UBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

grim [’07 202304

 

43““? A r. A
um i u [UU/

 

MIR 3 l 2008

 

032309

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRCJDateDuopes-pJS

 

A GESTURE-BASED ACCOUNT OF CONSONANT PATTERNS
IN KOREAN

By

Ok-Sook Park

A DISSERTATION

Submitted to
Michigan State University
in partial fiJlfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Linguistics and Germanic, Slavic, Asian, and African Languages

2002

ABSTRACT

A GESTURE-BASED ACCOUNT OF CONSONANT PATTERNS
IN KOREAN

By

Ok-Sook Park

By adopting basic tenets of Articulatory Phonology and Optimality Theory, this
study shows that seemingly unrelated consonant patterns in Korean can be combined and
treated as regular and related phenomena within a gestural approach

Consonant patterns such as Coda Neutralization (CN), Coda Consonant Cluster
Simplification (CCCS), Post-Obstruent Tensification (POT), Post-Nasal Tensification

(PNT), Post-Lateral Tensification (PLT), and Aspiration in Korean are investigated in
this study.

These consonant patterns are treated as separate, as exceptions, or as unrelated
phenomena in previous Korean literature. In addition, Korean linguistic studies (Cho
1990, Kim-Renaud 1974, SA. Jun 1993 and 1995, Rhee 1997, and Silva 1993) have
suggested implicitly and explicitly that various aspects of Korean consonant patterns
would benefit greatly from an approach that takes into account the interplay between
phonetics and phonology. However, few studies have been attempted that connect
phonetic characteristics to phonological explanations in explaining consonant patterns in
Korean. Therefore, this study examines the extent to which consonant patterns in Korean
are accounted for in terms of phonetic considerations in a given context.

One common phonological characteristic involved in CN, POT, PNT, PLT, and

Aspiration is neutralization in manner of articulation. Because manner of articulation is

A...

,.4n-

closely related to constriction degrees of articulators, CN, POT, PNT, PLT, Aspiration,
and other phonological phenomenon are analyzed using constraints sensitive to
constriction degrees in the vocal tract in a given context.

This study illustrates that a gesture-based approach, which is based on the
integration of phonetics and phonology, provides simplicity in explaining various aspects
of consonant patterns in Korean and makes it easy to grasp generalizations that lie behind
seemingly unrelated consonant patterns. Therefore, the analyses of consonant patterns in
terms of gestures provide unified and systematic explanatins for seemingly unrelated

consonant patterns in Korean. In addition, some idiosyncratic phonological phenomena

are better able to be explained and understood.

Copyright by
Ok-Sook Park
2002

ACKNOLWLEDGEMENTS

I would like to thank to those who have constantly shown their support and
encouragement in many ways. Among the many people, first, I would like to express my
gratitude to my academic advisor and chairperson, Professor Yen-Hwei Lin. I benefited
greatly from her valuable advice, comments and questions. Especially her academically
open—mind has encouraged me to not be afraid of talking about my ideas which are not
formed concretely yet, and has helped me in adopting the approach taken in this study. In
addition, I am thankful for her encouragement and patience even during times my work
was hardly progressing. Without her help, it would have been difficult to finish this

dissertation.

I am also grateful to the other members of my committee: Professor Alan Munn,
Professor Susan Gass, Professor Grover Hudson, and Professor Patricia Lunn. Their
comments, suggestions, and encouragement have been helpful in finishing this
dissertation. Especially Professor Hudson’s suggestions often helped make points clear
and allowed me to read my dissertation as a reader, not just as a writer.

I am also grateful to my other professors in Korea. Especially, Professor Sook
Hee Kim at Woosuk University, Professor Kum J a Chung and Professor Hae-Rhan Lee
who are both retired from Sookmyung Women’s university. Although there haven’t been

frequent contacts, their academic advice, support, and encouragement have been of great

help and provided motivation in continuing my study.

My thanks also goes to faculty and staff members associated with the Department
of Linguistics, the Center for the Enhancement of Linguistics Student Research, the
Center for Language Education And Research, and the Language Learning Center for
giving me the opportunity to work as a teaching or graduate assistant and providing
financial support to continue my education. In addition, I also want to express my thanks
to students that I have taught formally and informally. Their creative questions and
intellectual wonder have encouraged me to continue my studies with a sense of
responsibility.

Last, but certainly not least, I would like to thank my family and friends for
sharing their bottomless love and friendship in many different ways. My mother, sisters,
and brothers have encouraged me constantly no matter what the situation. The Moyers,
Margy and John Moyer, have also given me constant support and encouragement.
Especially, I am greatly indebted to Steve Moyer for being a patient listener, a critical
reader of my dissertation, and spending a lot of time in helping me to complete this

dissertation.

vi

TABLE OF CONTENTS

 

CHAPTER 1 THEORETICAL Issues

 

1.1 Introduction

1.2 Background and theoretical assumptions

 

 

1.2.1 Before Generative Phonology

 

1.2.2 Generative Phonology

1.2.2.1 Early Generative Phonology

 

 

1.2.2.2 Non-linear Phonology

 

1.2.3 Optimality Theory

 

1.2.4 Articulatory Phonology

1.2.4.1 The unit of representation in Articulatory Phonology

1.2.4.2 Distinctiveness and natural classes in Articulatory Phonology -------

 

 

1.3 Korean consonants

1.3.1 Phonological structure of Korean

1.3.2. Characteristics of Korean obstruents

1.3 .2. 1 Articulatory characteristics

1.3.2.2 Acoustic and aerodynamic characteristics

1.3.2.3 Constriction degree hierarchies

CHAPTER 2

CODA NEUTRALIZA'HON AND CODA CONSONANT CLUSTER SIMPLIFICATION ------------

 

 

 

 

 

 

2.1 Coda Neutralization

 

2.1.1 Introduction

 

2.1.2 Previous analyses

vii

12

22

23

36

36

39

4o

42

46

51

51

53

 

2.1.3 Patterns of obstruents in coda position 57

 

 

 

 

 

 

2.1.3.1 Patterns of stops in coda position 57
21.3.1.1 Introduction 57
2.1.3.1.2 Gestural characteristics of Korean stops 58
2.1.3.1.3 The gesture-based account of stops in coda position ----------- 62

2.1.3.2 Patterns of fricatives in coda position 73
2.1.3.2.] Introduction ' 73
2.1.3.2.2 Gestural characteristics of fricatives 75
2.1.3.2.3 Gestural characteristics of fricatives in coda position ---------- 77

2.1.3.2.4 Phonetic motivations for neutralization of fricatives

 

 

 

 

 

 

 

 

in coda position 81

2.1.3.25 The gesture-based account of fricatives in coda position mm 84

2.1.3.3 Patterns of affricates in coda position 95
2.1.3.3.] Introduction 95
2.1.3.3.2 Gestural characteristics of afi‘ricates 97
2.1.3.3.3 The gesture-based account of affn'cates in coda position ------ 99

2.1.4 Summary 103
2.2 Coda Consonant Cluster Simplification (CCC S) 107
2.2.1 Introduction 107
2.2.2 The gesture-based account of Coda Consonant Cluster Simplification ---- 112
2.2.2.1 The gesture-based account of CCCS in SK 112
2.2.2.2 The gesture-based account of CCCS in Kyung-Sang Korean ------- 127
2.2.3 Influences of non-surfaced consonants 133

 

 

2.2.4 Summary 137

viii

CHAPTER 3 POST-OBSTRUENT TENSIFICATION 140

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3.1 Introduction 140
3.2 The gesture-based account of Post-Obstruent Tensification 147
3.2.1 Introduction 147
3.2.2 Occurrence of Obstruent Tensification 147
3.2.3 Distribution of tense gestures and Post-Obstruent Tensification ----------- 151
3.2.4 Morphological conditions and distribution of tense gestures 160
3.2.5 The gesture-based account of Post-Obstruent Tensification 165
3. 3 Discussion 185
3.4 Summary 199
CHAPTER 4 OBSTRUENT PATTERNS AFTER SONORANTS AND /h/ 202
4.1 Introduction 202
4.2 Obstruent Tensification after sonorants 205
4.2.1 Introduction 205
4.2.2 The gesture-based account of tensification after sonorants 208
4.2.3 Summary 221
4.3 Patterns of /h/ and Obstruent sequences 223
4.3.1 Introduction 223
4.3.2 Previous studies 227
4.3.3 The gesture-based account of /h/ and Obstruent sequences 232
CHAPTER 5 CONCLUSION 243
REFERENCES 250

 

ix

Chapter 1 Theoretical Issues

1.1 Introduction

Traditionally, consonants are categorized based on both places and manners of
articulation. Manners of articulation classify sounds based on degrees of constriction in
the vocal tract. This study investigates Korean consonant patterns in terms of the degree
of constriction and presents a unified analysis of Korean consonant patterns within
Articulatory Phonology (Browman and Goldstein 1986 and 1990a).

The phonological patterns that are investigated in this study are known as Coda
Neutralization (CN), Coda Consonant Cluster Simplification (CCCS), Post-Obstruent
Tensification (POT), Post-Nasal Tensification (PNT), Post-Lateral Tensification (PLT),
and Aspiration or h—coalescence in Korean literature. CN deals with what consonants are
legitimate in coda position. CCCS refers to a phonological phenomenon in which a
consonant in a consonant cluster is selected as a legitimate consonant in coda position.
POT, PNT and PLT refer to phonological phenomena in which obstruents are tensed after
Obstruents, nasals, and the lateral /1/ respectively. POT is known to occur obligatorily
Within an accentual phrasel regardless of morpheme boundaries according to Jun (1993).

PNT and PLT are known to be sensitive to morphological conditions.

\

1 According to Jun 1993, an accentual phrase is determined by the syntactic and non-syntactic information
and composed of a word or words and forms an intonational phrase of an utterance. In the Seoul dialect,
also known as the standard dialect of Korean, an accentual phrase with one or two syllables has the rising
tonal pattern (Low (L) + High (1-1)) while an accentual phrase with three or more syllables has the rise-fall-
rise tonal pattern (LHLH). An accentual phrase is similar to the phonological phrase of Prosodic
Phonology (Selkirk 1986, Nespor and Vogel 1986, and Hayes 1989)

Aspiration or h—coalescence refers to a phenomenon in which sequences of /h/ +
/p, t, k/ or /p, t, k/ + /h/ are coalesced into aspirated obstruents, resulting in [ph, t", kh]
respectively.

In previous studies (Kim-Renaud 1974, Cho 1990 and Sohn 1987), these
phonological phenomena are treated with separate rules that are not relevant to each
other, or are treated to be irregular or exceptions. This study, by presenting an alternative
view of Korean consonant patterns, will show that these seemingly unrelated consonant
patterns are both systematic and related. By investigating Korean consonants in terms of
the degrees of constriction in the vocal tract, this study will lead to a better understanding
of Korean phonology and sound patterns.

Chapter 1 provides a brief history of linguists’ view of speech sounds,
frameworks of Optimality Theory and Articulatory Phonology, phonological structure of
Korean, and phonetic characteristics of Korean obstruents. Constriction degree
hierarchies for Korean obstruents are introduced at the end of chapter 1 as a basis of

analyses for this research. Chapter 2 investigates systematic principles that lie behind CN
and CCCS, suggests context-sensitive markedness constraints based on the principles,
and shows how the constraints interact with other constraints. Based on the context-
sensitive markedness constraints suggested in chapter 2, chapter 3 examines the phonetic
motivations for obstruents being tensed after obstruents (POT) and the relationship
between POT and acoustic or aerodynamic properties in (C)V1C1C2V2(C) contexts.
Chapter 4 investigates what phonetic and phonological motivations cause PNT and PLT

Sensitive to morphological or phonological conditions (4.2). The analyses of this

research show that there is no need to treat POT, PNT and PLT separately. POT, PNT,

 

and PLT result from interaction of the same phonetically grounded constraints. The
analysis of h-coalescence in Section 4.3 adds another example to the claim that the
overlap in inter-gestural timing between segments could cause neutralization (Steriade
1997: 76, Browman and Goldstein 1992, and Cho 1998a and 1998b). Chapter 5
concludes with implications of this study and possible directions of future research.

The analyses in this study are based on two main theoretical assumptions:
Articulatory Phonology which integrates phonetics and phonology (Browman and
Goldstein 1986, 1989, 1990a, 1990b, 1990c, 1991, 1992 and Browman 1994) and
Optimality Theory (Prince and Smolensky 1993 and McCarthy and Prince 1993a, 1994
and 1995). Articulatory Phonology is adopted for the units of representation and
Optimality Theory for relating an input to an output.

Throughout the history of speech studies, phonological theories have revolved
around two main questions: What is the unit of phonological representation? and how
does a phonological representation (an input) relate to a phonetic representation (an

output)? The following introduces the theoretical frameworks related to these two

questions and briefly describes the phonological structure of Korean consonants.

 

1.2 Background and theoretical assumptions

1.2.1 Before Generative Phonology

Prior to the late 19505 and early 19605, linguists were concerned with how speech
sounds are characterized articulatorily and acoustically (phonetics) and how sounds
function in a particular language (phonology). At this time, phoneticians classified speech
sounds on a physical basis: places and manners of articulation for consonants, with
consonants represented in columns and rows of a chart. The chart implies that there is a
one-to-one correlation between articulations and sounds. A one-to-one correlation
between articulations and sounds was developed into the concept that segments are
discrete, separable and perrnutable with an utterance being a linear concatenation of
segments. Thus, linguists paid attention to isolating the individual sounds rather than the
dynamics of speech.

After separating Speech into segments, linguists paid attention to how a sound (or
a segment) firnctions in a language, developing the concept of phonemes. For example, in

English, [p] in [peg] and [b] in [beg] make words differ in meaning. Thus, [p] and [b]

function as distinctive units (or phonemes) and are represented as /p/ and /b/ respectively.

In addition, if an English speaker pronounces [peg], [pheg] or [p’eg]', the word is ‘peg’

~k

1 . . . . . .
[p], [ph] and [p’]: Lax (or plain), aspirated and tensed vorceless bilabial stops respectively.

 

without differences in meaning. These sounds ([p] [ph] or [p’]) are considered as various
pronunciations for the phoneme /p/ in a given context in English. Thus /p/ (phoneme) is a
label for the set of [p, ph, p’] (allophones) in English.

In Korean, however, [p] and [b] do not make words different in meaning. A
Korean speaker can pronounce either [pap] or [bap] for the word /pap/ 'cooked rice'.
Both [p] and [b] are considered as one of various pronunciations of the phoneme /p/.
Thus, /p/ is a label thatis put on the set of [p, b] in Korean. /p/ is typically a voiceless
bilabial stop in both English and Korean. However, phonologically /p/ represents
different sets of allophones in English and Korean (/p/ for [p, p“, p’] in English and /p/ for
[p, b] in Korean). Thus the phoneme becomes an abstract concept rather than

representing a sound in its own right.

1.2.2 Generative Phonology

1.2.2.1 Early Generative Phonology

Early Generative Phonology began during the late 19505 and 19605 with Morris
Halle’s Sound Panem of Russian (1959) and Chomsky and Halle’s Sound Pattern of
m (1968). Generative phonologists have focused on what Speakers and listeners
know about a language’s sound system. Early Generative Phonology differs from
previous phonologies (or classical phonologies) in several points.

First, speech is composed of two levels: phonological and phonetic
representations, which are mediated by rules of phonology. The phonological
representation of each lexical entry contains phonologically distinctive features, which
determine the phonetic form of an item in all contexts according to the rules of the sound
Patterning in a particular language (Chomsky and Halle 1968:297). Phonological
representation contains information that is not predictable or describable by phonological
rules and is expressed with horizontal and vertical dimensions. The horizontal dimension
Stands for the successive phonemes of a language and the vertical dimension stands for

names of distinctive features, as shown in (1).

 

(1) /a k i / ‘baby’ in Korean
E-back " -sonorant " [-back]
-hight -continuant
-anterior
-coronal
-spread glottis
— -constricted glottis-

 

 

The (surface) phonetic representation indicates degree along a physically defined
scale and may thus assume numerous coefficients, as determined by the rules of the
phonological component (Chomsky and Halle 1968: 297).

Second, in classical phonology, phonemes were minimal units of phonological
descriptions. However, phonological descriptions with phonemes as basic units had the
problems of being redundant and failed to recognize generalizations. In Korean, for
instance, /p/, /t/ and /k/ are voiced between two vowels. In classical phonology, three
separate rules are needed for this voicing phenomena as in (2b) because /p/, /t/ and /k/ are
the minimal units of phonological description. This fails to recognize the generalization
that a set of sounds (/p/ /t/ and /k/) patterns together by sharing common features
(“Consonant -voice, -continuant, -constricted glottis, -spread glottis]).

Early generative phonologists claimed that human beings perceive speech in terms
0f Class rather than individual sounds. The strategy of classification of sounds is a binary
princ3iple, namely the presence or absence of a feature. It was claimed that sounds
represented by the presence or absence of a feature reflect what we know about our
la‘r‘guages. Thus, distinctive features, which are expressed by a binary principle, are used

In Order to capture ‘natural’ classes, to generalize about occurring phenomena, and to

formulate predictions about the behavior of class members. Thus, the voicing of stops

between two vowels is stated with a rule (2a):

(2) a. Early Generative Phonology
[+consonant, -continuant, -constricted glottis, -spread glottis]—>[+voice] / V_V

b. Classical Phonology
/p/—)[b]/ V_V

/t/ —>[d]/ V_V
/k/ -—)[g]/ V_V

1.2.2.2 Non-linear Phonology

Since 1970, the linear view of phonological representation as shown in (1) has
been challenged by studies of tones and stress (W 00 1969 and Leben 1973, and Liberrnan
and Prince 1977). Studies on tones, stress and vowel harmony have shown no one-to-one
relationship between a segment and a feature. Instead, segment patterns are independent
from feature or tone patterns, resulting in a contour tone or mismatch between the number
0f tone bearing units and tones. Thus, in non-linear phonology, features or prosodic
Stmctlws are placed on separate tiers (Clements and Keyser 1983, Goldsmith 1976, and
Van der Hulst and Smith 1982).

The non-linear view of phonological representation further developed into Feature

GeOmetry theories (Clements 1985, Clements and Hume 1995, McCarthy 1988, and

Sagey 1986). The basic principle of Feature Geometry theories is that features are not
only placed in separate tiers but also organized in a hierarchical tree structure, as shown in
(3). For instance, laki/ in (1) is composed of [-coronal], [-coronal], and [+coronal] at the
coronal tier as shown in (3). Coronals, which are dominated by the place node, may
pattern together with labials or dorsals in phonological rules involving the place node.

This is because the labial, coronal and dorsal tiers are daughters of the place node.

(3) Based on Sagey (1986)

i /

{-consonant] [+consonant] {-consonant]
[+continuant] [- continuant] [+continuant]

Laryngeal ISupralaryngeal Laryngeal JSupralaryngeal LaryngeatSupralaryngeal

Iplace lplace place
[+V0ice] {-voice /]I [+voice
[-labial] [-labial [-labial]
[-coronal] [- coronal] [+coronal]
[+dorsal] [+dorsal] [-dorsal]

Feature Geometry theories explain why a group of features patterns together in
assimilation and deletion by expressing dependency among features. The marriage
betWeen Feature Geometry theories and the early Generative Phonology’s assumption
that predictable information is absent in phonological representations led to
Underspecification theories, in which not all nodes of a feature geometry are present

photlological representations (Steriade 1987, Kiparsky 1982a and Archangeli 1988).

Although our understanding of phonological representation has increased greatly

with the non-linear view of phonological descriptions, some basic principles of early
Generative Phonology have been questioned. First, phonologists relied heavily on a series
of ordered rules or rules and constraints in explaining phonological adjustments between
an input and its output. However, rules are known to be too powerful to produce all and
only surface forms of a language. For instance, there is nothing that prevents replacement
of V_V with _{#, C} in (2a), resulting in [+voice] stops in coda position. However,
having [+voice] stops is less common (or more marked) than having {-voice] stops in coda
position in natural languages (Lisker and Abramson 1964, Lombardi 1995, Ohala 1983).

A theory of “markedness” is implemented to prevent occurrence of unnatural
phonological alternation in natural languages. Usually “markedness” is expressed as
constraints such as “sequences of obstruents within the syllable must agree for voicing
(*Otvoice-orvoice, Gussenhoven and Jacobs 1998:51).” or “no voiced stops in coda
Position (*+voice/_{#, C}).” However, the rule-based approach with constraints is
known to have “duplication problems” according to Gussenhoven and Jacobs (1998: 50)
and Kenstowicz and Kisseberth (1977: 136). A rule that a voiced Obstruent becomes

voiceless after a voiceless Obstruent is necessary to explain [backs] from /ba:k-;/z ‘backs’

In English. Native speakers of English also should have a constraint *avoice-orvoice so

that they are aware that *[aebs] and *[aepz] are ill-fonned syllable structures while [oks]

2\
Examples are from Gussenhoven and Jacobs 1998:45-50.

10

all

miiisillri L'p LLJ

‘ox’ and [as-dz] ‘adz’ are well-formed syllable structures in English. The devoicing rule

and the constraint *orvoice-orvoice work for the same purpose in English grammar but are
expressed differently. In the 19905, Optimality Theory (Prince and Smolensky 1993 and
McCarthy and Prince 1993 and 1995) presents an alternative view on explaining
phonological adjustments between input and output. Section 1.2.3 presents a brief
overview of Optimality Theory.

Second, a binary principle is known to be too simple to connect to the physical
reality of speech. Ladefoged (1972) claims that although two sounds are represented with
[+/-X], the differences in these two sounds are not based simply on [X]. These two
sounds may differ in length, shape of vocal tract, target positions, and in other ways.
Thus, in the mapping between phonological and phonetic representations, distinctive
features that are based on the binary principle are not adequate as an explanatory device,
even though distinctive features are based on phonetic characteristics.

Third, most generative phonologists described isolated individual sounds and
treated them out of context without paying attention to the dynamics of speech as
classical phoneticians did. However, the traditional view of speech as the concatenation
0f di Screte, separable and permutable segments, as shown in (1) and (3), has been changed
by teChnological advances. Electromyography, for instance, allows linguists to study the
motor control of the vocal tract during speech. Technological advances enabled linguists
to realize that it is rarely possible to be sure about the beginning and ending of various

SourIds. Phoneticians and some phonologists started to become aware of how non-linear

11

and non-discrete speech is and became reluctant to treat speech sounds as ‘discrete units’
and ‘static phenomena’ as in Classical Phonology and early Generative Phonology.

With the view of speech as continuity, phonologists have suggested several
different approaches in order to overcome the problems of classical and early Generative
Phonology. Speech is viewed as a ‘dynamic’ rather than a ‘static’ phenomenon in these
approaches. These approaches are often referred to as phonetically motivated
phonology. Among the phonetically motivated approaches to Speech, Articulatory
Phonology of Browman and Goldstein (1989 and 1990a) is significant in unifying
phonological and phonetic descriptions. Section 1.2.4 provides a brief overview of

Articulatory Phonology as presented by Browman and Goldstein (1989 and 1990a).

