LINGUIST List 15.658

Sat Feb 21 2004

Diss: Modified Issue 15.648: Phonology: Kabak

Editor for this issue: Tomoko Okuno <>


  1. Baris.Kabak, The Perceptual Processing of Second Language Consonant Clusters

Message 1: The Perceptual Processing of Second Language Consonant Clusters

Date: Sat, 21 Feb 2004 14:32:59 -0500 (EST)
From: Baris.Kabak <>
Subject: The Perceptual Processing of Second Language Consonant Clusters

Institution: University of Delaware
Program: Department of Linguistics
Dissertation Status: Completed
Degree Date: 2003

Author: Baris Kabak 

Dissertation Title: The Perceptual Processing of Second Language
Consonant Clusters

Linguistic Field: Phonetics, Phonology, Psycholinguistics 

Subject Language: English (code: ENG ) Korean (code: KKN )

Dissertation Director 1: William J. Idsardi

Dissertation Abstract: 

Listeners perceive epenthetic vowels within consonant clusters that
violate the phonotactics of their first language (L1) (Dupoux et al.,
1999). The present study tests two approaches towards phonotactics for
their predictions about perceptual epenthesis (PE) effects. While the
string-based approach predicts PE due to consonantal contact
restrictions in the L1, the syllable-based approach predicts PE as a
result of L1 syllable structure conditions. These two approaches are
tested with Korean ESL speakers whose L1 bans certain consonants in
coda position (e.g., *[c], *[g]) while allowing others (e.g., [k],
[l]). Korean also disallows certain heterosyllabic contacts such as
*[km] and *[ln], which are realized as [nm] and [ll] due to
nasalization and lateralization, respectively. In a perceptual
experiment, clusters that contain (1) coda violations (e.g., *[cm],
*[gm]), and (2) contact violations (e.g., *[km], *[ln]) in Korean are
used in nonce pairs comparing these clusters with their vowel-present
counterparts (e.g., *[phakma] vs. [phakuma]), and also to their likely
output forms (e.g., *[phakma] vs. [phaNma]). The results from a
discrimination experiment using English and Korean listeners indicate
that the English group successfully discriminates all clusters. The
Korean group is also successful in most clusters except for those
where the cluster incurs a coda violation (e.g., *[c.m],
*[j.t]). Finally, the Korean group does not confuse illicit clusters
with their likely output forms. These results show that PE arises when
there is a syllable structure violation, rather than a contact
violation, confirming the syllable-based approach. The successful
discrimination of clusters with voiced codas (e.g., *[gt]) suggests
that voicing is perceptually suppressed leading to a licit coda (i.e.,
[g.t] as /k.t/). That illicit clusters are not misperceived as their
likely output forms indicates that assimilation rules are not relevant
for perception. Therefore, theories that explain phonological
alterations based on perceived similarity between input and output
forms (e.g., Steriade, 2001b) cannot account for Korean listeners'
perception. Finally, frequency cannot explain present findings
because, despite their zero frequency in Korean, only certain illicit
clusters induce PE. Instead, a perceptual model using onset and coda
detectors is proposed and is shown to account for PE effects in a
straightforward way.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue