LINGUIST List 20.3119

Wed Sep 16 2009

Qs: Empirical Resources on Consonant Cluster Typology

Editor for this issue: Elyssa Winzeler <>

        1.    Steve Parker, Empirical Resources on Consonant Cluster Typology

Message 1: Empirical Resources on Consonant Cluster Typology
Date: 15-Sep-2009
From: Steve Parker <>
Subject: Empirical Resources on Consonant Cluster Typology
E-mail this message to a friend

I am looking for input (data) on tautosyllabic consonant clusters. Supposethat a syllable begins with two adjacent consonants, followed by a vowel:CCV. Technically this is called an initial demisyllable. I am aware of twocompeting claims/proposals about what kinds of consonants arecross-linguistically unmarked or preferred in this type of situation, bothbased on the notion of relative sonority. For the sake of simplicity, letus assume a common five-way sonority scale:

V (vowel)G (glide)L (liquid)N (nasal)O (obstruent)

(1) One approach posits that specific languages can place a minimumsonority distance requirement on onset clusters, such that Spanish, forexample, allows OL but not *ON or *NL, since obstruents differ from liquidsby two steps on the sonority scale, whereas obstruents plus nasals ornasals plus liquids differ by only one sonority rank. One implication ofthis is that there could or should exist languages in which the onlypermissible onset clusters consist of an obstruent followed by a glide,such as /py/, /kw/, etc., whereas OL onsets, such as /pl/ or /tr/, are notattested. Works such as Steriade (1982) and Selkirk (1984) are examples ofthis general theory.

(2) A different approach is the Sonority Dispersion Principle proposed byClements (1990). In this theory the three segments in an initial CCVdemisyllable prefer to be evenly spaced apart in terms of relativesonority. This leads to the claim that OL (obstruent + liquid)syllable-initial clusters are universally preferred over OG (obstruent +glide). One implication of this is that there could or should existlanguages in which the only permissible onset clusters consist of anobstruent followed by a liquid, such as /pl/ or /tr/, whereas *OG onsets,such as /py/, /kw/, etc., systematically do not occur.

I am preparing to carry out a major cross-linguistic study in which I testthe claims of these two competing approaches on a robust sample oflanguages, preferably a set of languages which is genetically andgeographically balanced. Evidence for or against these two theories couldpotentially come from different areas of the phonology:

(1) inventory of attested syllable patterns(2) relative frequency of different types of syllable patterns(3) child language acquisition data(4) dynamic morphophonemic alternations

The latter, for example, would consist hypothetically of an underlyingcombination of morphemes which might otherwise be expected to surface as OL(obstruent + liquid), but which instead is realized phonetically as OG(obstruent + glide), or vice-versa. To illustrate, /pla/ > [pwa] or /kwa/ >[kra], etc.

The general research question which I am trying to tease apart is, whichtype of initial cluster, OL or OG, is truly unmarked in the languages ofthe world? My general impression at this point is that the answer to thisissue is mixed, with some languages showing a preference for OL, and othersindicating that OG is default. I think there are also other languages inwhich these two types of clusters are more or less evenly preferred.

What I am looking for is hard empirical and statistical evidence fromindividual languages, or even better yet from many languages, in responseto this dilemma. I would especially like to know if any published surveysor typological databases already exist which address this issue, or whichwould allow me to perform searches to answer these questions? In addition,I would be happy to hear about electronic dictionaries and/or text corporain relevant languages which would lend themselves to easily counting uniquewords (types) or tokens of forms containing such clusters (OL and/or OG).

By the way, I am aware of the difficult issue of interpreting OGVsequences, such as [kwa], in terms of whether the [w] is really in theonset or the nucleus, whether it is a separate consonant or justlabialization of the [k], etc. So I would especially value cases oflanguages in which there is a clear answer to these questions.

Thank you very much,

Steve ParkerGraduate Institute of Applied Linguistics and SIL International

Linguistic Field(s): Phonology                             Typology