1.2.3 Optimality Theory

Although several different approaches on how to represent phonological units
have been suggested, approaches on how to relate a phonological representation (input) to
a Phonetic representation (output) did not change significantly until the Optimality
Theory (OT) proposed by Prince and Smolensky (1993) and McCarthy and Prince (1993
and 1995). In derivational approaches, a grammar is composed of a series of rules or
rules and constraints (Chomsky and Halle 1968, Paradis 1988) while in OT, a grammar is

conlposed of output constraints.

12

 

There are three components in OT: lexicon, generator and evaluator. The lexicon
contains contrastive properties of morphemes and provides input specifications. The
generator generates output candidates based on an input. The evaluator evaluates output
candidates against a set of ranked output constraints and selects one candidate as the most
optimal output (‘3’), as shown in (4). Using hypothetical languages, the process of

selecting the most optimal candidate among output candidates is explained in (7).

 

 

 

 

 

(4)
Leixicon Evaluator
A
f \
V Input Constraint 1 Constraint 2 Constraint 3
Generator Candidate 1 *
Candidate 2 *
C" Candidate 3 *

 

 

 

 

 

* : A critical violation.

One of the main issues that arises in OT is what an input Should be. Constraints
such as Morpheme Structure Condition (Kiparsky 1982b) are abandoned on the
assumption that no constraints are imposed on the input (Richness of Base) in OT. As
there are no constraints on the input, many inputs could be produced. However, based
on the acquisition perspective, OT generally accepts the assumption that in the absence
0f Eimpifical evidence for one input over another, the input should be the one that is
closest to the output (Lexical Optimization). (For further discussion on what an input
should be, see Prince and Smolensky 1993: 191-96, Archangeli and Langendoen 1997: 13,

Kager 1999: 19-34 and Ito, Mester and Padgett 1995).

13

 

Using hypothetical languages, L1 and L2, the process of selecting the most
optimal candidate among ouput candidates are explained. In Ll, stops are voiced between
two vowels and are voiceless in all other contexts as shown in (5a). In L2, there are no
voiceless stops between two vowels while the contrast between [+/-voice] stops is

maintained in outputs, as in (5b).

(5) a. L1 b. L2
[+voice] {-voice] [+voice] [-voice]
[aha] [Ra] [an] [aha] [ha] [ah] [pa] [an]
[aria] [ta] [at] [ada] Ida] [ad] [ta] [at]
[383] [ka] [3.1:] [383] [gal [a2] [ital [als]

First, it is necessary to identify a set of relevant constraints. All languages use
some voiceless stops while only a subset of languages uses voiced stops. Voiced stops
are more marked than voiceless stops (Maddieson 1984). (6a) represents a context-free
markedness constraint for the avoidance of voiced stops in languages.

HOWever, in some languages where no words are contrasted by [+voice] and [-
VOiced] Stops, there are cases when {-voiced] stops are voiced between two vowels (ex.
Korean)_ Thus, voiceless stops are not allowed between two vowels. (6b) represents the
disallowance of voiceless stops between two vowels.

In addition, there is a force in the grammar of a language that demands that features
in mm“ and output segments be identical. (6c) demands identity of correspondence

Segments in terms of [+/-voice].

14

(6) a. *[+voice]: Stops must not be voiced.
*[-voice]N_V: No voiceless stops between two vowels.

c. IDENT-IO[+/-voice]:
Correspondent segments are identical in feature [+/-voice].

IfIDENT-IO[+/-voice] is ranked higher than *[+voice] in a language, voiceless and
voiced stops are contrastive as satisfaction of IDENT-IO[+/-voice] takes precedence over
satisfaction of *[+voice]. However, no words are contrastive by [+/-voice] in L1. Thus,

IDENT-IO[+/-voice] is ranked lower than *[+voice] in L1 (*[+voice] >> IDENT-IO[+/-
voice]3).

The constraints, *[+voice] and *[-voice] /V_V, are in conflict. The constraint
*[+voice] demands that stops not be voiced. *[-voice] /V_V demands that stops be

voiced between two vowels. If *[+voice] is ranked higher than *[-voice] /V_V, no words

are voiced between two vowels. The undominated *[+voice] has to be satisfied first

before satisfaction of *[-voice] /V_V. In L], voiced stops appear between two vowels
but not in other contexts. Thus, satisfaction of *[-voice] /V_V takes precedence over
Satisfaction of *[+voice], resulting in the ranking hierarchy *[-voice] [V_V over *[+voice].
By transitivity, *[voice]/V_V is ranked higher than IDENT-IO[+/-voice] in L1 as shown in

(7)

 

3\
A>>~B2 A is higher than (dominates) B in a ranking hierarchy

15

 

(7) L1: *[-voice] /V_V >> *[+voice] >> IDENT—IO[+/—voice]

 

 
       

 

l*l*l
!: A critical violation. Shaded cells: Constraints not relevant to select the most optimal
candidates. A constraint in a left-hand column dominates constraints to its right.

 

 

 

 

 

 

 

 

 

 

 

In (7A), (7Aa) violates the undominated constraint *[-voice] N_V because of [p]
betVveen vowels. (7Ab) violates the context-free markedness constraint *[+voice] by
haVin g [+voice] stop [b] and violates a faithfulness constraint IDENT-IO[+/-voice] because
I‘VOi ce] in input is changed to [+voice]. Although (7Ab) has two violations of
ConStraints, (7Ab) is selected as the most optimal candidate. Thus, satisfaction of high

anked constraints IS more important than satisfaction of low ranked constraints 1n OT.

16

In (7B), although there is a [+voice] stop [b] in input, constraints and their hierarchy
suggested in (6) still select the same [aba] as the most optimal output.

In (7C) and (7D), both output candidates [ap] and [ab] satisfy the highest
constraint *[-voice] N_V because *[-voice] /V_V demands stops be voiced between two
vowels. In fact, the constraint *[-voice] N_V is not the relevant constraint for [ap] and
[ab]. (7Cb) and (7Db) violate *[+voice] as they have the voiced consonant [b] in output.
(7Cb) satisfies IDENT-IO[+/-voice] as the {-voice] stop [p] is maintained. However,
(7Db) violates IDENT-IO[+/—voice] as correspondent segments are not identical in [+/-
voice] features.

The comparison of tableaux (7A) to (7B) and (7C) to (7D) shows that stops are
voiced between two vowels but not in any other context regardless of candidate’s input
voicing specification in L1. Thus (7) Shows that rather than feature specification or

conditions that are imposed on an input, it is the interaction of output constraints that
determine the most optimal outputs (Kager 1999: 19).

In L2 (2b), [+/-voice] stOps are contrastive in outputs. However, {-voice] and
[+V0ice] are neutralized to [+voice] between two vowels. Consider the fact that [+/-
Voice] stops are contrastive in outputs. If IDENT-IO[+/-voice] is ranked higher than
aki‘l‘Voice], outputs maintain the [+/-voice] specification of their input.

Minimal pairs such as [ba] vs. [pa] and [ag] vs. [ak] in L2 suggest that the

faithfulness constraint, IDENT-IO[+/-voice] is ranked higher than *[+voice] as [-voice] and
[+voice] are contrastive except between two vowels. Between two vowels, there are no

[‘Voice] stops although {-voice] and [+voice] are contrastive in L2, resulting in positional

17

L

neutralization. This means that {-voice] in inputs is changed to [+voice] in ouputs
between two vowels, resulting in satisfaction of *[-voice] N_V and violation of IDENT-
IO[+/-voice]. Thus, the constraint *[-voice] N_V should be ranked higher than IDENT-
IO[+/-voice]. Otherwise, {-voice] and [+voice] stops would be contrastive regardless of

their positions. (8) shows the constraint hierarchy and evaluation of constraints for L2.

(8) L2: *[-voice]/ V_V >> IDENT-IO[+/-voice] >> *[+voice]

 

 

 

 

 

 

 

 

 

 

 

A. 7
. *[-voice]/ V_V IDENT—IO[+/-voice]

B.
*[-voice]/ V_V IDENT-IO[+/-voice]

C.

D.

 

 

 

 

 

[apa] in (SA) and (8B) is not the most optimal candidate as it violates the
und0minated *[-voice]/ V_V. [apa] in (SA) satisfies the constraint IDENT-IO[+/-voice]

ut the satisfaction of IDENT-IO[+/-vorce] 15 not relevant 1n (8A) as it Vlolates the highest

18

 

constraint *[-voice]/ V_V. In (8A), [aba] violates IDENT-IO[+/—voice] but [aba] is a more
optimal candidate than [apa] because of satisfaction of *[-voice] N_V. Contrary to L1,
the faithful constraint IDENT-IO[+/-voice] plays a role in selecting the most optimal
candidates in L2 as shown in the comparison of (8C) and (8D). In (8C) and (8D), the
most optimal candidates always satisfy the faithfulness constraint IDENT-IO[+/-voice].

The comparison of (7) to (8) Shows that language differences stem from language-

particular rankings of violable but identical constraints. In OT, language universality is
expressed in the same constraints that are violable. Language variations are expressed in
terms of language-particular constraint rankings as shown in (7) and (8). Markedness is
expressed with negative or positive constraints as in *[+voice], *NO CODA (syllables are
open), or ONSET (syllables should begin with consonants). Language specific
phonological patterns result from interactiOn of markedness constraints and faithfulness
ConStraintS.

For instance, voiced stops and voiceless stops are in complementary distribution
in L1 because they never appear in the same environment. The constraint ranking *[-
V0iCe]/V_V over *[+voice] results in [+voice] stops between two vowels in L1. The
con Straint ranking *[voice] over IDENT-IO[+/-voice], in turn, disallows [+voice] stops in
all Other contexts as *[+voice], a context-free constraint, eliminates all voiced stops

r egardless of their positions.

Contrary to L1, voiced stops and voiceless stops appear in the same

enVironments in L2 because of interaction between IDENT-IO[+/-voice] >> *[+voice].

ITIOVVever, the constraint ranking *[-voice]/V_V over IDENT-IO[+/-voice] results in no [-

19

 

voice] stops between two vowels. Thus, effects of IDENT-IO[+/-voice] appear when
stops do not occur between two vowels as in (SC) and (8D). In (8C) and (8D), the most
optimal candidates do not violate IDENT-IO[+/-voice]. Thus, the faithfulness constraint,
IDENT-IO[+/-voice], plays a role in selecting the most optimal candidates. In L2,
positional neutalization of [+/-voice] results from the domination of the context-sensitive
marked constraint *[-voice]N_V over the faithfulness constraint, IDENT—IO[+/-voice] and
context-free marked constraints *[+voice].

Data in this research shows positional neutralization. Thus, in this study,
context-sensitive markedness constraints are ranked higher than faithfulness constraints in
general, as in L2. Relevant markedness constraints are introduced in each chapter.
Relevant faithfulness constraints are introduced below based on Correspondence Theory
(CT) suggested by McCarthy and Prince (1995).

CT (McCarthy and Prince 1995) deals with issues of faithfulness between an

iIlput and an output. Constraints in CT basically demand that 81 (input, base, etc.) be
ideIltical to 82 (output, reduplicant, etc.). Three of the constraint families that play an

1important role in this research are introduced in (9).

20

(9) McCarthy and Prince (1995: 16)

A. The MAX Constraint Family: (No Deletion)
Every segment of S] has a correspondence in S2.

B. The DEP(ENDENCE) Constraint Family: (No Insertion)
Every segment of S2 has a correspondence in SI

C. The IDENT(ITY) (Feature) Constraint Family (IDENT-IO[F])

(N 0 Feature adjustment)
Let or be a segment in SI and [3 be any correspondence of or in S2.

Ifor is [7F], then B is [7F].

(Correspondent segments are identical in feature F).

The MAX constraint family favors the preservation of every element of S] in S2

Thus, the deletion of a segment is disfavored in the MAX constraint family. In the
DEPENDENCE (DEP) constraint family, every segment of 82 requires a correspondence in
S] The epenthesis of a default segment or other segment is disfavored in the DEP
Constraint family, because the epenthetic segment does not have a relevant segment in
inPut. The IDENTITY (IDENT-IO) (Feature) constraints favor the maintenance of

Underlying features in output. For instance, phonological changes such as deaspiration of
a Segment (/ph/ -) [p]) and palatalization of a segment (/5/ -> [1]) are disfavored.

As it will be discussed in 1.2.4, in Articulatory Phonology, gestures that are
Specified with phonological goals such as constriction degree, constriction location,

Constriction shape, stiffness and timing are phonological units. Thus, the IDENTITY

“DENT-IO) (Feature) constraints are revised to incorporate the concept of gestures as

b - . . .
asl c umts of phonological representation:

21

 

(10) The IDENT(ITY) (Gesture) Constraint Family (IDENT-IO[G])
Gestures (constriction degree, constriction location, constriction shape,

stiffness, and timing) in 81 are maintained in S2.
(Correspondent segments are identical in Gestures (G)).

1.2.4 Articulatory Phonology

In Articulatory Phonology, speech is considered as a two-layered process that is
related to planning (phonology) and execution (phonetics). Phonology is the planning of
how each utterance is to take place. Phonetics is thought of as taking over the plan and
executing it.

According to Tatham (1999), one of the motivations for modeling speech as a
tWO-layered process is based on the difference between planning and execution. Tatham

gives an instance from Estuary English. A plan for each word, cap, cat, cack, cab, cod and
Gag, is to end with a different stop, which is either voiceless or voiced. However the plan
is SOrnehow altered in the execution. Thus, a listener differentiates words with voice and
VoiCeless stops by the length of the proceeding vowels rather than the voicing quality of
final stops. In other words, phonologically (or in planning) they are differentiated by the

VoiCing quality of final stops. Phonetically (or in execution), the difference is determined

by the length of vowels.

22

 

According to classical and early generative phonology, these words are
differentiated by voicing quality, regardless of phonological and phonetic levels. Thus,
classical and earlier phonologies need another level that interprets the difference between
planning and execution. At this level, [+/- long] may have to be used, even though it is
not even a distinctive feature in Estuary English. Thus, classical or earlier phonologies

either proliferate features and grammar, or miss generalizations that are present in a given

COI'ItCXI.

1.2.4.1 The unit of representation in Articulatory Phonology

A gesture, which is associated with space and time, is the basic atom of
Phonological structure in Articulatory Phonology. AS a phonological unit, a gesture is an
abStract characterization of coordinated task-directed movements. A gesture is specified
With articulators involved, constriction degrees (CD), constriction locations (CL) and
C0nstriction shapes (CS) over time. The gestures for a given utterance are organized into a
large coordinated structure, or constellation, namely a gestural score, as schematically

Shown in (11).

23

 

(11) Gestural Scores for Korean /pm/, /p’ae/, /p”2e/

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a. /pae/ b. /p’a=:/ c. /p“ae/
p e p’ 88 1)" as
[Open]
open [mid] [closed]
Glottis cl sed L D
open
[closed] [closed] [closed]
Lips closed :22 ¥
—> Time -—> Time -> Time

In (11), each enclosed box represents an active articulator in the vocal tract, in this
case the glottis and lips. The height of each box represents the constriction degree of the
relevant articulators. The width of each box represents the time that an articulator takes
t0 Complete given gestures. The order of boxes across the time line represents the order
0f gestures along the time line. Thus, lip gestures with the [closed] constriction degree

Occur before glottal gestures in (11a, 11b, and 11c).

However, (11a), (11b), and (1 lo) differ from each other in constriction degree of

the glottis. Plain /p/ in /pm/ (11a) has the [mid] constriction degree of the glottis, and /p’/
in /p’a/ (11b) has the [closed] constriction degree of the glottis. Aspirated /ph/ in /phae/

(l 1 C) has the [open] constriction degree of the glottis. The [closed] constriction degree
r ef ers to a tight articulatory closure for that particular gesture (Browman and Goldstein
I 989: 226). [closed] permits glottalization or oral stops, depending on the set of

arti Culators involved. The [closed] constriction degree of glottlal gestures creates

24

the oral and nasal stops. The [open] constriction degree is a wide articulatory opening
that permits aspiration or nonnasal sonorants, depending on the set of articulators
involved. For instance, the [open] constriction degrees at the glottis and at the
supralayngeal level generates aspiration (Browman and Goldstein 19892238). The [mid]
constriction degree of the glottis for the plain /p/ refers to an articulatory value that falls
between the [closed] constriction degree of the glottis for the tense /p’/ and the [open]
constriction degree of the glottis for the aspirated /ph/. The constriction degree values of
the glottis for Korean obstruents are discussed in detail in section 1.3.2.2.

The dotted lines in (11) represent the point where the [closed] constriction degree
of the lips is released. At the burst of the oral closure, the width of the glottis is different
according to types of /p/. Thus, aspirated /p"/ has the widest opening ([open]) of the
810ttis at the lipcclosure burst. Tense /p’/ has the narrowest opening ([closed]) of the
81 Ottis at the lip-closure burst.

As a phonetic unit, gestures are characterizations of the movements of articulators

thr()lrgh space and time. A gestural score also shows which gestures are co-articulated
With other gestures and how long they are co-articulated. The curved lines within the
boxes in (11) characterize degrees of constriction over time. Thus, in (11c), the opening
of the glottis reaches its maximum when the lip closing is released in /pha/.

The gestures for a given utterance characterize the actual observed movement of

arti Culators and function as units of contrast. In distinctive feature theories, contrasts of
Sounds are represented by the presence or absence of features. In Articulatory

p
hoIlology, the contrasts of sounds are represented by the presence or absence of a

25

 

Phonology, the contrasts of sounds are represented by the presence or absence of a
gesture, the degree of constriction, the timing of gestures and the organization of gestures
(Browman and Goldstein 1986 and 1991). For example, (11) above shows that the
different degrees of constriction of the glottis are distinctive in Korean stops.

Gestures as phonological units differ from both features and segments. A gesture
is larger than a feature; a feature refers to one specific character of sounds such as voicing
or nasality. A gesture in a gestural score includes information related to articulators,
constriction degrees, constriction locations and constriction shapes (CS) over time.
Gestures are smaller than segments as gestures are linked together to form a segment
(Browman and Goldstein 1989:210-11).

The phonological representations of Articulatory Phonology contain more
information on a segment than the phonological representations of Underspecification
theories do (Steriade 1987, Archangeli 1988, Archangeli and Pulleyblank 1993 and Paradis
and Prunet 1991). For instance, in a language where [+/-voice] are contrastive as in

English /tie/ and /die/, /t/ has all the relevant feature specifications, as in (12a) below.

(12) a. /t/ b. /t/ (based on Paradis and Prunet 1991)
Glottis open

Clgsed l/\ [mid ] [ Root

 

 

 

 

 

 

Totlgue Tip open alveolar 0 Place
t [closed]
[:321
closed

 

26

 

 

 

 

The active articulators for /t/, such as Glottis (GLO) and Tongue Tip (TT) are
specified, including information on the location of constriction (alveolar) in (12a).
However, the place node under the root node, as Shown in (12b) above, is enough for the

phonological representation of /t/ in Underspecification theories proposed by Archangeli

1988, Archangeli and Pulleyblank 1993 and Paradis and Prunet 1991.

1.2.4.2 Distinctiveness and natural classes in Articulatory Phonology
There are two ways of grouping sounds into natural classes in Articulatory
Phonology. First, natural classes are grouped based solely on the moving articulators

used, as shown in (13a).

(13) Articulatory Geometry Tree (Browman and Goldstein 1989:223)

 

Vocal Tra_ct (VT)
Oral
Articulator set GLO VEL TB TT LIPS
Assoiciated [CD] [CD] [CD] [CD] [CD]
parameter [CL] [CL] [CL] [CL]
dimensions

VT: Vocal tract GLO: Glottis VEL: Velum TB: Tongue Body TT: Tongue Tip
CD: Constriction Degree CL: Constriction Location

27

‘)P'_

0‘.

‘s-o-

3..“
v

 

 

 

 

Gestures as independent moving articulators are specified with constriction
degrees (CD) such as [open] or [closed], constriction locations (CL) such as [alveolar] and
[palatal], and other parameters“, as shown in (13). Articulatory Geometry (13) groups
gestures into a hierarchy based on the anatomy of the vocal tract and classifies gestures
into natural classes. Thus, [t, d, 5, z, n, t], d3, n, 1] form a natural class by using the same
arti culator, Tongue Tip (TT).

In addition, the vocal tract is viewed as a tube or a series of tubes in Articulatory
Phonology, as shown in (14). There are five basic tubes (nodes) in the vocal tract that are
connected: Glottis, Nasal, Lateral, Central, and Lips (14). The active moving articulators
0f gestures, such as the tongue tip and the velum, reside in these tubes. The vocal tract
(VT) as a tube is terminated by glottis and lip gestures. The simple nodes and/or the
terminators of tubes (Glottis and Lips) form tube junctions such as Tongue, Oral,

SuPralaryngeal (SL) and VT junctions. These tube junctions are related to the

suPerordinated nodes of feature systems.

7\
Othe

of 1 associated parameter dimensions are constriction shape [CS] and stiffness. Except for the stiffness

01rd e glottis, these dimensions are not directly relevant for analyses in this research. Thus, they are
tted from the articulator geometry tree (13).

28

(14) Tube Geometry (Browman and Goldstein 1989:236)

VT
SL
Oral
=>Tube GLO Nasal Lateral Central Lips

VEL TB TT

SL: Supralaryngeal

The second way of grouping sounds into natural classes is by using a constriction
degree hierarchy (CDH) within the vocal tract based on Tube Geometry (14). Browman
and Goldstein (1989223 7) state that the CD at each superordinate is predictable “from the
CD of the tubes being joined and the way they are joined” (see 15c).

If tubes are joined in parallel as in (15a), air flows through connected tubes at the
same time. Thus, the effective CD value of the compound tube is equal to the widest CD
Value of the connected tubes (B value in (15a)). If tubes are joined in series as in (15b), air
flows through connected tubes one after another. Thus, the effective CD value of the
compound tube is equal to the narrowestCD value of the connected tubes (A value in

(1513)). Based on aerodynamic principles, Browman and Goldstein (1989:237-39) predict

C011Striction degrees for superordinate nodes as in (15d).

