epenthesis) driven by a clear and simple syllable contact law, and struggling a bit. This was partly motivated by this paper, referenced in a previous blog post, which suggests that the distribution of clusters in proto-Algonquian was motivated by such a law. It seems that illicit clusters were repaired either by consonant deletion or vowel insertion.

The syllable contact law suggests that languages for the final consonant of a preceding syllable to be more sonorous than the initial consonant of the next syllable. This makes the permissible or common heterosyllabic clusters more complex than just the combination of all legal codas and all legal onsets. Proto-Algonquian, for example, seems to have mostly permitted sonorant codas before stops and fricatives, but fricative codas only before stops.

So this raises an interesting question: are then any examples from modern languages with regular rules motivated by a strong, unified version of the syllable contact law / sonority hierarchy across syllable boundaries, and not just by sonority constraints within the syllables themselves, by more restrictive / ad-hoc cluster breaking rules, or where the SCL is merely a tendency created by historical sound changes? Attempting to find examples has been harder than I thought at first.

Sound Changes and Phoneme Distributions

As Blevins points out in Evolutionary Phonology, there are plenty of frequent sound changes which weaken coda consonants, and which can create phoneme distribution patterns in some languages reminiscent of the SCL. But I’m specifically interested in languages where the SCL has the status of an active phonological rule which applies to morphological processes like compounding, not just the structure of individual morphemes in isolation.

Possible Success: Consonant Fortition in Turkic

Davis suggests that the SCL is an active constraint in Kazakh. The examples he gives are from clusters created by suffixation. When a suffix is added which creates a cluster of ascending sonority, the second consonant is fortified to a stop. Here are some of the examples given of the same suffix with different preceding consonants (I think the vowel changes are due to vowel harmony): alma-lar BUT murin-dar, kol-ma BUT kes-pe. But there are alternative analyses for this which seem compelling because they account for simple patterning of initial consonants, which cannot be explained by the SCL.

Partial: Epenthesis

So far, I haven’t found any mention of comprehensive enforcement of the SCL via epenthesis. The examples I know of languages with widespread vowel epenthesis don’t seem to work quite this way. Most seem like Halh Mongolian, which is sensitive to sonority within the syllable, but increasing sonority across a syllable boundary doesn’t seem to be an issue. The syllabification algorithm produces outcomes like /nɔirstʰ.ɮɔ/ and /xa.rais.nas/ which would be ruled out by any hard SCL constraint.

A few languages have active rules relating to sonorants in particular:

• In Mohawk, /e/ is inserted to break up clusters where the second element is a sonorant /n/, /r/, /w/
• In Winnebago, if there is a voiceless obstruent followed by a sonorant /r/, /n/, or /w/, the vowel following the sonorant is copied between the two consonants. This is called Dorsey’s Law.

Other languages with epenthesis have rules related to stops and/or sounds with a clear release in particular. In Kalam, according to Blevins and Pawley, a cyclic epenthesis rule breaks up all clusters within words where the first consonant has a release (including stops, nasals etc.). The only clusters not broken up are intervocalic clusters that begin with a continuant: /s/, /j/ or /w/. Compare: /wlk/ = [wulɨk] with obligatory epenthetic vowel and no internal clusters, and /as-ket/ = [asker], [asɨɤer] with only optional ɨ insertion.

The reason that stops especially tend to favour pre-vocalic position and induce vowel insertion is almost certainly because (as Blevins states) their release is important to identify them, and so will be present even if not pre-vocal, and that release is easy to mistake for a short vowel of some description. This is also a possible explanation for Dorsey’s Law: the release of the preceding obstruent was reinterpreted as a copy vowel and turned into a phonological rule instead of a phonetic detail.

This seems to make intermediate sonority consonants the issue. It’s easy to find languages which systematically insert epenthetic vowels if the first element of a heterosyllabic cluster is from the class of lowest sonority consonants (stops, affricates), and possible to find languages which systematically insert epenthetic vowels if the second element of a heterosyllabic cluster is a sonorant. But what about a language which show epenthesis if an intermediate sonority consonant is part of a rising sonority cluster, but not if it’s part of a falling sonority cluster?

To pick a simple three class case:

First / Second	Stop / Affricate	Fricative	Sonorant
Stop / Affricate	Epenthesis ✓	Epenthesis ✓	Epenthesis ✓
Fricative		Epenthesis ????	Epenthesis ✓
Sonorant			Epenthesis ✓

The stop first and sonorant second epenthesis rules are simple and consistent. Note that the top row is completely filled, and the final column is completely filled. But the intermediate class of consonants needs to show switching behaviour to be sure we’re dealing with a true synchronic SCL rule. I’ve not found a language yet with this switching behaviour as a phonological rule in combination with the other rules, and in combination with enough classes to be sure we’re dealing with a sonority hierarchy.

There are some suggestions that epenthetic vowel placement in loanwords might count in some languages, but loanword phonology has its own complexities.

Does the SCL Exist?

I was expecting to find stronger evidence for the SCL as a rule in some language when I started looking. The fact that the examples provided are mostly weak and not very systematic makes it unclear if the SCL exists as an active rule driven by a sonority hierarchy in any language. It may well exist as a phonotactic tendency driven by universal phonetic factors and sound change, and some of those changes may be generalised into active rules that cover some (although generally not all) of the SCL space. But it’s not clear to me that the SCL is more than an epiphenomenon, and it does seem pretty hard to find a language with a strong SCL rule creating productive synchronic alternations.