[Author Login]
Title:Onset Cluster Typologies
Authors:Steve Parker, Stephen Jay
Comment:published at https://www.diu.edu/academics/opal/
Length:10 pages
Abstract:Most formal models of phonology invoke sonority to explain the set of consonant clusters allowed in specific languages. While such proposals account for many languages, they ultimately prove too weak when confronted with more exhaustive data sets. Part of the problem is that theories of onset cluster typology have not been tested against inventories from hundreds of languages. When larger samples are considered, formal models fail to accommodate the attested combinations of syllable-initial consonant clusters (σ[C1C2V]) in many languages.

Previous approaches to the phonotactics of consonant clusters include the following:

-MSD: Minimum Sonority Distance (Steriade 1982, Selkirk 1984, Levin 1985)
-SD: Sonority Dispersion (Clements 1990)
-SR: Sonority Rise (Flemming 2008)
-SA: Sonority Angle (Fullwood 2014)

Each of these models calculates the gradient harmony of consonant clusters in terms of a different mathematical formula based on the distance between the sonority indices of C1, C2, and sometimes the following vowel. However, none of them are able to handle the full range of empirical facts. For example, MSD favors obstruent+glide (OG) clusters, so it cannot produce a language where C2 must be a liquid without additional mechanisms. Conversely, SD evaluates obstruent+liquid onsets (OL) as unmarked, so it cannot generate languages where C2 is always a glide. All of these approaches are therefore partially right, yet partially wrong.

In this paper we highlight a new model called Minimum Distance to Offset (MDO). Like MSD approaches it sets a minimum threshold for sonority distance between C2 and C1 in each language. However, it further arranges onset cluster types into several continua with a constant C2 in each scale. For example, the glide offset continuum consists of a fixed ranking encoding the universal preference OG ≻ NG ≻ LG ≻ GG.

Each of these competing models is compared against a database listing the inventory of permissible onset clusters in over 300 languages worldwide. The results indicate that previous approaches massively undergenerate the attested typological combinations and are therefore too restrictive. In contrast to this, the MDO proposal successfully accounts for the full array of language types. While the flexibility inherent in the MDO model is necessary, we argue that it is not overly powerful. We note many hypothetical language types it cannot produce.

We conclude by showing how the preference scales of the MDO approach can be implemented using markedness constraint families in a stringency relationship (de Lacy 2004, 2006). For instance, the following constraints target increasingly more inclusive combinations of onset clusters, with C2 fixed as a glide: *{GG}, *{GG,LG}, *{GG,LG,NG}, *{GG,LG,NG,OG}. This subset of the model allows for languages in which C2 can only be a liquid, nasal, obstruent, or combination thereof.
Type:Paper/tech report
Area/Keywords:phonology, onset, consonant clusters, sonority, minimum distance, typology
Article:Version 1