29

(15) The airflow value/
The effective CD of the compound tube
a. Tubes in Parallel

0-i-
xiii..." .n = B value (the widest CD)

 

b. Tubes in Series
g-i-‘ = A value (the narrowest CD)
> .n
um...» Gib = A value (the narrowest CD)
c. Effective CD values for the superordinate tubes

(Browman and Goldstein 1989:237-39)

...When tubes are joined in parallel, the effective CD of the compound
tube has the CD of the widest component tube, that is, the maximum CD.
When they are joined in series, the compound tube has the CD of the
narrowest component tube, that is the minimum CD...

d. Percolation of CD up through Supralaryngeal node
(Browman and Goldstein 19892237)

Nasal [CD] = VEL [CD]

Lateral [CD] = TB [CD]

Central [CD] = MIN ('I'T [CD], TB[CD])

Tongue [CD] = MAX (Central [CD], Lateral [CD])
Oral [CD] = MIN (Tongue [CD], LIPS [CD])
Supra [CD] = MAX (Oral [CD], Nasal [CD])

Using (15d) as an example, the CD of the SL tube is predicted from the CD of the
or al tube and the CD of the nasal tube. As air can flow through the oral tube and the nasal
tube at the same time, they are connected in parallel. Thus, the CD of the SL node is

equal to the maximum CD of the oral or nasal tubes.

30

 

;’—-e-
. fiui . .. I'l—

In addition, the VT tube at the highest level is described in terms of the
characteristics of the ‘output’ flow such as occlusion, resonance and noise as shown in
(16a). This is because the airflow at this highest level is also influenced by the initiator
action from the lungs as well as the constriction degrees of the glottis and the SL node.

Thus, occlusion is either from the [closed] CD at the SL level or from the [closed] CD of

 

the glottis. Resonance results from the [open] CD at the SL level and [critical] CD at the
glottal level. Fricatives, which have turbulent airflow and noise, are from other

combinations of CDs at the SL level and the glottis. (16b) shows default CD values for

 

basic tubes when articulators in basic tubes are not active.

(16) a. VT output
Occlusion = SL [closed] OR GLO [closed]
Resonance = SL [open] AND GLO [critical] 5
Noise = Otherwise
(Browman and Goldstein 1989223 8)

b. Default constriction degree values for basic tubes and tenninators
Nasal [CD] = Closed
Lateral [CD] = Central [CD]
Central [CD] = Open
Lip [CD] = Open
Glottal [CD] = Critical (Crit)

(Browman and Goldstein 19892239)

(17) illustrates how the CD of the vocal tract for /t/ and /n/ are determined based on (15)
and (16).

\

5
[Critical] is an articulator value that pemrits frication (turbulence) or voicing, depending on the set of
Culators involved (oral or laryngeal) (Browman and Goldstein 19892225). A [critical] value within the

o
ral Ilode creates frication while the [critical] CD of the glottis creates voicing.

31

 

(17) Constriction degree hierarchies for /t/ and /n/

a. /t/
VT (Occlusion)
SL [closed]
Oral [closed]
Tongue[closed]
=>Tubes Glottis Nasal Lateral Central Lips
[closed] [closed] [closed] [open]
I |
VEL TB TB TT
l l
[mid] [closed] [closed]
/t/
b /n/
VT (Resonance)
SL [open]
Oral
[closed]
Nosed]
”Tubes Glottis Nasal Lateral Central Lips
[open] [closed] [closed] [open]
I |
VEL TB TB TT
l |
[critical] [open] [closed]

In (17a), when the TB is not an active articulator within the central tube, the
default CD value of the TB is [open] as the default CD value of the central tube is [open]

(See (16b)). Ifthe default CD value of the TB is [closed], the default CD value of the

32

central tube will be [closed] because the central node is formed by the serial connection of
the TB and the TT, and thus takes the minimum CD value of the TB and the TT, In this
study if TB gestures are not active within the central node, its default [CD] value, [open],
is not specified in order to avoid confusion with the [open] CD value of TB for the lateral
/1/ within the lateral tube.

In (17a), the CD at the central level is [closed] because the CD value of the central
level is equal to the minimum CD value of the TB or TT, as shown in (15d). The central
tube and the lateral tube are connected in parallel and form the tongue tube. If two tubes
are joined in parallel, air can flow simultaneously through the two tubes that are
connected. Thus, the higher node, namely the tongue node, has the maximum CD value of
the central tube or the lateral tube. However, the CD value of the lateral tube is the CD
Value of the central tube by default (see 16b). Thus, the CD at the tongue level is
[Closed].

The tongue and lip tubes are connected by the oral tube in series. If tubes are
lOined in series, air flows through tubes one after another. Thus, the oral tube has the
narrowest CD value of the tongue or the lip tubes. Although the CD value of the lip tube
is [open], this is not the CD of the active gestures. Thus, the CD value of the oral tube is
equéll to the CD value of the tongue node.

The SL node is formed by the oral and nasal tubes in parallel. The CD value of
the nasal tube is [closed] by default. Thus, the CD value at the SL level is [closed]. The
highest tube VT is influenced by glottal gestures, the SL node, and air from the lungs,

he'J‘acterrzmg the overall ‘output’ airflow in terms of occlusron, resonance, and norse

33

 

(frication). AS occlusion is characterized by the [closed] CD either at the glottis or at the
SL level, /t/ is occlusion because the CD value at the SL node is [closed].

The constriction degrees of hierarchy for [n] (17b) are the same up through the
oral tube. However, the CD at the SL level for [n] is different from that of [t]. As the SL
tube is connected in parallel with the nasal and oral tubes, the CD value at the SL level is
equal to the maximum CD of the nasal or oral tubes. Thus, the constriction degree at the
SL level is characterized by [open] for /n/. At the highest level, [n] is characterized as
resonance because it has the [critical] CD of the glottis and the [open] CD at the SL level.

There are two main points in CDH that distinguish it from traditional feature

systems. First, the CD of all levels are present at any given time regardless of whether
Some gestures are actively producing the constriction or not. Second, the CD of basic
tllbes and superordinated tubes can group sounds into natural classes at any level as the
Articulatory Geometry of (13) does. /t/ and /n/, for instance, form a natural class at a
loWer (Oral) level of the hierarchy rather than the Vocal Tract or Supralaryngeal levels. /t/
and /n/ do not form a natural class at the lowest Nasal, SL, or VT levels (tubes). If a
Phonological pattern in a language is related to the Oral level or within the Oral level, /t/
and /n/ will pattern together. Otherwise [t] and [n] will not pattern together.

The CD at various levels can serve to categorize and distinguish phonological units
in different ways. Browman and Goldstein (1989:240-41) claim that divergence among
traditional feature systems in classifying sounds lies in the differences of the levels at
W11ioh they classify sounds. Thus, Ladefoged defines [+/-sonorant] at the VT level while

Chornsky and Halle (1968) define [+/-sonorant] at the SL level. Thus [Y] and [h] do not

34

belong to [+sonorant] in Ladefoged’s classification system but [‘2] and [h] belong to

[+sonorant] in Chomsky and Halle’s.

Each gesture possesses an inherent spatial aspect and an intrinsic temporal aspect.
Thus, gestures are co-articulated in space and in time in a given context, generating

contextual variations (Browman and Goldstein 1989:211-21). When gestures are

overlapped in time, targeted gestures are achieved when the gestures are on different
articulatory tiers or tract variables. An example is h—coalescence, which sequences of /h/
+ /p, t, k/ or /p, t, k/ + /h/ are merged into aspirated obstruents. In h-coalescence, targeted
glottal and oral gestures are achieved although these gestures are overlapped in time. h-
Coalescence will be discussed in Chapter 4 in detail. Gestures are blended when the
geStures are on the same articulatory tier or tract variables. When two gestures are on the
Thus, they

Sarne tier, the two gestures are in competition to perform different tasks.

cannot overlap without perturbing each other’s tract variable motions.

As manners of articulation are concerned with how tight the constriction of
airflow is, CDH is closely related to the manners of articulation of traditional phonology.
In the remaining chapters, CDH is adopted in analyzing Korean consonant patterns that
ar e related to manners of articulation. As will be illustrated in the following chapters,
CDH provides generalizations for Korean consonant patterns in a simple and unified

Way.

35

1.3 Korean consonants

In order to provide background for subsequent discussion, this section introduces

the inventory of Korean consonants, Korean syllable structure, and articulatory, acoustic

and aerodynamic characteristics of Korean consonants.

1.3.1 Phonological structure of Korean

As illustrated in (18), for the same place of articulation, there are two or three
different types of obstruents that differ in the manner of articulation. These Obstruent
tYpes are labeled as plain or lax, tense and aspiration. They are specified with features

such as [+/-constricted glottis] and [+/-spread glottis] in literature (Cho 1990, Sohn 1987

and Silva 1993)

(18)

Korean consonant inventory

    

 

 

 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

Stop ________________________ k. .........
........................ k L-_ .----_-
\ Aspirated J“ th kh
Affricate __'_ _____________________ E ................
........................ 9'--. ------ .-----_
\ ch
I:r‘icative __________§ ____________________ h ---
\ S.
fiasal m 11 U
@quid I

 

 

 

 

Not all consonants of the inventory in (18) appear in all syllable positions. All

the consonants in the above inventory except /1]/ appear at syllable-initial position.
However, there are limitations on what consonants can appear in syllable-final position.
Only unreleased stops [p‘, t", k‘], nasals [m, r1, r3], and the lateral [1] occur before a
consonant or at the word-final position (_ (C, #}). This limitation on what sounds can
appear before a consonant or at the word-final position causes obstruents to be altered in

manners and places of articulation. In syllable-final position, bilabial, alveolar, and velar

obstruents are merged into [p’, t", k’] respectively. Palatal and glottal obstruents, /c,
C',Ch, h/, are merged into [t1].
Although there are several different views on Korean syllable structure (Ahn

1985, Kim 1986, Sohn 1987 and Kang 1992), the syllable structure adopted in this study

iS that shown in (19). Various studies on Korean syllable structure agree that Korean

Syllable structure is composed of [(C) (G) V (C)],,.

(19) Korean Syllable Structure (based on Ahn 1985 and Sohn 1987)
0'

R
N

I
(C) (G) v (C) G: Glide

37

 

Unless the second consonant of onset consonants is a glide, Korean does not allow
consonant clusters in onset or syllable-final positions as shown in (19). In word-medial
position, only two consonants are allowed in an output, one for coda and the other for the
onset of the following syllable, as in the examples of (20a). This even applies to cases

where morphology creates more than two consonants in word-medial position as in the

examples of (20b).
(20)
Input Output
a /palp+o/ [paLho] ‘step on’
/op§+9/ [op1 . s’o] ‘do not have’
b /pa[p+1go/ [pap1 .k’o] ‘step on and’
/9p_s+1go/ [op’ . 1_(’o] ‘do not have and‘
a. 1pf+1_cc/ [ip’ . k’o] ‘recite and’
+ : morphological boundaries (Verb stem + inflectional suffix)

: a syllable boundary

The examples in (20a) illustrate that when there are two consonants between two vowels,
the two consonants are maintained even though there may be phonological alternations
and the edges of morphological and syllable boundaries do not agree. When there are three
(or possibly more) consonants between two vowels as in (20b), only two of the
consonants are maintained in the output, resulting in consonant deletion. /l/ and /S/, for
instance, are not maintained in the outputs of (21a), (21b) and (21c). In addition, the

plain /p/ and aspirated /p“/ become [p’] in syllable-final position as shown in (21a), (21b)

and (21c).

38

 

(21) a. b. c.

Input /pa1pko/ /9 p 5 ko/ /i l phko/
ll ll
Output[pa p’ko] [op1 ko] [3“: p‘ko]

“(Va V \4 VV,

A comparison of (21a) to (21b) reveals that consonant position in the input does not
cause a segment to be deleted. The first segment after the first vowel is deleted in (21a)
while the second segment after the first vowel is deleted in (21b). (21a), (21b) and (21c)

reveal the preference for [p‘] over [1] or [s] as the syllable-final consonant. Contrary to
[s] in (21b), both [1] and [p‘] in (21a) and (210) are possible candidates to be in syllable-
final position. However, [p’] is maintained in the output, but not [1]. Chapter 2 explores

issues related to phonological alternations in syllable-final position.

1.3.2 Characteristics of Korean obstruents

In this section, characteristics of Korean obstruents in word and syllable-initial
Position are summarized. This is because obstnients in word and syllable—initial position
aChieve target gestures that are planned, contrasting all types of consonants in the
Consonant inventory (18) except [Q]. The other reason is that consonants in word-medial
and final position are more likely to be influenced by the preceding and/ or following

consonants.

39

1.3.2.1 Articulatory characteristics

Hirose, Lee, and Ushijima (1974) Fujimura (1972) and Kagaya (1974) have

observed the movements of the laryngeal muscles in Korean obstruents by using

fiberscopes and electromyographs and concluded that different types of Korean

ob struents result from

(Hiroese et a1. 1974: 151).

“the coordinated actions of the intrinsic laryngeal muscles”

According to Hirose, et a1, different types of Korean

obstruents have the following activity patterns of the laryngeal muscles:

 

 

 

 

 

 

 

 

(22) The activity patterns of the intrinsic laryngeal muscles for Korean obstruents:
(From Hirose, Lee, and Ushijima 1974)
CV Plain Tense Aspirated
Interarytenoid suppression of suppression of i, suppreSSionOfallfl V
(INT) abductor muscles is abductor muscles is fj‘adductor muscles of {if}
less dominant than less dominant than iithe larynx just before ff
that of aspirate that of aspirate ithe oral closure
obstruents. ob struents. igggrelease
.3 steep increasern' ’
factrvrty after the
‘ lirsuppreSSion .......
Lateral no significant increase ”a sharp increase in an intermediate level
Cricoarytenoid in the activity before i aetrVity immediately, . of activity but not
(LCA) the oral closure release. before the oral closure significant.
.\ Igrelease "‘ ‘ ~ . .
Vocalis (VOC) no significant increase amarked increase in an intermediate level
in the activity before [jactrvrty immediately of activity but not
the oral closure release. ffbefore the release significant.
resulting in an """"
increase in inner
tenSion of the vocal.
folds and constriction
k of the glottis

 

40

 

 

 

 

 

Orbicularis minimalacnwwln

 

(his (00) iiworriéinediaiflnoSi.tion439.33

Glottal Width begins to close earlier closes more rapidly stays Wideopenuntil
relative to the oral than plain stops and a theoralclosure
closure release. complete contact of release

the vocal process , j _ .
before the oral closure '
release. 'i " '

 

 

 

 

 

 

(22) illustrates that Korean obstruents are grouped differently according to the
activity of laryngeal muscles (Grayed areas vs non—grayed areas in (22)). The activity
patterns of INT separate aspirate obstruents from plain and tense obstruents while the
activity of LCA and VOC separate tense obstruents from plain and aspirate obstruents.
According to Hirose, et al (1974: 151), a marked increase in the activity of VOC before
the oral closure release in tense obstruents is closely associated with “an increase in inner
tension of the vocal folds” and “constriction of the glottis during or immediately after the

arti culator closure”.

Hirose et a1. (1974:152) also found that the activity patterns of 00 remain the

SElme for plain, tense and aspirate obstruents. However, the activity of 00 is minimal for
plain obstruents in word-medial position Thus, the activity patterns of 00 separate
Plain obstruents from tense and aspirate obstruents in the word-medial position.

Kagaya (1974) recorded the glottal width for different types of Korean
obstruents, using the glottal images over time. The results of Kagaya’s study are
Summarized in (23) with graphics. The glottal width of tense stops consistently
decreases and achieves complete contact earlier before the oral closure release (23 a). The

oral closure release occurs immediately before VOT. The glottal width of aspirate

41

 

ob struents increases gradually, reaches its maximum value at or around the oral closure
release and decrease rapidly after the oral closure (23b). The oral closure release occurs
l40ms to 220 ms before voice onset time. The glottal width of plain obstruents gradually

decreases and comes to a near close at or around the oral closure release (23c).

 

 

 

(23) a. Tense obstruents b. Aspirate obstruents c. Plain obstruents

Glottal r .

width 3! - .
increase 3: g :

‘1 g .

\ ii i l

, fl
'"" Time
wwwww : the oral closure release _ _ - _ :voice onset time (V OT)

1.3.2.2 Acoustic and aerodynamic characteristics

Acoustic studies of Korean obstruents agree that different Voice Onset Time
(VOT) values are one of the important properties for different types of Korean
obstruents (Han and Weitzman 1970, Hardcastle 1973, Kagaya 1974, and Kim 1970). In
general, tense obstruents have the shortest VOT after the oral closure release and aspirate
obstruents have the longest VOT. The VOT values of plain obstruents fall between those
of tense and aspirate obstruents as shown in (23). Kim (1970) suggests that the
difference in VOT is nothing but a function of the glottal opening at the time of the oral

release. The wider the glottal Opening is at the time of the oral release, the longer the

42

VOT is. Thus the constriction degree of the glottis for three types of Korean obstruents

are schematically represented as follows:

(24) The constriction degree of the glottal aperture for Korean ob struents

tense stops [closed]
plain stops [mid]
' aspirated [open]

In addition to different VOT values, qualities of voice onset such as fundamental
frequency value, are known to be important properties for the perception and production
0f Korean obstruents (Han and Weitzman 1970, Kagaya 1974, and Kim 1965).

The fundamental frequency after tense obstruents is highest. The fundamental
frequency after aspirate obstruents is higher than plain obstruents (Han and Weitzman
1970: 117). The fundamental frequency is closely related to stiffness of the laryngeal
muscles and inner-tension of the vocal tract and the pharynx (Kagaya 1974 and
I”iardcastle 1973). Hirose, et al. (1974:151) also observed that marked increased activity
of VOC and LCA before the oral closure release of tense obstruents “resulted in an
increase in inner-tension of the vocal cords” and “ in constriction of the glottis during or
immediately after the articulatory closure”. Thus there is a close correlation between high

fundamental frequency in tense and aspirate obstruents and active gestures of VOC and

LCA.

43

In addition, Hardcastle (1973) and Kim (1965) found that the duration of the first
full glottal cycle of a vowel following Obstruent release is always longer after plain
obstruents than after aspirated or tense obstruents. This phenomenon is due to slow and
weak airflow after plain obstruents according to Kim (1965). The large amount of airflow
after aspirated and tense obstruents is closely correlated with tension or stiffness of the
vocal folds according to Kim ( 1965: 351).

Han and Weitzman (1970) and Kim (1965) found that differences in the duration
of increased air pressure behind the oral closure and the speed of intensity build-up after
each type of obstruents are also important properties in the production and perception of
Korean obstruents. Duration of the increased air pressure is longer in tense and aspirate
Obstruents than in plain obstruents. The speed of intensity build-up after tense and

aSpirate obstruents is more rapid than that of lax obstruents due to subglottal pressure
and airflow. Tense and aspirate obstruents have higher subglottal pressure than plain
Obstruents.

(25) summarizes acoustic and aerodynamic characteristics of Korean obstruents
and correlates these characteristics to articulatory gestures of Korean ob struents based on
eirrperimental studies of Korean obstruents (Hardcastle 1973, Han and Weitzman 1970,

Hirose, et a1. 1974, Kagaya 1974, Kim 1965, Lee, and Smith 1973):

(25) The acoustic and aerodynamic characteristics of Korean obstruents in CV contexts:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Plain Tense Aspirated Sources
a. VOT intermediate shortest longest glottal width
b. f0 at the voice onset lowest highest higher stiffness from the
activities of VOC
and LCA
c. Duration of the longest shortest slightly amount airflow
initial glottal cycle for longer than from the lung
the following vowels the tense
obstruents
d. Stiffiiess Slack most stiff stiff activities of VOC
. and LCA
e. Air flow smallest large large glottal width and
stiffness
f. Subglottal pressure low highest high stiffness,
air flow
g. Duration of the short long longest stiffness,
ressure build-up air flow
h. Speed of the pressure gradual Sharp sharp stiffness,
build-up air flow

 

Based on the articulatory, acoustic, and aerodynamic characteristics of Korean ob struents,

the following gestures for the glottal articulator are suggested for Korean obstruents in

thi 8 study.

(26)

Plain ; [mid]
tense: [closed]
aspirated: [wide]

Constriction Degrees

 

Stiffness
[slack]

[high stiffness]
[stiff]

45

1.3.2.3 Constriction Degree Hierarchies

Korean obstruents are characterized by three types of constriction degrees:
[closed], [mid], or [open] for the glottal gestures at the time of oral articulator release as
shown in (26). Based on (26) and the framework of Articulatory Phonology, the

following constriction degree hierarchies are assumed for Korean obstruents.

(27) Constriction degree hierarchies for Korean consonants

a. Oral Stops
VT Occlusion

SL
[closed]

   
  

Oral
[closed]

Tongue
[closed]

Glo Nasal Latera/Entral Lips

[closed] [closed] [closed] [closed]
tensed /p’, t’, k’ / [closed] /\
Plain /p, t, k/ [mid] VEL TB TB TT
aSpirated /ph, t“, kh/ [open] [closed] [closed] [closed] [closed]
/k’, k, kh/ /t’, t, W /p’, p, p“/

46

b. Affricates

tensed /c’ /
plain /c /
aspirated /ch /

C. Fricatives

tensed /s’ /
plain /s /

VT Occlusion

\
SL

[closed/critical]

    
  

Oral

   

[closed/critical]
Tongue
[Clo/sled/critical]
Glo Nasal Lateral Central Lips
I [closed] [closed] [closed/critical] [open]
[closed] I
[mid] VEL TB TB TT
[open] [closed] [closed/critical] [open]
/c’, c, ch/
VT Noise (Frication)
SL
[critical]
Oral
[critical]
Tongue
[critical]
Glo Nasal Later Central Lips
I [closed] [cl sed] [critical] [open]
(1 /\
[closed] VEL TB TB TT
[mid] [closed] [critical]
/S, s’/

47

d./h/

/h/

e. Nasals

VT Noise (F rication)

SL
[Open]

Oral
[Open]

  

Tongue
[Open]
Glo Nasal Latermntral Lips
[closed] [open] [open] [open]
i /\
VEL TB TB TT
[open] [closed] [open][open] [open]
VT Resonance
\
SL
[Open]

 
   

Oral

[closed]
Tongue
[closed]
Glo Nasal Lateral Central Lips
I [open] [closed] [closed] [closed]
[critical] I I

VEL TB TB TT

[open] [closed] [closed] [closed]

/rj/ /n/ . /m/

48

 

f. Lateral VT Resonance

\
SL

[Open]

Oral
[Open]

  

Tongue
[Open]
Glo Nasal Lateral Central Lips
I [closed] [open] [closed] [open]
[critical] I I
VEL TB TB TT
[open] [closed] [open]
/1/ /l/

The above CDH shows that both stops and affricates have similar characteristics
at the oral, supra laryngeal, and vocal tract levels. Thus, the prediction is that, everything
being equal, stops and affricates pattern together in Korean phonology. Nasals and the
lateral /l/ have a similar characterization at the vocal tract level and the supralaryngeal
level. However, nasals separate themselves from the lateral /l/ at the oral level. Nasals,
StOps and affricates have a closed degree of constriction at the oral level. Thus, another
Prediction is that nasals, stops, and affricates will sometimes pattern together in
Phonology. /h/ has a characterization similar to nasals and the lateral /l/ at the
SUpralaryngeal level and the oral level. Thus, it is predicted that /11/ could pattern together
with nasals and the lateral /l/ in some cases. However, fh/ separates itself from nasals and
the lateral /l/ at the vocal tract level, and forms a natural class with fricatives. /h/ and

fricatives have noise output at the vocal tract level. However, fricatives are likely to form

49

a natural class with stops and affricates. Fricatives, stops, and affricates have a similar
characterization at the supralaryngeal and oral levels by having a narrower constriction
degree than nasals, the lateral /l/ and /h/.

The following chapters illustrate how natural classes formed by the CDH function

together in Korean consonantal phonology, especially manner alternations.

50

Chapter 2
Coda Neutralization and Coda Consonant Cluster Simplification

2.1 Coda Neutralization

2.1.1 Introduction

There are limitations on what consonants can appear in word-final position or

before a consonant in Korean (_{# or C}). Sonorants /m, n, I], l/, are allowed in word-

final position or before a consonant, as shown in (1).

(l) Word-final position Before a consonant
/kam/ -> [kam] ‘persimmon’ /kamki/ —> [kamgi]l ‘cold’ noun
/kan/ —-) [kan] ‘liver’ /kanja13/ —> [kanjarj] ‘soy sauce’
/kon/ —> [kon] ‘ball’ /kogki/ —-> [koggi] ‘a bowl’
lol/ —) [91] ‘spirit’ lolkul/ —> [olgul] ‘face’

of. /kam +i/—>[ka.m_i] ‘persimmon + subject marker’

/ka;1 + i/ —-) [kani] ‘liver + subject marker’
/kon+i/ —> [koni] ‘ball + subject marker’
/olkul+i/ -—> [91. gu.ri] ‘face + subject marker’

 

‘ Lax stop voicing between two [+voice] phones is not discussed here for simplicity.

51

In the case of obstruents, only unreleased [p", t", k’] appear at word-final

position or before a consonant, as shown in (2). Stops, affricates, and fricatives are

neutralized to unreleased stops (S’) in word-final position or before a consonant,

resulting in Coda Neutralization (CN). (2) shows patterns of obstruents in word-final
position or before a consonant. (2d) illustrates evidence for neutralization of obstruents

in syllable-final position.

(2) Patterns of obstruents in coda position

 

Input Output Examples2
a. Stops /p, ph/ [p"] I /afi‘/ [ap‘] ‘front’
It. t“/ [1‘] /pa;2/ [par] ‘field’
“a k’, k"/ [k’] i /nak_’si/ [nak‘ .s’i] ‘fishing’
/_ {#, C}
b. Affricates /c, 0’, ch/ [t1] maeh/ [nat‘] ‘face’
c. Fricatives /s, s’, h/ [t’] J /nas/ [nat’] ‘sickle’
/noh+ko/[not1 .lg'jo]‘and place (verb)’
d- lath/ + /i/Subcject marker [apl’i] ‘front,
/pafi/ + /e/ Location mm... [pa._t_he] ‘at the field’
/ 1133.7 Verb stem + / 3‘] declarative Ina-Ea] ‘fiSh (verb),
/nacf/ + /e/ Location marker [nacfe] ‘at the face’
/nas/ + /i/3ubejec( marker [nasi] ‘sickle’
/noh/ + /a/declal~dfive [no_.a]‘place (verb)’

In addition, complex onsets and codas are not allowed in Korean syllable
structure, as shown in (20) in section 1.3.1. For convenience’s sake, the word-final
position and the position before a consonant are referred to as coda positions in this

study. Because Korean does not allow a consonant cluster in onset and coda positions,

 

2) . , . . . . .
‘ There are no occurrences 01 /p / and /t / 1n word-final prismon anti belore a consonant.

52

the word-final position and the position before a consonant are equivalent to a coda
position.

The limitations on how many and what types of consonants are possible in coda
position result in CN and Coda Consonant Cluster Simplification (CCCS). The purpose
of this chapter is to investigate systematic principles in CN and CCCS and to present

output constraints based on these principles.

2.1.2 Previous analyses

CN is described in previous studies (Cho 1990 and Sohn 1987) as a phonological
adjustment in the coda that delinks laryngeal features, [+continuant], and [-anteri or] under
the place node. Thus, stops with laryngeal features /S’, Sh/ (S=stops), continuants /s, s’,

h/, and palatals /c, c’ ch/ are prevented from occurring in coda position, as shown in (3).

53

(3) Coda Neutralization

 

 

a. Sohn (1987: 266) b. Cho (1990:106)
N,
|
Core Skeleton X C °
Root node 0 0 Root node
Laryngeal tier ' ' Supralaryngeal tier Laryngeal ° Place [+continuant]

7‘ 1‘
[OP] [-anterior]
Manner tier 0 Place tier

i 3‘

[+continuant] [-anteri or]

N’: noun phrase, :5: delinking of features, [orF] =[+ spread glottis, + constricted glottis]

According to the CN in (3), delinking of laryngeal features such as [+spread
glottis] and [+constricted glottis] and delinking of [+continuant] are changes in manner of
articulation. Features such as [spread glottis] and [constricted glottis] refer to how tight
the glottal opening is. [continuant] refers to how tight an oral opening is so that air can
flow continuously, creating frication. Thus features such as [spread glottis], [constricted
glottis] and [continuant] concern constriction degrees of articulators in the vocal tract.
However, delinking of [-anterior] changes place of articulation. Thus, CN is considered as
phonological adjustments in coda position that change both manner and place of
articulation as shown in (3).

However, CN as delinkings of laryngeal features (3a) or the laryngeal node (3b),
[+continuant], and [anterior] under the place node in (3) does not support one of the

premises that Feature Geometry theories are based upon. Based on Clements (1985) and

54

Clements and Hume (1995)’s claim that certain features as a group recurrently participate
in phonological rules and constraints, Kenstowicz (1994:147) claims that “two features
can thus be expected to co-occur in rules or constraints only if they share a common node
in the tree”. (3) shows that the common node which is shared by laryngeal features,
[+continuant] and [-anterior] is the root node in both Sohn’s (1987) and Cho’s (1990)
studies.

Thus, CN should be expressed as delinking of the root node according to
assumptions of Feature Geometry theories. If the root node is delinked, either deletion of
a segment or geminate occurs in languages. However, Korean data in (2) shows that CN is
neither deletion of a segment nor formation of a geminate. Thus, CN as delinkings of
laryngeal features, [+continuant] and [-anterior] needs to be reanalyzed or modified.

In addition, CN as represented in (3) differs from other phonological processes in
Feature Geometry Theories. Generally, in Feature Geometry Theories, a phonological
rule operates on a single node rather than several nodes simultaneously. For instance,
voicing neutralization in word-final position in German is a phonological rule that
operates on the laryngeal node that dominates [+/-voice] features. Assimilation of coronal
consonants to the following labial or dorsal consonants in Korean is a phonological rule
that operates on the place node. Thus, in Feature Geometry Theories, a phonological
process is expressed as an operation on a single node that groups certain features together.
However, the phonological rules in (3) operate on several nodes simultaneously, affecting

both manner and place of articulation.

55

This chapter argues for the following four points: First, CN is neutralization in
manner of articulation that occurs systematically in coda position in Korean. Second,
Korean CN is better understood with phonetically motivated output constraints. Third,
CN results from impoverished articulatory gestures in a given context. The mending of
impoverished gestures may be language specific. In the case of Korean, there is a
relatively higher ranked output constraint that requires no contrastive constriction degrees
in the vocal tract when relevant gestures to obstruents are impoverished. Thus, output
forms in (2) result from the best possible satisfaction of output constraints. Fourth,
different rankings of the same constraints are responsible for different patterns in CCCS
between the Standard Korean (SK) and the Kyoung-Sang Korean (KSK).

The patterns of stops, fricatives and affricates in coda position are discussed in
section 2.1.3.1, section 2.1.3.2 and section 2.1.3.3 respectively. Based on the analysis of
stops, fricatives and affricates in coda position, I will demonstrate in section 2.2 that

there are principled reasons behind CCCS in Korean regardless of dialects.

56

2.1.3 Patterns of obstruents in coda position

2.1.3.1 Patterns of stops in coda position

2.1.3.1.1 Introduction

The delinking of laryngeal features such as [spread glottis] and [constricted glottis]

in (3a) or the laryngeal node (3b) is mainly proposed for bilabial, alveolar, and velar stops.

These stops are neutralized into unreleased stops [p", t", k’] respectively in coda

position without changing their places of articulation, as shown in (4a) and (4b).

(4) Patterns of stops in coda position

a. Word final position

/ap_h/ —) [ap’] ‘front’
/pafl/ —> [pa_t_‘] ‘field’
/ip/ ——> [ip‘] ‘mouth’
/puar_<_*1/ —> [pualg "] ‘kitchen’

b. Before a consonant

/ap_h+kwa/ —> [ap ‘ . k’wa]3
/pafl+kwa/ —> [pat1 .k’wa]
/ip+kwa/ —) [ip_1 .k’wa]
/pualg_h+kwa/ —> [puak‘ . k’wa]
/nal_<_’ta/ —> [nal_<'1 .t’a]

 

‘front and’

‘field and’

‘mouth and’
‘kitchen and’
‘fish’ (declarative)

3 Tensification after unreleased stops (Post-Obstruent Tensification) is discussed in Chapter 3.

57

c. Between vowels

/ai‘+e/ —> [ad‘s] ‘ in the front’
/pafl+e/ —>[pa._t'_‘e] ‘in the face’
/ip+e/ —> [i.b£] ‘in the mouth’
/pu9k§+e/ —>[pu9.li’e] ‘in a kitchen’
/nakj+a/—)[na.k_’a] ‘fish +declarative’

Underlying forms in (4a) and (4b) are justified by the comparison of (4c) to (4a)
and (4b). For instance, if the words in (4a) have underlying forms that are the same as the

phonetic forms in (4a) and (4b), /puok"/ for 'kitchen' and /nak"t'a/ for ’fish +declarative',

there is no way to predict when unreleased stops become tense, aspirated or voiced stops
between two vowels, as shown in (4c). Thus, underlying forms in (4a) and (4b) are

justified. Therefore, different types of stops are neutralized to [p", t", k"] in coda

position without changing their places of articulation, as shown in (4a) and (4b).

2. 1. 3. 1. 2 Gestural characteristics of Korean stops

Before presenting a phonetic-based output constraint for patterns of three types of
Korean stops in coda position, an overview of gestural characteristics of Korean stops is
in order. Glottal gestures for stops continue after the oral-closure release in syllable-
initial position, as in (5a), according to Steriade (1997216) and Hirose et a]. (1974). The

schematic CDs of glottal gestures after oral closure and release are presented in (5b) based

58

on experimental studies of Hirose et al. (1974), Kagaya (1974), and Kim (1965). At the
time of oral burst, the CDs of the glottis for aspirated, tense and lax stops are [open],
[closed], and [mid] respectively, as mentioned in (24) section 1.3.2.2. The glottal gestures
across time are illustrated in (SC) for three types of Korean obstruents. Glottal gestures
begin with opening gestures and end with closing gestures after reaching its targeted

opening gestures over time (Browman and Goldstein 1986 and Steriade 1997).

(5) Gestural characteristics of Korean stops in CV- position:

a. Timing (Steriade 1997116)

Glottal Gestures [----glottal open/close ------]
Oral Gestures [ ----- oral closure release---]¢
contextual cues for laryngeal features

b. Glottal CDs at the time of or afier the oral burst

oral burst
oral closure i glottal openings
Aspirated stops:

 

 

[Open]
Tense stops:
[closed]
Lax stops:
[mid]

 

c. Glottal Gestures
Time -) opening closing
Aspirate stops —

Lax stops “

Tense stops A

59

In general, the CD of lax stops is determined by contexts in Korean phonology.
Lax stops are slightly aspirated, voiced, tensed or unreleased depending on contexts where

they appear, as shown in (6).

(6) a. b. c.
Lax /p, t, k/ Aspirated /ph, t“, kh/ Tensed /p’, t’, k’/

#_ slightly aspirated

[tal] ‘moon’ [flal] ‘problem, mask’ [rial] ‘daughter’
V_V voiced

[agi] ‘baby’ [apha] ‘sick’ [nak’ayo] ‘to fish, informal’
VCr_V tensed

[mak1 .1128] ‘stick’ [kuk1 . 120] ‘land of a country’ [kalr1 . fii] ‘sash’
V_C2V unreleased unreleased unreleased

[kak1 .k’i] ‘separately’ [ap’ .k’o] ‘and carry’ [nak1 .t’a] ‘fish, statement’
_# unreleased unreleased

[ak]] ‘devil’ [k’it‘] ‘end’

Based on this context sensitivity of lax stops, Kagaya (1974) and Fujimura (1972)
suggest that there is no dynamic target CD in the glottis for lax stops, although the CD of
lax stops is [mid] in the syllable and word-initial position. In this study, based on Hirose
et al. (1974), Kagaya (1974), Ladefoged (1973) and Kim (1970), I adopt the premise that
the target CD of the glottis for lax stops is [mid], which is an intermediate value that

comes between those of aspirated and tense stops. This is because lax stops are

60

contrastive when they appear in the syllable and word-initial position, as shown in [tal]
‘moon’, [thal] ‘problem or mask’, and [t’al] ‘daughter’.

In addition to the contrastive glottal gestures for different types of Korean
obstruents, as shown in (5b), there are also three articulatory stages for obstruents at the
oral level. These three articulatory stages, which occur across time are shutting, closure
and release, as shown in (7). Stops are characterized by the [closed] CD value in lips, the

TT or the TB.

(7) Oral gestures

Time —)
Shutting Closure Release

: [closed] for stops/

The oral gestures shown in (7) are co-articulated or overlapped with the contrastive
glottal gestures across time, as shown in (5a). Co-articulation of contrastive glottal
gestures (5b), the oral articulator gestures (7), and other gestures such as the velum
gestures, and the amount of airflow from the lungs create the overall output characteristic
of stops at the VT level, namely occlusion, as shown in (8). The CDs in active
articulators, represented in bold face in (8), are input gestures for different types of
Korean stops, functioning as contrastive units in the input. The combined effects from
the CD of each articulator at a given time are reflected in the CDH for three types of

Korean stops in (8). Thus, (8) is considered to be the CDH in the input for three types

61

of Korean stops. (8) illustrates that three types of Korean stops have different CDs at

the glottis but the same CD, [closed], at the oral articulators.

(8) The CDH for three types of Korean oral stops

VT (occlusion)

SL
[closed]

 
   
 

Oral
[closed]

Tongue)
[closed]

Glo Nasal Later/Entral Lips

[closed] [closed] [closed] [closed]
tensed /p’,t’, k’/ [closed] | /\

plain /p,t, k/ [mid] VEL TB TB TT

aspirated /ph,t", kh/ [open] [closed] [closed] [closed]
/k, k', kh/ /t, t’, th/ /p, p’, ph/

2.1.3.1.3 The gesture-based account of stops in coda position

Gestures of moving articulators are likely to be reorganized according to the
context, creating different constriction, acoustic, and aerodynamic effects in the vocal tract
(Byrd 1996, Steriade 1997, Ohala 1983, and Browman and Goldstein 1991b). However,
the presence or absence of gestural modifications, degrees of gestural modification, and

types of modification in a given context are language specific.

62

In addition, gestural modifications are sensitive to linguistic factors such as
gestures involved for a given utterance, prosodic categories and positions, and morpheme
categories, as documented in the experimental studies of Byrd (1996), Byrd and Tan
(1996), Browman and Goldstein (1990b), Cho (1998a and b), Kim and Jongman (1996),
and Kim (1998).

Stops in coda position are more subject to gestural reduction spatially and
temporally than are stops in onset position (Byrd 1996 and Cho 1998a and b). Kim
(1998), for instance, compares releasedness between English and Korean stops in syllable

and word-final position.

(9) Kim (1998: 352 and 359)

 

 

 

 

 

 

 

 

Acoustic characteristics of coda consonants English Korean
a. aspiration together with oral burst 51.2% 0%
b. brief low amplitude oral burst after silences 35.7% 83%
c. silence corresponding to oral closure 13.1% 17%

 

Coda stops of English /p, t, k/ are characterized by aspiration together with an oral
burst (51.2%), a brief low amplitude oral burst after silence (35.7%), or silence
corresponding to oral closure (13.1%), as shown in (9). Coda stops of Korean /t, t“, s/4
are characterized by either a brief low amplitude oral burst after silence (83%) or silence

corresponding to oral closure (17%).

 

‘ /s/ is included in Kim’s study (1998) because the phoneme /s/ is realized as [t‘] in coda position as is the
phoneme /t/ (see 2.3.1.2 for further discussion).

63

Kim (1998) claims that if the oral burst, which is closely related to the removal of
oral articulators after the oral closure, is interpreted as [+release], both English and
Korean stops are released. Both languages have the oral burst (86.9% in English and 83%
in Korean). Therefore, English and Korean stops differ from each other by the presence
or absence of aspiration after the removal of oral closure, as shown in (9a).

Kim also observes that some English coda consonants have a diffusing-rising
shape that is similar to a released onset consonant but Korean stops do not have this.
Kim (362) claims that ““release” is associated with the presence of a pulmonic-egressive
airstream which flows through the oral tract after the removal of the oral closure, before or
during the articulation of a following segment.”

As acoustic and articulatory characteristics of a segment are not independent from
each other, it may be assumed that the presence or absence of a pulmonic-egressive
airstream after the oral-closure release is closely associated with the presence or absence
of the targeted glottal gestures after the oral-closure release. Especially, the presence or
absence of aspiration after the removal of oral closure in (9a) is closely correlated to CDs
of the glottis. Therefore, Kim's (1998) acoustic definition of [+/-release] could be stated
in articulatory terms: [+/-release] articulatorily refers to the presence or absence of the
targeted glottal gestures after the oral-closure release, before or during the articulation of a
following segment.

Therefore, articulatorily, Korean stops in coda position are unreleased because
targeted glottal gestures are either absent or impoverished. Especially, different CD

values of the glottis for different types of stops are either absent or impoverished after

64

the oral-closure release. This absence of targeted glottal gestures, in turn, changes the
overall CDH of the vocal tract because a CDH results from CDs of moving articulators
that are active concurrently. (10) shows the output CDH for three types of Korean
stops in coda position.

In order to avoid confusion, a CDH with dotted lines, as in (10), represents an
output CDH, and a CDH with a solid line, as in (8), represents an input CDH hereafter.
The comparison of (8) and (10) indicates that there is discrepancy between the planning

of an utterance (8) and executing the plan in a given context.

(10) The CDH for three types of Koran oral stops in coda position

VT (occlusion)
r-.-
I” ‘ SL
,’ [closed]
I I' ~~~~~~
Il ” - ~
1’ ,’ Oral
,’ ,’I [closed]
I, I, ' a ’«\
I I , I s
I I Tongue) \
I I \
,’ ,’ [closed] \
. I A \\
[I ‘ ‘ .
. \ .
Glottal Nasal Lateral Central Lrps

[closed] [closed] [closed] [closed]
I I F x I
I ’ ‘~ I

I
VEL T'B TB TT r
[closed] [closed] [closed] [closed]

[k’] [t’] [P’]

The CDH in (10) reflects the fact that contrastive CDs at the glottis are absent in

Korean when stops are in coda position. Although CDs of glottal gestures are absent, the

65

overall output characteristic in VT remains as occlusion because the CD value at the SL

level is still [closed]. Based on (10), unreleased stops (8’) may be defined as stops

whose targeted glottal gestures are either absent or weakened but have the [closed] CD
value at the SL level.

Based on the fact that consonants are likely to be unreleased in word-final
position or before a consonant (Hudson 1995), a marked constraint that is sensitive to

releasedness of stops and context, is suggested in (11).

(11) "‘ Contrastive Constriction Degrees of glottis /_{#, C}

(*CCDGLoms /_{#, C} hereafter)

No contrastive constriction degrees of the glottis when stops appear in word-final

position or before a consonant.

(11) demands that stops not be contrastive in terms of the CD of the glottis before
a consonant or in word-final position. However, the constraint (11) conflicts with the
faithfiilness constraint suggested in (10) in section 1.2.3. The faithfulness constraint is

shown below.

(12) IDENT(ITY)-IO(Gestures) (IDENT—IO[G])
Let 0t be a segment in SI and B be any correspondent of on in S2.

[fa is [7G], then [3 is [7G].

(Correspondent segments are identical in gestures (G))

66

The faithfulness constraint (12) demands that a surface form (output) be identical
to its lexical entry (input). Any deviation of the surface form from its lexical entry
violates the faithfulness constraint (12). If the context-sensitive markedness constraint
(11) dominates the faithfulness constraint (12) in a language, contrastive CDs of the
glottis are absent or weakened in coda position, resulting in positional neutralization in
terms of the CD of the glottis. If the faithfulness constraint (12) dominates the context-
sensitive markedness constraint (1]), stops in coda position are contrastive with regard to
the CD of the glottis, resulting in no positional neutralization.

Two generalizations are drawn from data (4). First, contrastive CD values for
different types of stops are absent or weakened in coda position, namely CN. As a
consequence, coda stops do not have identical glottal gestures between surface forms and
their lexical entries, as can be seen in (8) and (10). Second, no efforts are made to
compensate for the lack of contrastive glottal gestures in coda position. Therefore, in
Korean, the constraint *CCDGLomS/_{#, C} is ranked higher than IDENT-IO[G], as
shown in (13). Otherwise, the contrastive CD values of the glottis would be maintained

in the output.

67

(13) a. *CCDGLmvns/_{#, C} >> IDENT-IO[G]

 

 

 

 

 

 

 

A. /ip_"/ ‘leaf’ B lip/ ‘mouth’
/ h/: GLO[open], SL[closed], Lips[closed] /p/: GLO[mid], SL[closed], Lips[closed]
aw *ccoGLoms IDENT—IOIG] /ip_/ *ccoGwms IDENT-
L{#, C} / #, c IOtG]
a. in“ *l ' i 9 .‘ " a. in“ *l * GLO:
, " . {open} ' ,
b.ip’ *! _* (GLO): b.ip’ *! *GLO: ,
_ v‘ [closed] . [closed] '
c.ip *! 1, ._"‘(GLO): {may c.ip *!
train" * (GLO) cram" * (GLO)

 

 

 

 

 

 

 

 

In the examples of (13), any output candidates with [ph, p’, p] violate the
constraint *CCDGLO—mS/_{#, C}. This is because stops cannot be contrastive with
different CDs of the glottis in positions where different CDs of glottal gestures are

impoverished or weakened, as shown in (10). The unreleased bilabial stop in [ip’]

indicates that there is only the [closed] CD of the lips in coda position without any
targeted glottal gestures. However, candidates (13Ad) and (13Bd) violate IDENT-IO[G]
because gestures in the input are not maintained in the output. However, satisfaction of
*CCDGLomS/_{#, C} takes precedence over IDENT-IO[G] in Korean.

Note that the lack of active glottal gestures in [p’] in [ip’] ((13Ad) and (l3Bd))

does not necessary mean that there are no glottal gestures at all for obstruents in coda
position. I follow Browman and Goldstein (1992) and Sawashima et al. (1980) in
assuming that the default CD of laryngeal gestures is [critical] or a small degree of glottal
opening. The default [critical] CD of glottal gestures is as narrow as or narrower than that

of voicing. However, unlike the [critical] or [closed] CD of glottal gestures for voicing,

68

the default [critical] CD of glottal gestures in coda position does not function as a
contrastive unit because of the phonetic implementational conditions that will be
discussed later in this section. Therefore, the default [critical] CD of glottal gestures in
coda position does not violate *[critical]/ _{#, C} because the default [critical] CD of the
glottis is not contrastive in coda position.

According to Steriade (1997), where cues refer to acoustic properties that
influence the perception of sound categories, laryngeal neutralization is closely related to
the distribution of cues to a relevant contrast. For example, Steriade states that [voice] in
obstruents is more likely to be contrastive between two vowels (V_V) than in word-final
position (__#). This is because acoustic cues to [voice] such as VOT value, F0, and F 1
values at the onset of voicing in the following vowel, are lacking in word-final position.
Thus, cue—based constraints such as Context cue5 (voice) are suggested as a part of
phonological grammar. Context cue (voice) demands [voice] not be contrastive in
positions where acoustic cues to [voice] are absent.

The gesture-based account of laryngeal neutralization of coda stops in (13)
parallels the cue-based analysis of laryngeal neutralization of Steriade (1997). For
instance, different CDs of the glottis are closely related to different VOT values according
to Kim (1970) and Ladefoged (1973). In addition, different VOT values are known to be

critical acoustic cues in distinguishing different types of obstruents in Korean (Han and

 

5 Contextual or transitional cues refer to cues that are distributed before and after the oral closure stage of
obstruents. Onset cues: Cues that are distributed before the oral closure. Offset cues: Cues that are
distributed after the oral closure. Internal cues: Cues that are distributed during the oral closure (Steriade
1997).

69

Weitzman 1970 and Kim 1970). The lack of contrastive CDs of the glottis in coda
position means there is also a lack of contrastive VOT values for different types of
Korean stops in coda position. Therefore, the lack of contrastive glottal gestures in coda
position is also closely related to laryngeal neutralization.

(14) displays articulatory or aerodynamic properties of different types of Korean

stops distributed in # _ V, V_V, and _ {#, C} contexts.

 

(14)

Cin#_VorV__V Cin_{#,C}

different CDs of the glottis at oral- no different CDs of the glottis at
closure release oral-closure release

different oral closure duration. not observable

tense and aspiration : relatively long
(J. 1. Han 1997)

heightened subglottal pressure for slightly heightened subglottal
tense stops pressure for all stops

 

(14) shows that the constraint *CCDGLoms/_{#, C}, which is responsible for
laryngeal neutralization, is closely related to the distribution of gestures in a given context.
For example, different CD values of the glottis are not contrastive in positions where
relevant gestures are absent or weakened. Ifthe meaning of cues used in Steriade (1997) is
expanded to include articulatory gestures or aerodynamic properties that are employed to
distinguish one sound from another in production, the gesture-based account of stops in
coda position provides one more example to strengthen Steriade's claim that "relative

poverty of cues induce neutralization" (Steriade 1997 : 7).

7O

I propose that unreleased stops in _ {#, C} are more similar to tense stops in
Korean than to aspirated or lax stops in two aspects: heightened subglottal pressure and
the default [critical] CD of the glottis. First, unreleased stops are likely to have
heightened subglottal pressure because air from the glottis is captured behind the area
where a [closed] CD of an oral articulator is formed. The heightened subglottal pressure
of unreleased stops is also closely related to the patterns of Post-Obstruent Tensification
(POT) in Korean, as will be discussed in Chapter 3.

Second, the CD of the glottis for unreleased stops is [critical] by default as is that
of voiced obstruents. Browman and Goldstein (1992:239) assume the default laryngeal
gesture is appropriate for voicing if there is no specification for the laryngeal gesture.
However, Korean coda stops cannot be voiced because one of the requirements for
obstruents to be voiced is not met in Korean. For stops to be voiced, both a slightly
adducted status of the glottis and sufficient airflow through the adducted glottis are
required (Ohala 1983:294). In the case of Korean coda stops, the latter condition is not
met because there is no airflow through the adducted status of glottis as shown in Kim’s
study (1998).

In summary, the different CDs of the glottis after oral—closure release are very
critical in distinguishing different manners of Korean stops. However, the different CDs
of the glottis are not contrastive in coda position because the targeted CDs of the glottis
(5) are not observable or not achieved after oral-closure release. The context-sensitive
markedness constraint (1 1) takes into account implementation conditions that are related

to contrastive CDs of the glottis in coda position. (11) requires no contrastive CDs of the

71

glottis for different types of stops before a consonant or in word-final position where the
relevant CD gestures of the glottis are absent. Thus, laryngeal neutralization in data (4)
results from the interaction between the context-sensitive markedness constraint (11) and

the input-output faithfulness constraint (12), as shown in (13).

72

2.1.3.2 Patterns of fricatives in coda position

2.1.3.2.1 Introduction

According to Feature Geometries of Sagey (1986), McCarthy (1988), and Cho
(1990), /s, s', h/ form a natural class of fricatives by sharing [+continuant]. When /s, s'/
occur between vowels, as in (15a), corresponding segments of /s, s’/ have identical
gestures in the output. When /h/ occurs between two vowels, /h/ is deleted as in [no_a]
from moha/ ‘to place + declarative’. However, neither the oral fricatives /s, s’/ nor the

glottal fricative /h/ appear in coda position, as shown in (15b). Instead, /s, s', h/ are

merged into the unreleased alveolar stop [t’] in coda position.

(15) a. Fricatives between vowels b. F ricatives in coda position
Input Output Input Output
/nas+ i/ [nasi] ‘sickle + sub’ /na§/ [nap] ‘sickle’

/u1+os’+9/ [u,19.§’g]‘wept+declarative’ /ul+os’+ko/[u.ro_t’.k’o] ‘Wept’

/noh:a/ [no_a] ‘place +declarative’ /noh+ko/ [1101‘ ,flof ‘and place’
(verb)’

In previous analyses based on Underspecification Theories (Cho 1990 and Sohn

1987), the change of /s/ to [t’] in coda position is represented as delinking of

[+continuant] as in (16a). The change of /s’/ to [t’] is represented as delinking of

 

6 /h/ + /p, t, k/ are coalesced into aspirated obstruents, resulting in [p”, th, kh] in the fast, informal, or casual
styles of speech. However, in the slow, formal, or careful styles of speech, [tph, gt”, tkh] are heard.

73

[constricted glottis] and [+continuant], as in (16b). The change of /h/ to [t’] is

represented as delinking of [spread glottis], as in (16c).

(16) a. /s/ b. /s’/ c. /h/
/ \ / \ / \
Laryngeal 0 0 Supralaryngeal - o e o

I ’6

>«< >|<
a< [constricted glottis] [spread glottis]
[+continuant] [+continuant]

In addition, the alveolar unreleased stop [t1 ], which is a default consonant in Korean, is

inserted for /h/ at the later stage of derivation, resulting in the surface forms in (15b).

Therefore, neutralization of /s, s’, h/ to [t’] in coda position are treated with separate

rules, such as delinking of laryngeal features and/or [+continuant].

In this analysis, I will show that the neutralization of /s, s’, h/ to [t’] results from

an effort to satisfy output constraints, especially *Contrastive Constriction Degree (CCD)
/_ {#, C}. The constraint *CCD/_ {#, C} demands there be no contrastive CDs at the
vocal tract in word-final or before a consonant. The parallelism between the pattern of

fricatives /s, s’, h/ and the pattern of stops in coda position will be shown in 2.1.3.2.2.

74

2.1.3.2.2 Gestural characteristics of fricatives

Gesturally, /s, s’/ have the [critical] CD of TT at the alveolar region, as shown
schematically in (17a). /s/ and /s’/ are distinguished from each other by different CDs of
the glottis. /s/ is characterized by the [mid] CD at the glottis and /s’/ by the [closed] CD
at the glottis.

/h/ is characterized by the [open] CD at the glottis without any obstacle in the oral
cavity, as shown in (17b). The [open] CD at the glottis for /h/ refers to an articulatory

value that is relatively wider than those of the glottal stop [7] or voiced stops, thus

creating weak noise or aspiration.

(17) Gesture scores of fricatives - modified from Kagaya (1974: 164)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(a) /s, s’/ (b) /h/
[alveolar]
TT 9‘ [critical] _____/ TT
[Open]
[mid] for /s/
GLO [é ; GLO
[closed] for /s’/
—-) Time —> Time

According to Browman and Goldstein (1992), the peak of glottal gestures for oral
fricatives occurs while [critical] CD of TT gestures is maintained, as shown in (17a). In
other words, [mid] or [closed] CD of glottal gestures synchronizes with [critical] CD of
the TT for /s, s’/. The synchronized timing of glottal gestures relative to oral gestures

arises in order to create turbulent noise efficiently for /s, s’/ (Iverson and Salmons 1995).

75

This is because the amplitude of turbulent noise is increased by the presence of an
obstacle in the oral cavity.

The co—articulation of different CDs at the glottis and the [critical] CD of TT
gestures for the oral fricatives /s, s’/ and the [open] CD at the glottis for /h/ are important
articulatory gestures. Assuming the ‘average’ amount of airflow through the vocal
cavity, the co-articulation of glottal gestures and the [critical] CD of TT gestures for /s,
s’/ and the [open] CD of the glottis for /h/ are closely correlated to static acoustic targets
of fricatives, namely continuous frication noise (Johnson 1997). /s, s’/, namely
[+strident], are known to have higher turbulent noise than other fricatives such as /x/ and
/h/ (Shriberg and Kent 1995).

The CDs of TT gestures and glottal gestures creates the CDHs in (18a) and (18b)
for the oral fricatives /s, s’/ and the glottal fricative lb], respectively. /s, s’, h/ form a

natural class at the VT level by having the characteristic of frication or noise, as shown in

(is).

(1 8) a. /s, s'/ VT (frication or noise)

SL
[critical]

   

Oral
[critical]

Tongue
[critical]

Glo Nasal Lateral Central Lips
[closed] [critical] [open]
I l /\
tensed /s’/ [closed] VEL TB TB TT
plain /s / [mid] [critical]
/s, s'/

76

(b) /h/ VT (frication or noise)

SL
[Open]

   

Oral

[Open]
Tongue
[Open]
/\ _
Glo Nasal Lateral Central Lrps
[closed] I [open] [open]
I /\
VEL TB TB TT

[Open]
/h/

2. 1. 3. 2. 3 Gestural characteristics of fricatives in coda position

In this section, I will discuss how the [critical] CD of TT gestures for /s, s’/ and
the [open] CD of glottal gestures for /h/ are adjusted gesturally in coda position. An
output CDH, which is created by gestural adjustments of fricatives in coda position, is
illustrated in (19). The phonetic motivations for gestural adjustments of fricatives in coda
position are discussed in 2.1.3.2.4.

Kim and Jongman ‘s (1996) study is introduced first because their study provides
insight for understanding CN in Korean. In one of their experiments, V1 duration in
CV1.sV2 and CV1.t"V2 sequences is examined to see if the length of V1 is influenced by
the underlying manner of following consonants (/s, thl). This experiment uses words such
as [ki_.sa] from /kisa/ ‘newspaper article’ and [kifla] from /ki_tha/ ‘etc.’. Kim and

Jongman observe that V1 duration before [8] is longer than V1 duration before [th], as

77

vowels before fricatives within a syllable are longer than vowels before affricates in
English (Dorman and Raphael 1980, Johnson 1997, and Ferrero et al. 1982). Thus, Kim
and Jongman conclude that vowel duration is sensitive to the manner of a following
consonant in Korean.

In other experiments, Kim and Jongman (1996) investigate /t, t”, s/ in syllable and

word-final position in terms of production and perception. Kim and Jongman measure

vowel durations in words such as [mut’] from /mus/ ‘many’ and [mut’] from /mufl/

‘land’. These words are chosen to see if vowel durations in surface forms with the

identical coda consonant [t’] are influenced by underlying manners of following
consonants /s, t"/, as in [mut’] from /mus/ and /muth/. In surface forms, average vowel

durations before underlying fricatives and stops are 100.5 ms and 101.75ms respectively,
which is not a significant difference. Therefore, Kim and Jongman conclude vowel
durations in surface forms are not influenced by underlying manners of following
consonants if following consonants appear in word-final position or before a consonant.
Furthermore, in Korean, /s/ at the word-final position or before a consonant is
realized as silence corresponding to oral closure rather than relatively long turbulent

noise as in English. The average closure duration of [t1 ], which is relevant to underlying

/t, th and s/ in coda position, is 73.5 ms for /t/, 74 ms for /t”/ and 73.25 ms for /s/.

Therefore, there are no significant differences in closure according to different manners

of underlying consonants, resulting in neutralization of /t, t“, s, s’/ to [t’] in coda position.

Dorman et al.’s study (1980) is comparable to Kim and Jongman’s study (1996).

After examining /j/ in 'dish' and /tj/ in 'ditch’, Dorman et al. conclude that in English,

78

acoustic properties for fricatives in coda position are not different from acoustic
properties for fricatives in onset position. Acoustic properties of fricatives evident
regardless of syllable positions are slow rise time of fricative noise and long duration of
fricative noise.

A comparison of Kim and Jongman (1996) and Dorman et al. (1980)’s studies
indicates that different languages use different gestures for underlyingly identical
segments in the same prosodic position. For example, as discussed earlier, the [critical]
CD of TT gestures for /s, s’/ and the [open] CD of glottal gestures for /h/ are responsible

for continuous or long duration of fricative noise. In English, the CD of TT gestures for

oral fricatives /j/ in coda position is [critical], as Dorman et al’s study shows. However,

in Korean, the [critical] CD of TT gestures for oral fricatives is absent in coda position.
Instead, the [critical] CD of TT gestures is adjusted to the [closed] CD of TT gestures in
coda position, acoustically resulting in silence, as shown in Kim and Jongman’s study. In

addition, I assume that the [open] CD at the glottis for /h/ is also absent in coda position

because there are no active glottal gestures in coda position, resulting in [t’] from /h/.

Any adjustment of targeted CD gestures, in turn, creates a different CDH.
Therefore, I suggest (19) as an output CDH for underlying fricatives /s, s’ h/ in coda
position. (19) reflects gestural adjustments made in the TT and glottis in coda position.
In (19), the targeted [critical] CD of the TT for /s, s’/ and the [open] CD of the glottis for
/h/ are adjusted to [closed] CD of TT gestures in coda position because of the phonetic

implementation condition that will be discussed in 2.1.3.2.4.

79

(19) Output CDH for underlying fricatives /s, s’, h/ in _{#, C}

VT (occlusion)
r ~ .
I ‘~.~~

’I’ SL
I closed

II [I’- - - -] ----

I I

,’ ,’ Oral
:' / [closed]
I, ” a A
I I ’ ’ S‘
x Tongfié) ~.
1' ,' [closed] \\
I I’ [A 8 \ ‘\‘
Glo Nasal Lateral CEntral Lips
[closed] [closed] [closed] [open]
I I h s
. ' I ~ .
VEL TB TB TT
[closed] [closed]
[t ’ ]

In (19), there are no contrastive CD of glottal gestures for fricatives in coda
position, as three types of Korean stops do not have contrastive CDs of glottal gestures in
coda position (see (10) in section 2.1.3.1.3). The [critical] CD of TT gestures and
contrastive CDs of glottal gestures for fricative /s, s’, h/ are not achieved due to phonetic
reasons that will be discussed below. Instead, the [critical] CD of TT gestures and
contrastive CDs of glottal gestures are adjusted to the [closed] CD of TT gestures. This
adjustment changes the overall output characteristic of the vocal tract from frication to
occlusion.

In (19), the CD value at the central node is [closed] because the central node takes
the minimum CD value of TT or TB gestures because the TT and TB are connected in
series. The CD value at the tongue node is [closed] because the lateral and central nodes

are connected in series. Remember that the default CD of the tongue body gestures is

80

[open]. The predicted CD at the oral node is [closed] because the tongue and lip nodes
are connected in series. The predicted CD at the SL level is the maximum CD of the
nasal or oral nodes due to their parallel connection. However, both nasal and oral nodes
have [closed] CD in (19), therefore resulting in the [closed] CD at the SL level. The
[closed] CD at the SL level, in turn, creates occlusion, which results from either the
[closed] CD of glottal gestures or the [closed] CD at the SL level (see (16) in section

1.2.4.2).

2. 1. 3. 2. 4
Phonetic motivations for neutralization of fricatives in coda position

Now it is time to discuss why fricatives in word-final position or before a

consonant behave as they do in Korean. First, I will discuss phonetic motivations for

phonological adjustments of /s, s’/ to [t’].

Two points from previous discussion are important to understand neutralization of

/s, s’/ to [t’] in coda position. The first point is articulatory and acoustic characteristics

of [+/-release]. Articulatorily, [+/-release] refer to the presence or absence of targeted
glottal gestures after oral-closure release. Acoustically, [+/-release] refer to the presence
or absence of airflow escaping through the cral cavity after oral—closure release. The
second point is Browman and Goldstein’s (1992) observation that the peak of glottal
gestures occurs simultaneously while the [critical] CD of TT gestures for /s, s’/ is

maintained in order to create continuous frication effectively, as shown in (l 7).

81

I suggest that the primary phonetic motivation of [t’] from /s, s’/ in coda position

is closely tied to the lack of active glottal gestures in coda position for obstruents in
Korean. The glottis and TT gestures are the active articulators involved in /s, s’/ and they
co-articulate together in onset position for /s, s’/ as shown in (17). However, glottal
gestures are absent in coda position. Therefore, gesturally, there is no peak of glottal
gestures synchronized with the [critical] CD of TT gestures for /s, s’/ in coda position.
The lack of active glottal gestures, in turn, create difficult conditions for creating
turbulent noise in coda position. Therefore, the main source for fricatives, turbulent
airflow, is not present or is weakened significantly in coda position. As a consequence,
the [critical] CD of TT gestures cannot be maintained because there is no air to be
controlled to maintain continuous frication noise. Hudson (1995:662) states that syllable-
final consonants are likely to be neutralized because syllable-final consonants are not
salient acoustically. Therefore, phonetic implementation conditions for underlying /s, s’/
are weakened or impoverished in coda position. Furthermore, the lack of air escaping
through the glottis is likely to promote silence corresponding to oral closure, which is a
critical property of stops in coda position (Dorman et al. 1980), resulting in neutralization
of Is, s’ h/to [t‘].
The question raised is why the [critical] CD of TT gestures for /s, s’/ is adjusted to

the [closed] CD of TT gestures, resulting in [t’] instead of [p1 , k’]. I claim that the

adj ustment of the [critical] CD of TT gestures to the [closed] CD results from an effort to
minimize phonological adjustments. Consider the fact that the tongue tip as an articulator

for /s, s’/ is specified with the [alveolar] CL and [critical] CD as shown in (17). In the

82

adjustment of the [critical] CD of TT gestures to the [closed] CD, the articulator and CL

are still maintained in coda position.

Discussions on phonetic motivations. for /h/ to [t’] in coda position are divided
into two areas: phonetic motivations for the preference of [t’] over other unreleased
stops, [p’ , k " ], and the relationship between the phonetic realization of [t’] and the style
of speech. I will provide phonetic motivations for the preference of [t"] over other

unreleased stops, [p ’, k1 ], below. The second point will be discussed in detail in chapter

Neutralization of /h/ in coda position parallels neutralization of /s, s’/. /h/ is the
glottal fricative which is specified with the [open] CD of glottal gestures. However, as

mentioned earlier, active glottal gestures are absent in syllable-final position in Korean,

as shown in (19). Any of [p’, t’, k’] for /h/ in coda position seems an equally
reasonable candidate to appear in coda position. The lips for [p’ ], TT for [t’] or TB for
[k’] are absent in the input for the glottal fricative /h/. Therefore, any adjustments using

any of these articulators incurs violations of IDEN-IO(Gesture), which demands that
corresponding input-output segments be identical in gestures.

However, I believe that the adjustment of /h/ to [t’] rather than to [p " , k’] occurs
because of the following articulatory implementation conditions for [t’] in coda position:

First, the TT movement is quicker than the TB movement for [k’] or lip movement for

83

[p ’ ]. Second, the tongue tip is more pliable than the tongue blade and lips. Third, during

/h/, there is a tendency for the tongue tip to rest on the floor of the mouth (Byrd 1996).

Thus, it will take more time or efforts for the tongue or the lips to travel to make [k’] or

[p’] (Byrd 1996). All thisngs being equal, speakers tend to use gestures distributed in a

given context (J. Jun 1995, Kirchner 1998 and 2000). Therefore, a segment with the

tongue tip articulator, namely [t’], rather than segments with tongue body or lip

articulators, namely [p’, k’ ], is produced as in [not1 .k_h‘o] from /noh + ko/.

Thus far, it has been discussed that the absence of active glottal gestures and lack
of air escaping through the oral cavity for obstruents in coda position create articulatorily
and acoustically unfavorable conditions to maintain [open] CD of glottal gestures for /h/
and [critical] CD of TT gestures for /s, s’/. Thus CN in Korean emerges as a by-product

of phonetic implementational conditions in a given context (Steriade 1997).

2.1.3.2.5 The gesture-based account of fricatives in coda position

A task at hand is to establish a link between phonetic implementational conditions
for fricatives in coda position and fricative neutralization in coda position. Output
constraints suggested for fricative neutralization must be sensitive to releasedness of
fricatives in coda position. This is because fricative neutralization in coda position is

motivated by the lack of active glottal gestures, as stop neutralization in coda position is

84

motivated by the lack of active glottal gestures. In addition, output constraints should be
sensitive to the fact that the [critical] CD of TT gestures is not maintained due to the lack
of synchronized gestures of the glottis and the TT and the absence of airflow escaping
though the oral cavity to maintain [critical] CD of TT.

Before proposing output constraints for fricative neutralization in coda position, it
is necessary to examine output CDHs for oral stops ((10) in section 2.1.3.1.3) and an

output CDH for fricatives shown in (19). These CDHs can be united as in (20) below.

(20) Output CDHs for three types of oral stops and fricatives in _ {#, C}

VT (Occlusion)
r ~ .
I ‘~.~~
’1' SL
I, [closed]
’I I, -----
,’ ,’ Oral
:' 1’ [closed]
’ I, , A
f, I,” Tongue ’ ‘s‘
’l I/ [closed] \\
I I I ‘s ‘\
Glo Nasal Lateral Céntral Lips
[closed] [closed] [closed] [closed]
I I ”s I
I r ’ “ I
VEL TB TB TT :
[closed] [closed] [closed] [clo'sed]

[k’ l [t’] [P’l
A generalization drawn from (20) is that only occlusion, which is specified with
[closed] CD at the SL level, can appear in coda position. Occlusion is specified for either
[closed] CD of glottal gestures or [closed] CD at the SL level. However, the [closed] CD

of glottal gestures of occlusion is not active due to the lack of active glottal gestures in

coda position. Therefore the [closed] CD at the SL level plays an important role in

85

Korean. To have the [closed] CD at the SL level, the TB, TT, or Lips have to be [closed].
Therefore, there are no contrastive CDs of articulator gestures in coda position for
fricatives and stops regardless of their target gestures. For example, contrastive CDs of
glottal gestures for stops and fricatives are absent or weakened and the [critical] CD of
TT gestures for /s, s’/ are absent in the output, as shown in ( 15b). The absence of
contrastive CDs at the articulators, in turn, influences other higher levels, resulting in
occlusion at the highest level VT, as shown in (20).

Based on the generalization above, the output constraint *CCDGLOTTIS /_{#, C},
which is responsible for neutralization of three types of Korean stops in coda position, is
expanded and modified as in (21) to incorporate neutralization of stops and fricatives in

coda position.

(21) *Contrastive Constriction Degrees (CCD)/ _ {# ,C}
Contrastive CD at the glottis and'at the SL level for obstruents is not allowed
when obstruents appear in word-final position or before a consonant.
The constraint (21) requires that obstruents not be contrastive in terms of the CD at the
glottis and at the SL level before a consonant or in word-final position.
Two points must be mentioned related to the constraint (21). First, CCD in (21) is
subdivided into [open], [mid], [critical], and [closed]7 (Browman and Goldstein 1989:
209). Languages use different CDs of an articulator depending on obstruents involved.

For example, Korean uses [open], [mid] and [closed] CDs of the glottis for obstruents,

 

7 According to Browman and Goldstein (1989), [narrow] CD comes between [critical] and [mid] and they
say that [narrow] is useful in differentiating vowels.

86

resulting in aspirated, lax, and tensed obstruents respectively. Contrary to Korean,
English uses [mid] and [critical] CDs of the glottis for obstruents, resulting in voiceless
and voiced obstruents respectively. In addition, an articulator cannot have two CD values
simultaneously. Therefore, stops cannot be aspirated and tensed at the same time because
the glottis cannot have [open] and [closed] CDs simultaneously.

Second, the comparison of (18) to (19) indicates that within a language there is
preference for one CD over another in a given context. Take an example from [critical]
CD of TT gestures for fricatives. The [critical] CD of TT gestures for /s/ is contrastive
with the [closed] CD of TT gestures for /t/ in onset position in English and Korean.
However, in Korean, only the [closed] CD of TT gestures is allowed in coda position. In
English, both [critical] and [closed] CDs of TT gestures for /s, t/ are allowed in coda
position.

I modify (21) to include language difference and CD preference in a given context

as in (22).

(22) *Contrastive Constriction Degrees (CCD)/ _ {#, C}

Contrastive CDs at the glottis and at the SL level for obstruents are not allowed
when obstruents appear in word-final position or before a consonant.

*[open] /_{#, C} >>*[mid] /_{#, C} >>*[critical] /_{#, C} >> *[closed] /_{#, C}

The fact that the [closed] CD of TT gestures is allowed instead of the [critical]
CD of TT gestures in coda position ranks the constraint *[closed] /_{#, C} to be the
lowest in the ranking hierarchy. The fact that /h/ is not allowed but a sonorant is allowed

in coda position in Korean and English justifies the ranking of *[open]/_{#, C} over

87

[critical]/_{#, C}. As shown in (23), CDs at the SL level for /h/ and a sonorant are [open]

but CDs of glottal gestures are [critical] for a sonorant and [open] for /h/.

(23) a. sonorants b. /h/
VT VT
/\ /\
GLO SL GLO SL
[critical] [open] [open] [open]

A A

However, English and Korean do not allow [h] at word-final position or before a
consonant. Therefore, the [critical] CD at the glottis is preferred over the [open] CD of
glottal gestures in coda position. Therefore, *[open]/_{#, C} dominates *[critical]/_{#,
C} in the ranking hierarchy. In English, /h/ is not allowed in coda position but voiceless
obstruents appear in coda position. Therefore, *[mid]/__{#, C} is located between
*[open]/_{#, C} and *[critical]/_{#, C}.

The phonetic motivation for CN and phonetic based constraints having been
introduced above, it is now possible to apply the constraints and their rankings to actual
data and evaluate possible candidates.

Languages differ from each other by the ranking of IDENT-IO(Gestures) ((12) in
section 2.1.3.1.3) related to *CCD/_{#, C} (22). If IDENT-IO(Gestures) is ranked higher

than *[mid]/ _{#. C} in a language, the language would allow fricatives such as [f, v, s,

z, 0, 6 j, 3] and plain stops such as [p, t, k] but not [h] in coda position as in English.

I propose that IDENT-IO(Gestures) is ranked lower than *[closed] /_{#, C} in

Korean as in (24) because there are no fricatives or stops with contrastive CDs in coda

88

position. I will show below that CN in Korean results from the effort to satisfy output

constraints (24) most optimally.

(24) *[open] /_{#, C} >>*[mid] /_{#, C} >>*[critical] /_{#, C} >>
*[closed] /_{#, C}>> IDENT-IO(G)

(25) shows evaluation of candidates from a lexical entry ending with the fricative
/s/. In (25), *CCD/_{#, C} is violated if a consonant in coda position has contrastive
CDs either at the glottis or at the SL level. This is because the output airflow of VT is
determined by CDs at the glottis or at the SL level when the right amount of airflow from

the lungs is assumed.

(25) /na§/ ‘sickle’

GLO [mid], SL [critical], Oral [critical], TT [critical]
/na§/ *[open]/ *[mid]/ *[critical]/ *[closed]/ IDENT-IO (G)
_{#, C} _{#, C} _{#. C} _{#, C}
a. [nas] *!(GLO) *(SL)
b. LnaS’L *!(SL) *(GLOI.'tiGLO:1§lctds‘ed]."‘
>c. [na]
d [nat] *! (GL0) 91'fEQ-775‘f*(SL)*GL0lmtdl
. :iff: *TTIclosedl
6. [nat’] *! (GL0) 1:57;?jS "i’ff-iQil‘féi‘ i *(SL) "GLOilloti'b‘éfili‘ff
f. [nat’] *!(GL0) ‘21.fGLOiV-iclOsedlt;s:
*(SL) if.¥TT‘:’-,[closed]A,.75%
‘7’ g. [nat ’ ] *(SL) *TT[closed]
> indicates to the most optimal candidate with respect only to constraints given.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Any output candidate that has CCDs of glottal gestures is not the most optimal

candidate because of the lack of active glottal gestures in coda position, as discussed in

’89

neutralization of three types of stops in coda position. Contrastive glottal gestures are
impoverished or not observable in coda position. Thus, (25a), (25b), (25d), (25c), and
(25f), which have contrastive [open], [mid] and [closed] of glottal gestures, are not
optimal candidates. In addition, candidates with [critical] CD of TT gestures in coda
position as in (25a) and (25b) are not desirable to be optimal candidates. The lack of
active glottal gestures in coda position creates unfavorable conditions for completing
[critical] CD of TT gestures and creating continuous airflow for fricatives. Therefore,
among output candidates, only (25c>) and (25g?) are possibilities to be the most
optimal candidate. (25c>) deletes the underlying consonant /s/ in coda position. In (25g
'3’“), [critical] CD of TT gestures and [mid] CD of glottal gestures in the input are
adjusted to [closed] CD of TT gestures without any active glottal gestures, resulting in

the unreleased stop [t ’ ].

According to the constraint hierarchy suggested in (22), the candidate (25c>)
satisfies the constraint ranking most optimally. The constraint IDENT-IO(G) is satisfied
vacuously because there is no corresponding segment for input /s/ in (25c>). However,
no consonants in the input are deleted in Korean unless there are two consonants in coda
position, as will be shown in Coda Consonant Cluster Simplification (section 2.2).
Therefore, there is a constraint that prevents an input segment from being deleted.

In Correspondence Theory, a constraint that prevents a segment from being
deleted is Max-IOseg in (26), which requires every segment of 51 have a correspondent in
S; The comparison of (25c>) and (25g?) indicates tMax-IOseg is higher than

*CCD/_{#, C} as shown in (26) because no segment is deleted to avoid contrastive CDs

in the vocal tract.

90

(26) Max-10568: Every segment of the input has a correspondent in the output.
(From McCarthy and Prince 1995:16)

Max-Io...g >> *ccr) /_{#, C} >> roam-tom)

Inas/ ‘sickle’
GLO [mid], SL [critical], Oral [critical], TT [critical]

 

 

 

 

/na_s_/ | Max-103., *cco /_{#, C} mam-10(6)
rb. [nat‘] I *(SL)*TT[closed1

 

 

 

 

The most optimal candidate in (26) has [closed] CD of TT gestures, resulting in

the unreleased stop [t’], as shown in (26b). The adjustment of [critical] CD of TT

gestures to [closed] CD of TT gestures also changes CDs in higher levels, creating
[closed] CD at the SL level. Therefore, the constraint hierarchy of Max-IOscg above

IDENT-IO(G) is responsible for gestural disparity (or phonological adjustment of a
segement) between the input and the output, rather than deletion of a segment.

The next step is to determine if the constraints suggested in (26) can correctly
predict neutralization of /h/ in coda position in Korean. (27) illustrates a lexical item with

/h/ before a consonant.

91

(27) /nohko/ ‘and place (verb)’

/h/: GLO [open], SL [open], Oral [open]
/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]

 

/no_h+ko/ Maxie“, *CCD/_{#, C} roam-row]

 

 

*[open] *[mid] *[critical] *[closed]
a. nohko *!(GLO) -‘

 

m— 0 * (SL) ( T) ~

 

c. ncgkho "fKGLQI 7*?! (SL) j-Iiijs'}:*(rrfi'jixigg;

 

 

d. noth’kho f"!( GL0) ~ *(SL) ,- *(TT)
*! - * '1 , ‘1‘ 3 ' 43 . 7 3‘ ‘ in; ‘ [[1212 ;.

e. no.ko " “319

 

 

 

 

 

 

 

 

 

 

c:> indicates free variation with the output depending on the style of Speech 3 ~ . . . ‘

Although the output candidate (27a) maintains all input segments, it violates a
relatively high *[open]/_{#, C} constraint because it has the [open] CD at the glottis in
coda position. Candidates (27b), (27c), (27d), (27c), and (27f D) violate one of
*CCD/_{#, C} because they have either contrastive CDs of the glottis or [closed] CDs at
the SL level in coda position. In addition, candidates (27b), (27c), (27d), (27c), and (27f
C9) violate IDENT-IO[G] because these candidates have TT gestures that are absent in the
input. I mark insertion of TT gestures as one violation with respect to IDENT-IO[G].

Candidates (27b) and (27f 51>) violate a relatively low *[closed]/_{#,C} constraint.

However (27b) violates *[closed]/_{#,C} twice and (27f D) violates *[closed]/__{#,C}
once. [not’kho] (27f=>) is considered as a free variant of (27g) depending on the rate or
style of speech in Korean. Considering that gestures are spatio-temporal linguistic units,
the rate or style of speech will influence how gestures are co-articulated or co-produced

at a given time. Constraints related to these free variations are discussed in detail in

Chapter 4.

92

 

Candidate (27e) is far from being an optimal candidate because' there is no

correspondence for [h] in output. The comparison of candidate (27g?) to candidate

(27f'=i>) shows that (27g? ), in which /h/ and /k/ are coalesced into [kh], is the most

optimal candidate because both gestures of /h/ and /k/ are maintained in the output.

In order to include neutralization of stops and fricatives, constraints and the

constraint hierarchy in (22) are expanded and modified from (11) in section 2.1.3.1.3.

Therefore, constraints (22) are examined below for three types of Korean stops to see if

constraints and their hierarchy in (22) select the output correctly.

Lexical entries in (28) have aspirated and lax bilabial stops at word-final position.

(28) A. /iph/ ‘leaf’ From (13)

GLO [0

en], SL [closed], Oral [closed], Lips [closed]

 

/igh_/

Max-

*Contrastive CD/ {#, C}

 

103%

*kmmfl

*DnM]

* [critical]

* [closed]

Imam-row]

 

or
a.rp

*KGLO)

 

b.ip’

H_’G“J

*KGLO)

(GLO)

 

c.ip

*KGLO)

(GL0)

 

 

G’dip"

 

 

 

 

 

 

 

i’lQLQIfiiii7 

 

B. /ip/ ‘mouth’ From (13)

GLO [mid], SL [closed], Oral [closed], Lips [closed]

 

/ip/

Max-
1033}

*Contrastive CD / {# , C}

mam-tom]

 

flbmml

*[mid]

*[critical]

*[closed]

 

3. ip11

*!(GLO)

' " .. .1} fjf(SL)f

«23:12:;mm

 

b.ip’

*(SL)

*KGLO)

(919

 

c.ip

*KGLO)

— ~ (SL)

 

 

¢’d.ip’

 

 

 

 

*(SL)

 

 

 

93

 

Any candidates with segment deletion are winnowed from being most optimal
because they violate a relatively high constraint MAX-108.3. Any candidates with CCDs
of the glottis violate *CCD/_{#, C} because glottal gestures cannot be contrastive in coda
position due to the lack of glottal gestures. In candidates (28Ad) and (28Bd), contrastive
CDs of the glottis are absent in coda position. Compared to (28Ab) and (28Bb),
candidates in (28Ad) and (28Bd) violate *CCD/_{#, C} only once because of the [closed]
CD at the SL level in coda position. In other candidates such as (28Aa), (28Ba), (28Ac),
and (28Bc), relatively higher constraints within *CCD/_{#, C} constraints are violated.
Therefore, (28Ad) and (28Bd) are the most optimal candidates.

Up to this point I have shown that CN of stops and fricatives can be analyzed
systematically in terms of CD, which is closely related to manner of articulation.

In previous studies of Korean phonology, CN of stops and fricatives was
described as delinking of different features such as [spread glottis], [constricted glottis]
and [+continuant]. As a consequence, it has been difficult to analyze neutralization of
stops and fricatives in a unified and systematic way. The previous analyses based on
traditional Feature Geometry Theories at times hinder linguists in finding regularities,
generalizations or parallels among different phonological phenomena. This is because it
has been difficult to find phonetic reasons why features such as [spread glottis],
[constricted glottis] and [+continuant] are involved in CN together in Feature Geometry

theories.

94

2.1.3.3 Patterns of affricates in coda position

2.1.3.3.1 Introduction

Data (29) shows neutralization of affricates /c, ch/ to the unreleased [t’] in coda

position. No words end with the tensed /c’/ in Korean.

(29) a. Word final position
/nac/ -> [nat’] ‘day’
/nacl’/ —> [nat’] ‘face’
c. Between two vowels
/na§+£/ —> [nage]
/ magil +e/ —> [nagsil e]

b. Before a consonant

/nag_:c’am/ ——> [na_t1 . c’am] ‘nap’

/nac’_‘ kalim/ —> [nat’ . k’a.rim] 8
‘shyness of children’

‘in the daytime’

‘in the face’

According to previous studies (Cho 1990, Sohn 1987 and Kim-Renaud 1974), the

delinking of [+constricted glottis], [+spread glottis] and [-anterior] as in (30) is

responsible for neutralization of affricates /c, ch/ as unreleased [t’] in coda position.

 

8 /1/ becomes /r/ between two vowels in Korean.

95

(30) Patterns of affricates in coda position

Cho (1990: 106)
0 Root node

Laryngeal " Place

 

[in ..

[-anterior]

In this section, I will argue two points. First, there is no need for delinking two
separate features for patterns of affricates in coda position, as show in (30). I claim that
because CN is neutralization in manners of articulation in coda position, CN should be
analyzed without reference to a place feature such as [-anterior].

Second, neutralization of /c, c’, ch / as unreleased [t’] in coda position is in parallel

with neutralization of stops and fricatives in coda position. Neutralization of affricates
results from the most optimal satisfaction of output constraints in (22), as does

neutralization of stops and fricatives.

96

2.1.3.3.2 Gestural characteristics of affricates

In order to understand patterns of affricates in Korean phonology, gestural
characteristics of affricates must be understood first. According to Kehrein (1999),
affricates /c, c“, c’/ result from promoting acoustic transparency by changing the manner
of stop /t/. Kim (1997) claims that Korean affricates are characterized as [-continuant,
+strident] in the alveolar location. Based on these studies, I assume that TT gestures for
affricates are specified with two different CDs: [closed] and [critical] at the TT level, as

shown schematically in (31).

(31) Gestural Score of affricates

—> Time

TT

 

[closed] [criticfl
(
/

 

\ 7' [open] for /ch/
GLO \ - - [mid] for /c/
gr [closed] for /c’/

The two different CDs of the TT are co-articulated with three different CDs of

 

 

 

 

 

 

the glottis depending on types of affricates: [open] CD of the glottis for /c"/, [mid] CD of
the glottis for /c/ and [closed] CD of the glottis for /c’/.

Affricates begin with oral closure as do stops, as shown in (31). However, the
collected air behind an oral closure escapes through the oral cavity in various ways. In the

case of stops, the collected air behind an oral closure escapes through the glottis right afier

97

oral-closure release. In the case of affricates, the collected air behind the TT closure is
held continuously by the [critical] CD of TT gestures in order to create intense frication
for the latter part of affricates.

[closed] and [critical] CDs of TT gestures and constrastive CDs of glottal gestures

for affricates create the CDH in (3 2).

(32) CDHs for affricates /c, c’, ch/
VT (occlusion/frication or noise)

SL
[closed/critical]

 
   

Oral

 

[closed/critical]
Tongue
[closed/critical]
/\
Glottal Nasal Lateral Central Lips
[clo ed] [closed/critical] [open]
I I /\
tensed /c’/ [closed] VEL TB TB TT
plain /c / [mid] [closed/critical]
aspirated /ch/ [open] /c’, c, ch/

In (32), at the highest level, VT, the overall characteristic of output airflow is
either occlusion or frication depending on which CDs of TT gestures are counted. This
may be understood by reference to controversial issues concerning the representation of

affricates discussed by Sagey (1986), Lombardi (1990), Rubach (1994), and Steriade

(1989).

98

2. 1. 3. 3. 3 The gesture-based account of affricates in coda position

In this section, I will discuss adjustments of target gestures of the TT and the
glottis for affricates in coda position. In addition, phonetic motivations for gestural
adjustments of affricates in coda position are discussed. Gestural characteristics of
affricates mentioned above allow us to understand why affricates in coda position behave
as they do in Korean.

As stops and fricatives in coda position are faced with adverse conditions for
completing release gestures, affricates in coda position are also faced with adverse
conditions for completing release gestures.

One of the disadvantages of affricates in coda position is that there are no
observable glottal gestures at oral-closure release or at the time of [critical] CD of TT
gestures. The absence of glottal gestures at the time of TT gestures for affricates creates
unfavorable conditions for distinguishing three types of affricates.

In addition, the lack of glottal gestures in coda position decreases
implementational conditions for achieving the [critical] CD of TT gestures that is
associated with the latter part of affricates. During the latter part of [critical] CD of TT
gestures, there should be a fairly large amount of air escaping through the oral cavity in
order to create the acoustic characteristics of frication. However, the aerodynamic
conditions for frication are not met for affricates in coda position due to the lack of

airflow escaping through the glottis. When there is no airflow after the [closed] CD of TT

99

gestures is formed, [closed] CD of TT gestures for affricates in coda position are
maintained. Therefore, there is no need for delinking [-anterior] as in (30).

Therefore, neutralization of affricates as the unreleased alveolar stop [t’] in coda

position emerges from phonetic implementational conditions, which result from
adjustments of target gestures of affricates in coda positoin. (33) represents the output

CDH that reflects adjustments of target gestures of affricates in coda position.

(33) An output CDH for underlying affricates /c’, c, ch/ in _ {#, C}

VT (Occlusion)
I. ‘ ~‘ ~
,I’ " SL
,’ [closed]
I I ‘ ~ ~ ‘
I, I, ~‘ ‘ \.
[I I,’ Oral
,' [I [closed]
I, [I ‘ A
I I , l ’ ‘ \\
I, I, Tongue) ‘ \ \
,’ ,’ [closed] \
'I I ’A‘ \\|
GI ottal Nasal Lateral] CEntral Lips

[closed] [closed] [closed] [open]
I I " x
I I ’ ‘ ~

VEL T'B TB TT
[closed] [closed]
[t ’1

As shown in (33), there is parallelism in the behavior of stops, fricatives and
affricates in coda position. These obstruents are neutralized as unreleased stops in coda
position due to the lack of critical gestures such as contrastive CDs of glottal gestures and

the [critial] CD of TT gestures. Stops in coda position lack contrastive CDs of glottal

100

gestures. Fricatives in coda position lack contrastive glottal gestures, which are
synchronized with TT gestures. In addition, it is difficult for fricatives to maintain the
static [critical] CD of TT gestures due to the lack of air escaping through the glottis.

Considering that gestures of affricates are a combination of gestures of stops and
fricatives, parallelism in patterns of stops, fricatives, and affricates in coda position is
expected in a gestural approach. Affricates in coda position also lack contrastive glottal
gestures as well as TT gestures for the same phonetic reasons that stops and fricatives in
coda position do.

If affricates in coda position pattern together with stops and fricatives for the
same phonetic reasons that stops and fricatives in coda position do, constraints suggested
in (26) should correctly select the most optimal candidate for affricates in coda position.
Constraints and their hierarchy suggested in (26) are evaluated below in (34) for affricates

in coda position to see if they can correctly predict the output.

(34) /c’oc§/ ‘flower’

GLO [open], SL [closed/critical], Oral [closed/critical], TT [closed/critical]

 

    
   
 

 

I /c’o€/ *CCD/_{# or C} | IDENT-
' *[closed] IO[G]

 
   
 
 

 

      
 
   

 

 

 

 

 

 

 

 

 

 

 

 

 

H (GLO) (fill-ii.» '1 q . A a ,7 '

* ! (GLO) a}: GLO: [closedlf

* (SL) It"; 'I'I‘Icritical]. 21$

* ! i jjil ,_~t}:u(.SL)jg 7'.¥»GL0:[mid]

(GLO) “ '. f f . V “"‘j:"IT:[critical].4~

.___d- C’OL" | . *KGLO) 2 * I g :7 ; ° *(SU Y ;‘TT=lcriticall'-'i

e. c’o *! , 3 '_ - ‘
(ff. c’o];1 * (SL) * TI’:[critical]

 

101

Candidate (34a) is not the most optimal candidate because it has the [open] CD of
glottal gestures and the [critical] CDs at the SL level. The [critical] value at the SL level is
predicted from [critical] CD of TT gestures. IDENT-IO[G] is satisfied best in (34a)
because it does not change input gestures. Candidate (34b) is not an optimal candidate, as
it has the contrastive [closed] CD of glottal gestures. It also has the contrastive [closed]
CD at the SL level, violating *CCD/_{#, C}. Candidate (34c) is not an optimal candidate
because it has the contrastive [mid] CD of glottal gestures. (34d) is not an optimal
candidate because of the contrastive [open] CD at the glottis and the [closed] CD at the
SL level. Candidate (34e) is the least optimal candidate because it deletes /ch/. (34f)
violates *CCD /_{#, C} because of the [closed] CD at the SL level. However, violation of
*[closed]/_{#, C} is preferred to violation of Max-IOseg because Max-IOSeg is higher
ranked than *[closed]/_{#, C}. Therefore, constraints and their hierarchy in (26)

correctly select the most optimal candidate for affricates in coda position.

102

2.1.4 Summary

Hudson (1995) and Steriade (1997) claim that some phonological phenomena are
understood better when phonetic characteristics of a segment in a context rather than
prosodic units such as the syllable are taken into account in phonological descriptions.
Especially, Hudson (19952660) claims that “as a phonological feature with generally
recognizable phonetic characteristics, [released] may explanatorily replace the analytical
syllable boundary in phonological rules.” The analysis of CN in this study has provided
one more example to strengthen Hudson’s claim.

In this study, based on experimental studies in perception and production,
releasedness of obstruents is defined in terms of gestural and acoustic properties.
Gesturally, [+/- release] refers to the presence or absence of active glottal gestures at oral-
closure release. Acoustically, [+/- release] refers to the presence or absence of airflow.

Based on the fact that consonants are likely be unreleased in word-word final
position or before a consonant (Hudson 1995) and experimental studies of Korean and
English obstruents in word-final position (Kim-Jongman 1996, Dorman et al. 1980), I
propose the constraint *CCD/_{#, C} in (22), which is sensitive to context and
releasedness of obstruents. In the constraint *CCD/_{#, C}, the role of releaseness is
expressed in terms of CDs. This is because [+/-release] is closely associated with active
CDs of glottal gestures at the time of oral-closure release. For example, in Korean there
are no active glottal gestures at the time of oral-closure release, thus resulting in

neutralization of three types of stops as unreleased stops [p1 , t1 , k’] in coda position.

The unreleased stops are characterized by oral-closure gestures only, as shown in (10).

103

In addition, the presence or absence of active CDs of glottal gestures at the time
of oral-closure release consequently influences gestures of other articulators that are co-
articulated with glottal gestures, as explained in section 2.1.3.2.4. Therefore, by
suggesting the constraint *CCD/_{#, C}, which is sensitive to CDs in the VT, I have
shown that neutralization of obstruents in coda position, namely CN, can be analyzed in a
unified way.

In addition, I have claimed that CN is neutralization in manners of articulation,
which is closely related to the CD of the vocal tract. Therefore,as discussed in 2.1.2,
there is no need of suggesting different rules for stops, fricatives and affricates that delete
different features. Furthermore, deleting of [+anterior] is not relevant to CN because CN

as neutralization of manners of articulation is related to CDs not CLs.
I have claimed that interaction of *CCD/_{#, C}, Max-10863, and DENT-IO[G] is

responsible for CN in Korean. If obstruents are in coda position, only the [closed] CD at
the SL level is allowed in Korean due to the-lack of active glottal gestures. The lack of
active glottal gestures in coda position forces DENT-IO[G] to be violated. However,
having the [closed] CD at the SL level is preferred to deletion of a segment as long as
Korean syllable structure in (19) of section 1.3.1 is maintained. Therefore, satisfaction of
Max-IOscg takes precedence over satisfaction of *CCD /_{#, C}, as shown in (26).

In Korean literature, the analyses of CN can be divided into two types of
approaches according to features used. In the first approach, features used in the analysis
of CN are distinctive features used in Chomsky and Halle (1968) or Jackobson et al
(1952), as in Sohn (1987) and Cho (1990). In the second approach, phonetic features

such as [release] and A0 are used in the analyses of CN, as in Kim-Renaud (1974) and

104

Lee (1994). A0 refers to zero aperture of the oral cavity that stands for oral closure in
Steriade (1991).

The analyses of CN based on releasedness of obstruents have been attacked for
the reason that no sounds are contrastive in terms of releasedness in languages. However,
Hudson (19952661) suggests that [+/-unreleased] is contrastive in Hindi. In addition,
when releasedness is defined in terms of CDs of articulatory gestures, especially CDs of
glottal gestures, as in this study, the attack on the analyses of CN based on releasedness
of obstruents can be avoided. This is because, in Articulatory Phonology, CDs of
articulator gestures function like distinctive features of previous phonological theories.
Therefore, languages are distinctive in terms of the presence or absence of CDs in
articulatory gestures or the degree of constriction, as shown in (5).

In addition, representation of releasedness of obstruents in terms of CDs offers
simplicity in phonological descriptions. For example, Kim-Renaud (1974) describes CN

as in (35a):

(35) Kim-Renaud(l9742 116)

a. [+consonant, -sonorant] —9 [-release] /_ {$, [-release] } : $ syllable boundary
(one or more obstruents are unreleased in syllable-final position)

b. [-release] -—> [-aspiration, -tense, -delayed release, -continuant]
(An unreleased segment is unaspirated, non-tensed, non-fricated and
non-affricated.)

c. [+coronal, -anterior] —> [+anterior]

(An unreleased coronal segment is dental)

105

(35a) is suggested for unreleased obstruents in coda position. In addition to (35a),
Kim-Renaud suggests (35b) and (35c) to take care of residues that cannot be fully
explained by (35a) only. Basically, (3 5b) defines what the characteristics of non-released
obstruents are. (35c) is suggested for non-occurrence of /c, c’, ch/ in coda position.
However, like CN analyses based on distinctive features, it is difficult to see phonetic
reasons why [-release] of obstruents is related to [-aspiration, -tense, -delayed release,
-continuant] and [+anterior].

In this study, unreleased stops in coda position are closely related to the absence
of contrastive glottal gestures in coda position due to the lack of active glottal gestures in
coda position. The absence of contrastive CDs in coda position is also closely related to
[-aspiration, -tense, -delayed release, -continuant] in coda position because these features
are associated with CDs in the vocal tract. Therefore there is no need for (35a) and (3 5b).

In addition, non-occurrence of /c, 0’, ch/ in coda position is closely related to the
lack of airflow escaping through the oral cavity at the time of the [closed] CD of TT
gestures, which is the first part of TT gestures for affricates. The [critical] CD of TT
gestures for the latter part of affricates is not achieved due to the lack of airflow needed
to create turbulent noise, thus the [closed] CD of TT gestures is maintained. However,
CL of TT gestures for affricates are not altered in coda position in this study. Therefore,

there is no need to suggest (3 5c) in Articulatory Phonology.

106

2.2 Coda Consonant Cluster Simplification (CCCS)

2.2.1 Introduction

Input morphemes can have two consonants in Korean. If output forms in (36a)
are assumed to be input forms, there is no way to explain why [s’, s c’, ph] are inserted
before a vowel suffix, as shown in (36c). However, if input forms have two consonants,
as in input forms in (36a), output forms in (36a) are explained as deletion of a consonant
in syllable-final position while output forms in (36c) maintain input segments [s’, s c’,
ph]. Because input morphemes ending with two consonants can explain output forms in
both (3 6a) and (3 6c), input morphemes should end with two consonants as in (3 6).

When an input morpheme ends with two consonants, only one of them is

maintained in syllable-final position, as shown in output forms of (3 6a) and (36b).

(3 6) Input Output

a. /nol_<_s_/ [nok ’ ] ‘ soul’
/kap;s/ [kap’ ] ‘price’
/wekol_s/ [we.gol_] ‘stubborn’

b. Input Output
/n91_(§+kwa/ [nak1 . k’wa] ‘soul + and’
lane New Stem +/ta/d¢clmfivc [ant’ a] ‘to sit down + declarative’
lilpf/vcrb Stem +/ta/ declarative [in1 .t’ a] ‘to recite + declarative

c. Input Output
Mafi/ + /i/Subject Marker [nok1 .s’i] ‘soul’
”(313/ + /i/Subject Marker “(321 ° S’I] ‘pI‘ICC’
/wekols_u/ ‘ [we.gol. su] ‘stubborn’
/ ani/ Verb Stem + / a] Command Marker * [330,3] ‘ Sit down,
0132/ Verb Stem + / 8/ Command Marker [if-P119] ‘rCCite ‘

*Either /a/ or /e/ is selected depending on the vowel in the verb stem.

107

If morphemes ending with a consonant cluster are followed [by morphemes
beginning with a consonant, a sequence of VlquCMCan is created, as shown in (36b).
In the sequence of VlCn-2Cn-1CnV2, the first consonant Cn before V2 always appears as an
onset consonant of V2 in output forms, as shown in (3 6b). However, only one of CW2 and
CM appears as a coda consonant of V1, as shown in (3 6b). The choice between CM and
CM after V1 depends on gestural characteristics of Cm; or CM and dialects.

Data in (37) illustrates patterns of Coda Consonant Cluster Simplification (CCCS)

in Standard Korean (SK), known as Seoul Korean.

(3 7) Patterns of CCCS in SK
Input Output
a. kS /ngl£§/ [119k 1 ] ‘ SOLll ’
ps /kaps/ [kap’ ] ‘ price’
nc lanc+ta/ [ant’a] ‘to sit down + declarative’
nh /am+ko/ [ankho] ‘not to do + and’
/mam+so/ [mans’o] be a lot + statement’
lh /om+ko/ [olkho] ‘be right + and’
/o]_l_r_+so/ [ol_.s’o] ‘be right + statement’
18 /wekol_s/ [we.kol] ‘stubborn’
b. lph /ilp’:+ta/ [53" .t’a] ‘recite + declarative’
lp /n912+ta/ [1192‘ 133] ‘be wide + declarative’
lk /i&+ta/ [i511] ‘to read + declarative’
c. lm /salm/ [sam] ‘life’
/talm+ ta/ [tam.t’a] ‘to resemble + declarative’
(1. 1th /hal_fi+ta/ [hal_,t’ a] ‘lick + declarative’

When fricatives and affricates compete for a coda position with stops, nasals, and
/l/, fricatives never appear in coda position, as shown in (37a). Either an unreleased stop

or a sonorant appears in coda position. (37b) shows that when the lateral /1/ and stops

108

except /th/ compete for the coda position, unreleased stops appear in coda position instead
of [I]. When /1/ and /m/ compete for the coda position, /m/ is preferred over the lateral /l/

as shown in (3 7c). Note that in Korean, there are no input morphemes ending with /1n/,

/nl/, /lIJ/, or /rJl/, which further verifies a preference for nasals over the lateral /l/ in coda

position. (37d) shows that in a /l/-/th/ cluster, the lateral [1] appears in coda position

instead of the unreleased stop [t1 ]. When (37d) is compared with (3 7b), (37d) seems to

be an exception to the generalization that unreleased stops are preferred over the lateral
[1], as in (37b). However, the analysis of CCCS in this study will show that the

preference for the lateral [1] over the unreleased stop [t’] from a /l/-/th/ cluster is the

expected pattern when gestures involved in /l/ and /t”/ are taken into account. One point
to be mentioned regarding (3 7) is that all consonant clusters include a consonant whose
articulator is the TT. Consonants with the TT articulator are /s, n, l, c, th/.

There are two facts observed in CCCS regardless of dialect. First, consonants in

word-final position or before a consonant are either sonorants [m, n, I], l], or occlusions

.1

[p , t’ , k’], as mentioned in section 2.1. Therefore, consonants in coda position are

characterized with the following CDHs in output forms:

(3 8) Consonant CDHs in coda position in Korean.

a. Sonorant [m, n, r], 1] b. Occlusion [p’, t", k’]
YT YT
GLO SL GLO SL
[critical] [open] [critical] [closed]
‘1’.’--::2> ¢{:--.\:Zs

109

As shown in (38), Korean consonants in coda position have either [open] or
[closed] CDs at the SL level. (38a), in fact, is ambiguous whether the [critical] CD of the
glottis refers to the default CD of the glottis or the active [critical] CD of the glottis for

sonorants. If the [critical] CD of the glottis refers to the default CD of the glottis in coda

position, sonorants [m, n, r], 1] could be represented with [m’ , n’, r]1 , 1’]. This is
because [-release], which is represented by [’], is closely associated with the lack of

active glottal gestures while targeted oral gestures are completed. If the [critical] CD of

the glottis refers to the active [critical] CD of the glottis, sonorants [m, n, 13, l] are

represented with [m, n, 13, 1]. No matter whether the [critical] CD of the glottis in coda

position is the default CD or the active CD for sonorants, the fact is that the CD of the
glottis for sonorants is [critical].

Recall that in Articulatory Phonology, the default CD of the glottis for obstruents
is [critical]. As mentioned in 2.1.3.1.3, unreleased stops in coda position have the
[critical] or narrow CD of the glottis by default, which is similar to the [critical] CD of
voiced obstruents. However, unreleased stops in coda position tend to be voiceless
because of heightened sub-glottal pressure and the lack of sufficient airflow, which is
required for obstruents to be voiced (Ohala 1983).

The second fact observed in CCCS, regardless of dialect, is that the syllable

structure in (3 9) is strictly maintained in output forms.

110

(3 9) Korean syllable structure

0'

A

(C) (G) v (C) G: Glides

If morphemes ending with a consonant cluster appear in word-final position or
before a consonant, as shown in (3 6), the need to fulfill the syllable structure (39) creates
competition between two consonants for the coda position.

As mentioned in the previous section, no consonants in the input are deleted
unless there are two consonants in coda position. Max-10368, which demands that every
segment of the input has a correspondent in the output, is responsible for maintaining of
input segments in output forms. However, the constraint Max-103.:g is violated when a
morpheme ends with a consonant cluster or when a morpheme ending with a consonant
cluster is followed by a morpheme beginning with a consonant, as shown in (36). The
violation of Max-IOscg stems from conflicting demands between Korean lexical
representation and Korean syllable structure in (3 9).

The following section 2.2.2 investigates how these conflicts are resolved in
Korean grammar and how SK and Kyung-Sang Korean (KSK) differ from each other.

Section 2.2.3 examines how a consonant that fails to appear in coda position influences

the following consonants.

111

2.2.2
The gesture-based account of Coda Consonant Cluster Simplification

2.2.2.1 The gesture-based account of CCCS in SK

As a start, (40) evaluates output forms from a lexical entry ending with a stop-
fricative cluster against constraints suggested in (26), because both stops and fricatives
are obstruents. Recall that the constraint *CCD/_{# or C} is suggested primarily for
obstruents in coda position. Therefore, it is expected that constraints and their hierarchy
in (26) select correct output forms in an Obstruent-Obstruent cluster because two
obstruents compete for the coda position. Due to space, hereafter, target gestures of

consonants in a consonant cluster are given above tableaux, as in (40).

(40) Max-IOscg >> *CCD/_{# or C} >> IDENT-IO[G] (from (26))
/kap§/ ‘price’

/p/: GLO [mid], SL [closed], Oral [closed], Lips [closed]
/s/: GLO [mid], SL [critical], Oral [critical], TT [critical]

 

 

 

 

 

 

 

 

 

/kap_s/ Max- *CCD/ {# or C} DENT-IO[G]
103,,g *[open] *[mid] *[critical] *[closed]
>a. kags **!<GL0> *(SL> : ' *(SL) : *
web, kap‘ * *(s1.) *GLO
c. 1g; * “(GLO) » «sup n» a _
d. kag * ”(GLO) [7 - *(SL)
e. ka **!/p,s/ 7 ' . . .' l‘ .
f_ k3!“ * *(SL) *lTT(closed)
*GLO

 

 

 

 

 

 

 

 

112

/p/ is specified with GLO [mid] and SL [closed]. The [closed] CD at the SL level
for /p/ is predicted from the [closed] CD of Lip gestures. /s/ is specified with GLO [mid]
and SL [critical]. The [critical] CD at the SL level for /s/ is predicted from the [critical]

CD of TT gestures.

Candidate (40a>) satisfies the highest ranked constraint, Max-IOseg given in (40).
However, (40a) is not the actual output. The actual output is (40b?) which violates
Max-10368.

According to constraints and their hierarchy suggested in (26), (40b), (40c), (40d),

(40c), and (40f) should not be selected as the most optimal candidate. They violate the
highest constraint Max—10368. One of two consonants in a consonant cluster does not
have a correspondence in (40b), (40c), (40d), and (40f). Candidate (40e) is the worst

candidate because it violates the undominated constraint, Max-IOseg twice.
In the case of (40f), it is hard to determine if the unreleased [t’] corresponds to
either /p/ or /s/. If [t"] in (40f) corresponds to /s/, instead of /p/, (40f) violates IDENT-

IO[G] twice. Most importantly, the [critical] CD of TT gestures for /s/ is adjusted to the

[closed] CD of TT gestures. In (40), I considered the unreleased [t‘] as a correspondence
of /s/. However, if the unreleased [t’] corresponds to /p/, (40f) would incur serious

violations of DENT-IO[G] because Lip gestures as a whole are adjusted to TT gestures.
Candidates (40b), (40c), (40d), and (40f) are tied in violations of Max-IOscg.
Because candidates (40b), (40c), (40d), and (400 all violate Max-IOscg once, (40b) and

(40f) are possibilities to be the most optimal candidate. (40b) and (40f) satisfy *CCD/

_{#, C} most optimally, tying in violation of *CCD/ _{#, C}. However, (40b?) is the

113

actual output, even though constraint Max-IOscg is violated by (40b?) The selection of

(40b) over (40f) indicates a preference of minimizing gestural adjustments from the
input. In the case of (400, the [critical] CD of TT gestures are adjusted to the [closed]
CD, resulting in violations of DENT-IO[G].

When /s/ alone appears in word-final position before a consonant as in

/nas/—)[nat’], the adjustment of the [critical] to the [closed] CD of TT gestures is

preferred over deletion of /s/, as shown in (25). The constraint Max-10368 is responsible

for gestural adjustment in [nat’] from /nas/, instead of deletion.

The fact that the actual output is (40b?) even though it violates Max-IOscg,
implies that there is a higher constraint than Max-10368. A higher constraint than Max-

10868 is *COMPLEX that bans a consonant cluster in coda position, as shown in (41).

(41) *COMPLEX (*COMP, hereafter): a consonant cluster is not allowed in coda position.
*COMPLEX >> Max-IOscg >>*CCD/_{# or C}>> [DENT-IO[G]
/n9_k_s_/ ‘soul’

/k/: GLO: [mid], SL: [closed], TB [closed]
/s/: GLO: [mid], SL: [critical], TTicritical]

 

 

 

 

 

 

 

 

 

 

 

 

 

mag] *COMP Max- *CCD/ {# or C} IDENT-
IOsea *[open] *[mid] *[critical] *[closed] IO[G]
a. nag *9 Tm." ".“‘“’(GL0>' *(SLlf*(SL7
fit). my * .. ,, , ... . .._ . Ill-*(SL) . *(GLO)
0. nos " “(GLO) 3 "(SLIP , .7 ’7 ° ‘4
d. nak * *!(GLO) , "f."   *(SL)
e. no— "”keS/ . I 8' .7 I 7 Ar ' . - I
f. not“ * ‘ *7 N V I f f f 0*(SL) *(GLO)
' *TT CD

 

 

 

 

 

114

(41) evaluates mafi/ against *COMPLEX and other constraints in (40). (41a) is far

from being the most optimal candidate because it violates the highest constraint,

*COMPLEX. Candidates (41b), (41c), (41d), (Me), and (41f) violate Max-10368. In

particular, (41c) violates Max-IOS¢g twice and is excluded from being the most optimal

candidate. One of two consonants in a consonant cluster does not have a correspondence
in candidates (41b), (41c), (41d), (41c), and (41f).

As mentioned in the analysis of CN, glottal gestures for obstruents are
impoverished or weakened to function as contrastive units in coda position. Therefore,
candidates with contrastive glottal CDs, such as (41c) and (41d), are excluded from being
the most optimal candidate. Compared to (41f), violation of [DENT-IO[G] is minimal in
(41b). Therefore, in an Obstruent-Obstruent cluster, an unreleased stop wins the
competition for the coda position, as shown in (41b).

In morphemes ending with a stop-fricative cluster, as in (40) and (41), the
constraint *CCD_/{#, C}, which is responsible for CN, plays a crucial role in preventing
the occurrence of fricatives in coda position.

Constraints and their hierarchy suggested in (41) correctly select the output form
(41b) in an Obstruent-Obstruent cluster. Therefore, it is worthwhile to investigate
morphemes ending with a sonorant-Obstruent cluster to see if constraints and their
hierarchy in (41) correctly select output forms in a sonorant-Obstruent cluster. Note that
sonorants involved in sonorant—Obstruent and sonorant-sonorant clusters are always either
/n/ or /l/, as shown in (37).

(42) evaluates the /l/-/t”/ cluster, 3 sonorant-Obstruent cluster. Because *COMPLEX

is undominated in the constraint hierarchy, hereafter, unless necessary, candidates

115

violating *COMPLEX are excluded from candidate evaluation. Hereafter, candidates
violating Max-IOseg twice are also excluded from candidate evaluation, because the most

optimal candidates never delete two consonants in a consonant cluster, as shown in (41c).

(42) /hal_tf+ta/: ‘to lick + declarative’

/l/: GLO[critical], SL[open], Oral [open], TB[open], TT[closed]
/th/ : GLO[openLSL[closed], Oral [closed], TT [closed]

 

/hafl+ta/ Max- *CCD/ {#, C} IDENT-
IOscg *[open] *[mid] *[critical] *[closed] IO[G]

 

 

 

Wa. halt’a *

 

1" hag” * *’ .%*:'*cSL)fi .
(GLO) 'vav. h.,_._._, __,__-‘ _

 

 

 

 

 

 

 

 

 

c. hat1 t’a " I. I I H V ”(SL). 1 [.71. *(GIO)

 

As Korean allows either a sonorant or an occlusion in coda position, as shown in

(3 8), candidates ending with [l] or [t’] are possible optimal candidates. All candidates in

(42) violate Max-IOscg. (42a) satisfies *CCD/__{#, C} vacuously because *CCD/_{#,C}

is a constraint for obstruents in coda position. Candidate (42b) is excluded from being

the most optimal candidate because the coda consonant has contrastive CDs at the glottis

and at the SL level. In (42c), if [t’] corresponds to/t"/, gestures of /t"/ are adjusted to [t’]

in order to optimally satisfy *C(TD/_{#, C} at the expense of violating faithfulness
constraint IDEN’l‘-IO[G]. Therefore, a candidate ending with /l/, (42a), is selected as the
most optimal candidate. Constraints in (41) correctly select the most optimal candidate in
a sonorant-Obstruent cluster, as shown in (42). Therefore, in a sonorant-Obstruent cluster,
a candidate with a sonorant in coda position is selected as the most optimal candidate.
However, the conclusion drawn in (42) poses problems for morphemes ending

with sonorant-bilabial stop and sonorant-velar stop clusters (see (3 7b)). (43) evaluates

116

output forms from a morpheme ending with a /l/-/,rfl/ cluster against constraints and their

hierarchy in (41).

(43) /i]ph+ ta/ 2 ‘to recite+ declarative’

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/ph/: GLO bpen], SL [closed], Oral [closed], Lipsiclosed]

 

 

 

 

 

 

 

 

 

 

/ilp_':+ ta/ Max- *CCD/ {#, C} IDENT-
1086% *[open] *[mid] *[critical] *[closed] IO[G]
b. apnea * *IGLO) :<GL9>
’5 °' *9“ * ”(SL) (9L)

 

 

 

 

Candidate (43b) is excluded from being the most optimal candidate because it
violates *[open]/_{#, C}, which is the highest constraint within *CCD/_{#, C}.
Candidate (43c) violates *[closed]/_{#, C}, which is the lowest constraint within
*CCD/_{#, C}. The faithfulness constraint [DENT-IO[G] is also violated in (43c).

As in (42a), candidate (43a) satisfies *CCD/_{#, C} most optimally without
violation of DENT-IO[G]. Therefore, according to constraints and their hierarchy in (41),
(43a>) is the most optimal candidate. However, (43a>) is not the actual output. The
actual output is (4363’), which violates *[closed]/_{#, C} and DENT-IO[G]. The actual
output (43c?) is not consistent with the conclusion drawn in (42): in a sonorant-

obstruent cluster, a candidate with a sonorant in coda position is selected as the most

optimal candidate. The actual output (430 <9”) has the unreleased stop [p’] in coda

position.
To investigate sources of inconsistency in selecting output forms between (42)

and (43), gestures involved in /ph/ and /t"/ are compared in (44).

117

 

(44) /ph/: VT (occlusion), GLO [open], SL [closed], Oral [closed], Lips [closed]
/t”/: VT (occlusion), GLO [open], SL [closed], Oral [closed], TT [closed]

The overall output characteristics at the VT for /ph/ and /t”/ are occlusion. CDs at
the glottis are [open] for /ph/ and /th/. CDs at the SL level are [closed] for /ph/ and /th/.
CDs of oral articulators for /ph/ and /th/ are [closed]. However, /ph/ and /th/ differ from
each other in the oral articulator used. The active articulator for /ph/ is the lips and the
active articulator for /th/ is the tongue tip.

Therefore, I assume that the reason that constraints and their hierarchy in (41)
select (42a) and (43c) as the most optimal candidates is closely related to the oral
articulator being used. In (42), the consonant cluster is composed of the lateral /l/ and /th/
whose articulator is the TT. In (43), a consonant cluster is composed of the lateral /l/ and

/ph/ whose articulator is the lips. When everything is equal, the preference is given to the

lip articulator, as shown in the comparison of [ip’ .t’a] from /ilp_’f+ ta/ in (43) and [halt’a]

from /hafl+ta/ in (42). Therefore, the lip articulator is preferred over the TT articulator
in SK.
The evaluation of /talk/ in (45) also illustrates an asymmetrical pattern of CCCS

according to the oral articulators involved.

118

(45) /talk/ 2 ‘chicken’

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/k/ 2 VT (occlusion), GLO [mid], SL [closed], Oral [closed], TB [closed]

 

/taLk/ Max- *CCD/ {#, C} IDENT-
IOseg *[open] *[mid] *[critical] *[closed] IO[G]

 

 

> a. ta] *

 

b. .tafik * *!(GLO) w “*(SL);

 

 

 

 

 

 

 

 

 

... c.ta_lg1 * *!(SL) *(GL93)i_-*'

 

(45a>) satisfies constraints and their hierarchy in (41) most optimally but it is not
the actual output. Although (43 c) violates *[closed]/_{#, C} and [DENT-IO[G], (43c?)
is the actual output. Therefore, in a /1/-/1</ cluster, the actual output has in coda position

the unreleased [k’], whose oral articulator is the TB. Thus, in a sonorant-Obstruent

cluster, candidates with TB and Lip articulators in coda position are selected as the
output. In addition, in a /l/-/m/ cluster, although both /1/ and /m/ have the [open] CD at
the SL level, [m], whose active articulator is the Lips, is preferred over the lateral [l], as
shown in [sam] from /salm/ (37c).

Therefore, in competition for the coda position between two consonants, a
consonant with the TB or Lip articulators appears in coda position, as shown in (43c) and
(45c). I suggest a hierarchy of constraints that are sensitive to the oral articulators used
for a consonant as in (46).

(46a) requires that the oral articulator of an input segment be maintained in its

output correspondent. According to the oral articulators, [DENT-10A,} is subdivided into
IDENT-TBM, IDENT-LipsAn, and IDENT-TTAn. Therefore, IDENT-TBM, IDENT-LipsAn,

and IDENT-TTAn require that any correspondent of an input segment specified with the

TB, Lips, or TT maintain the TB, Lips, or TT respectively. The constraint hierarchy

119

(46b) is responsible for the preference of a consonant with the TB or Lip articulator over
a consonant with the TT articulator. (46b) indicates that a consonant with the TB or Lip
articulator is preferred over a consonant with the TT articulator when two consonants are

in competition for the coda position.

(46) a. IDENT-IOAn (An - Articulator):

Let or be a segment in the input, and B be a correspondent of or in the output.
If or is [y articulator], then [3 is [’y articulator].

(DENT-TBA“, IDENT-LIPSAn, IDENT-TTArt)
(TB A“, LipsAn, and TTAn respectively hereafter)

b. TBAn/LIPSAn >> TTAn

Before establishing rankings between TB An/Lips An >> TTAn in relation to other
constraints in (41), there are two important points to be considered. First, an evaluation
standard for the lateral [l] with respect to TBA“ and TTAn has to be set up because the

lateral /l/ is specified with TB [open] and TT [closed] in the input, as shown in (43) and

(45). Consequently, candidates with the lateral [l] in coda position either violate or

satisfy TBA“ and TTAn simultaneously, depending on how constraints TBA“ and TTArt

are interpreted in a grammar. If constraints TBA“, LipsAn, and TTAn are interpreted as
constraints that only allow consonants with TB, Lip, or TT articulators, candidates with

the lateral [l] in coda position violate TBA“ and T'I‘Art simultaneously because both the
TB and TT are active articulators for the lateral [1]. If constraints TBA“, LipsAn and

TTAn are interpreted as constraints that allow any consonant that involves TB, Lip and

120

TT articulators, candidates with the lateral [l] in coda position always satisfy both TB Art

and TIA“ simultaneously.

Based on the assumption that gestures are adjusted according to context, I suggest
that the distribution of [l] in a given context determines satisfaction or violation of TB An
and TTAn. Sproat and Fujimura (1993: 306) observe that ‘In syllable-final /l/s, the

(vocalic) dorsal gesture is attracted to the nucleus of the syllable, whereas the
(consonantal) apical gesture is attracted to the (right) margin of the syllable...In syllable-
initial cases, the reverse prediction is made’. In addition, Browman and Goldstein
(1992:166-67) state that for the lateral /1/, the peak of TB gestures occurs significantly
earlier than the peak for TT gestures in word-final position.

Based on Sproat and Fujimura (1993) and Browman and Goldstein (1992), I

assume that when the lateral [1] appears in word-final position or before a consonant,

constraint TTAfl is satisfied because the TT gestures of [l] are active in coda position.
When the lateral [1] appears in word-initial position or before a vowel, constraints TBA”

and TTAn are satisfied simultaneously because both TB and TT gestures occur
synchronously and are active in word-initial position or before a vowel.

The second important point to be noted is that constraints TBA“, LipsAn and TTAn
evaluate for sonorants and obstruents. Therefore, satisfaction of TTAn in a given context
means that consonants such as [n, l, t s, c] are allowed in the context. In Articulatory
Phonology, consonants such as [n, l, t s, 0] form a natural class by sharing the TT

articulator. Therefore, CDs within the TT articulator such as [close] or [mid], and CLs of

121

the TT such as [alveolar] or [palatal] are not considered in evaluation of TTArt. What

matters most is if the TT articulator in an input segment is maintained in its output
correspondent. As mentioned in 1.2.4.2, one way of grouping sounds into natural classes
is based solely on the moving articulators. Therefore, the TT articulator itself plays a role
in a grammar, as suggested in (46).

Having introduced an evaluation standard for the lateral /l/, I evaluate /talk/ below

in order to establish ranking between TBArt/ LipsAn >> TTArt with respect to other

constraints in (41). Recall (45) first. Without constraints TBA“, LipsArt, TTAn and their

hierarchy suggested in (46b), a candidate ending with the lateral [l] was considered as the
most optimal candidate because it satisfied *CCD/_{#, C} most optimally, as shown in
(45a>). However, (45a>) was not the actual output in SK. The actual output was
(45c? ), which violated both *CCD/_{#, C} and IDENT-IO[G]. This indicates that
violation of these constraints is tolerated in SK. When two consonants compete for the
coda position, maintaining the TB articulator in the output takes precedence over

satisfaction of *CCD/_{#, C} and IDENT-IO[G] in SK, as shown in (37b). Therefore,

TBAn/LipsAn>> TTM is ranked higher than *CCD/_{#, C}, as shown in (47).

122

 

(47) *Max-IOseg >> TB An/Lips Ad >> TTAn >>*CCD/_{#, C} >> IDENT—IO[G]
haw: ‘chicken’

/1/ 2 GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]

 

 

 

 

 

 

 

 

 

 

haik/ Max- TB A n/ TTAn *CCD/ {#, C} IDENT-
IOScg Lips An *[open] *[mid] *[critical] *[closed] IO[G]
b. ta]; * * *ttGLO) , , , «51.) , .
<3” c . tak1 * * *(SL) *(GLO) ,

 

 

 

 

 

Candidate (47a) satisfies *CCD/_{#, C} most optimally but violates TBA“ and
LipsAn, which are higher than *CCD/__{#, C}. The lateral [l] in coda position as in (47a)

violates TBA“ because the TB gestures of the lateral /l/ is not maintained or is weakened
significantly in coda position (Sproat and Fujimura 1993 and Browman and Goldstein
1992). However, satisfaction of TBAn/Lipsmt and TTAn take precedence over
satisfaction of *CCD/_{#, C} and [DENT-IO[G] in competition between two consonants
for the coda position. The TT articulator is significantly active when the latral [1] appears
in coda position.

Candidate (47b) is not an optimal output because it has the contrastive [mid] CD
at the glottis and the [closed] CD at the SL level in coda position. Although candidate

(47c) violates *[closed]/_{#, C} and [DENT-IO[G], it is the most optimal candidate
because it satisfies the relatively higher constraint TB An.
In order to see if TTAn should be ranked lower than *CCD/_{#, C}, (48) evaluates

candidates from /hal_t1+ta/. In a /l/—/th/ cluster of /hal_fl+ta/, the TT articulator is involved

in both /1/ and /th/. Therefore, for the correct output to be selected from candidates of

123

/halfi+ta/, it would be unnecessary to rank TTAn below *CCD/__{#, C} until it is proven

to be otherwise.

(48) /hafl+ta/ : ‘to lick + declarative’

/1/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]

 

 

 

 

 

 

 

/th/: GLO closed] SL [closed], Oral [closed], TT [closed]
/hal_t:’_+ta/ Max- TBAn/ TT,, *CCD/Jae, C} IDENT-
IO$6 LipSArt *[open] *[mid] *[critical] *[closed] IO[G]
@‘a halt’a * *
b. hafit’a * * *!(GL0) g ’ _- i from] » w
c hat1 .t’a * * a”(SL) efl."'(GLO)i'e‘-l.
d. hag.“ *!(GLO) v *(sL) ' . *(GLO)

 

 

 

 

 

 

 

 

 

 

Candidates with [l] or coronal stops in coda position violate TBA“ and satisfy

TTAn equally, as shown in (48). Therefore, candidates with /l/ or stops with the TT

articulator in coda position are evaluated against *CCD/_{#, C}. However, because /1/ is
not an obstruent, effects of *CCD/_{#, C} do not occur in candidates with the lateral [l]

in coda position, as shown in (48a). Candidates (48b) and (48d) with [th] and [t’] in coda

position violate *CCD/_{#, C} twice. There are contrastive CDs at the glottis and
[closed] CDs at the SL level. Although candidate (48c) violates *[closed]/_{#, C} at the
SL level and the faithfulness constraint, DENT-IO[G], it is a less optimal candidate than
(48a). Candidate (48a) does not violate both *[closed]/_{#, C} at the SL level and the
faithfiilness constraint, DENT-IO[G]. Therefore, candidate (48a) with [l] in coda position
is selected as the most optimal candidate because it satisfies output constraints most
optimally. Therefore, constraints and their hierarchy in (47) select the correct output

form as the most optimal candidate when /1/ and /th/ compete for the coda position.

124

Based on (48), it is predicted that when a sonorant and an Obstruent compete for
the coda position and the TT articulator is used for both the sonorant and the obstruent,
output forms preserve the sonorant in coda position. This is because both sonorant and
Obstruent satisfy TTAn, and a candidate with the sonorant in coda position satisfies
*CCD/_{#, C} most optimally, as shown in (48a). (37a) and (37d) illustrate that the
prediction is correct. For example, in (3 7a) and (37d), /n, l/ compete with /c, 8, th/ for the
coda position and /n, l, c, 3, th/ are specified with the TT articulator. Thus, output forms
have either [11] or [1] in coda position because they satisfy constraints (47) most
optimally.

[ha] .t’ a] from /hal_t’_’_+ta/ in (37d) appears to be an exception, compared to [tak’]
from /tal_k/ in (37b). [haLt’a] from /hal_t”+ta/ is not an exception, but can be analized as a
result of satisfying constraints in (47) most optimally, as shown in (48). As the ranking

of TTAn above *CCD/_{#, C} correctly selects the output when two TT consonants form

a consonant cluster as shown in (48), TTAn is ranked below TBAn/LipsAn and above

*CCD/_{#, C} until proven otherwise.

I conclude that constraints and their hierarchy in (49) are responsible for CCCS in

SK:

(49) *COMPLEX>>*Max-IOseg>>TBAn/LipsAn>>TTAn>> *CCD/_{#, C}>> IDENT-IO[G]

Before discussing KSK, (50) examines constraints and their hierarchy in (49). A
/n/-/h/ cluster is evaluated in (50) to see if (49) selects correct output forms given other

input consonant clusters.

125

(50) lanh+tal ‘don’t do’

/n/: GLO [critical], SL [open], Oral [open], TT [closed]
/h/ 2 GLO bpen], SL [open], Oral [open[

 

 

 

 

 

 

 

 

 

 

lanh+tal *COMP Max- TBAn/ TTArt :‘CCD/ {#, C} IDENTo
IOIeg . *[open] *[mid] *[crrtrcal] *[closed] IO[G]
LrpsAn
a. aim-Va H! _ *(GLO) , _ _ . ,. .. . . . _ ._ , .
' *(SL).
@‘b antha *
c. ah.t’a * *!(GLO)
d.at’.t’a * “(SL) 1*(GL0)1
. _ ”11*(SL') '
e. at.t’a * *! 133;? . ‘ _<‘;*(SL)1;:. ~ #(GLO).
(GLO) *(SL) ‘
f. ah.ta * *! *(GLO) ' .
" ‘ ~,*<SL2. ~
g. a.ta **l ' . l

 

 

 

 

 

 

 

 

 

 

In (50), because *COMPLEX is an undominated constraint, candidates that violate

*COMPLEX, as in (50a), are not the most optimal. Candidate (50g) is not the most optimal

candidate because it deletes two consonants in the input, violating Max-10m,g twice.
(50b), (50c), (50d), (50e), and (50f) are tied in violation of Max-10368. Candidate (50f),

with [h] in coda position, satisfies TBA”. There is no TB articulator in the input, which is
preferred in consonant competition for the coda position in a consonant cluster.
However, (50f) is excluded from being the most optimal candidate. In (50f), a consonant
with the TT articulator in the input, namely /n/, is not maintained in competition between
two consonants for the coda position. Instead, a consonant with the glottal articulator in
the input, /h/, is preferred over a consonant with the TT articulator, /n/. (50b), (50c),
(50d), and (50e) satisfy TTAn because the TT articulator in the input is maintained.
However, (50e) is not the most optimal candidate because of violation of *[mid]/_{#, C}

and *[closed]/_{#, C}. In addition, if [t] corresponds to /h/, [open] CDs at the glottis and

126

 

at the SL level in the input are not maintained in [t]. Hence, (50e) violates the
faithfulness constraint DENT-IO[G] at the glottis and at the SL level. (50d) violates
*[closed]/_{#, C} and the faithfulness constraint DENT-IO[G]. [open] CDs at the glottis
and at the SL level in the input are not maintained in (50d). Candidates (50b) and (50c)
satisfy *CCD /_{#, C} most optimally because both (50b) and (50c) end with the
sonorant [n]. However, constraints given in (49) select (50b) as the most optimal
candidate. This is because the glottal gestures in the input are maintained in (50b) by
coalescing /h/ and the following /t/ into [th]. Therefore, constraints and their hierarchy in
(49) select the output form correctly in other combinations of coda consonant clusters.
Thus, in SK, constraints sensitive to oral articulators are ranked higher than constraints

sensitive to CDs in the VT.

2.2.2.2 The gesture-based account of CCCS in Kyung-Sang Korean

Patterns of CCCS in KSK are not much different from those in SK except for the

sonorant-stop clusters of (3 7b), as shown in (51):

(51) Patterns of CCCS in KSK and SK in sonorant-stop clusters

KSK SK
Input Output Output
1P /na]p+ta/ [n91.t’ a] [nap1 .t’ a] ‘be wide + declarative’
lk ligfia/ [iLt’a] [$1621] ‘to read + declarative’
1th /hal_fl+ta/ [hal_.t’ a] [ha] .t’ a] ‘lick + declarative’

127

When competition for the coda position occurs in a sonorant-obstruent cluster in
SK, a candidate with an unreleased stop in coda position is selected as the most optimal
candidate if the articulators for obstruents are either the TB or the Lips. If the 'IT is the
articulator for an obstruent in a sonorant-obstruent cluster, the sonorant appears in coda
position, as shown in (48). However, sonorants appear in coda position as a result of
satisfying constraints in (49) most optimally.

However, in KSK, when competition for the coda position occurs in a sonorant-
obstruent cluster, a candidate ending with a sonorant is selected as the most optimal
candidate regardless of articulators of obstruents. Recall that the sonorant in a sonorant-

obstruent cluster is either /n/ or /l/, whose articulators are the TT in coda position.

When there are no constraints sensitive to oral articulators as in TBA“, LipsAn,

and TTAn, the most optimal candidates selected in sonorant-obstruent clusters have
sonorants in coda position, as shown in (43) and (45), and as in KSK in (51). Therefore,
one may speculate that there are no constraints such as TBA“, LipsArt, TTAn in KSK.
However, this speculation is not valid in OT because in OT, constraints are universal and
language differences stem from different rankings of universal constraints. If language
differences stem from different rankings of universal constraints, I assume that Korean
also shares the same constraints regardless of dialect, and dialect differences stem from
different rankings of the same constraints.

I argue that differences in CCCS patterns between SK and KSK stem from

different rankings of the same constraints. In support of constraint rankings in KSK,

recall (45). Before constraints sensitive to the articulators of consonants such as TBA”,

LipsAn, and TTM were considered, the most optimal candidate selected was a candidate

128

ending with a sonorant, regardless of the articulators of obstruents, as shown in (52a)

below.

(52) Max-IO$ch >> *CCD/_{#, C} >> DENT-IO[G]
lika/ ‘to read’

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/k/ 2 VT (occlusion), GLO [mid], SL [closed], Oral [closed], TB [closed]

 

 

 

 

 

 

 

 

 

/il_lgta/ Max- *CCD/ {#, C} IDENT-

1056 *[open] *[mid] *[critical] *[closed] IO[G]
era-item“) 7 _ H .. ,. ..
b. iy’a * *lLGLO) V;,71}:'{.f.,§*»7:i3.f{5;1'Ie'fffilgfiLi:f.ff.“‘(S_.L)gi'fiizlfé if}.iifff}{1735‘flii'f».I.f
c. ik’.t’a(SK) * “(SL) *(Gw)

 

 

 

 

 

When *CCD/_{#, C} is not dominated by TBA“, LipsAn and TTAn as shown in

(52), constraints in (52) select a candidate with the sonorant [l] in syllable-final position,

(52a?) as the most optimal candidate. In SK, candidate (52a) is not the actual output.
(52c) is the actual output in SK. Constraints such as TBA“, LipsAn and TTAn, and their

ranking above *CCD/_{#, C} are responsible for the selection of (52c) over (52a) in SK,

as shown in (47).
Contrary to SK, (52a) is the actual output in KSK. Thus, effects of TBA“, LipsAn
and TTAn are minimal. Therefore, I suggest that TBAn/LipsAn >> TTArt is ranked lower

than *CCD/_{#, C} in KSK, as shown in (53). Based on the fact that when two
consonants are in competition for the coda position, output forms always have a

consonant with TB, Lip, or TT articulators in coda position regardless of gestural
changes, I rank TB An/LipsAn >> TTArt above IDENT-IO[G] in KSK, as shown in (53).

A morpheme ending with a /1/-/k/ cluster is evaluated in (53).

129

(53) *COMPLEX>> Max-IOscg>>*CCD/_{#, C }>>TBAn/LipsAn >>T'I'An >>IDENT-IO[G]
/ifl(_ta/ ‘to read’

/l/: GLO[critical], SL[open], Oral [open], TB[open], TT[closed]

 

 

 

 

 

 

/k/ 2 VT (occlusion), GLO[mid], SL[closed], Oral closed], TB[closed]
/ifl<_ta/ ax- *CCD/J#,C} TB Art/ TTArt IDENT-
IOSCS *[open] *[mid] *[critical] *[closed] LiRSAn IO[G]
a. il.t’a * 1 *
b. .i_.t’a * ”(GLO) 7f?![1571.752-.;7'5I*%.7f5375.“:39'1”",(3111iflif .;7f'ff2;.‘l‘lg‘i'F-j'e
c . ik’t’a "‘ *!(SL) *ICGLQl-Ef

 

 

 

 

 

 

 

 

 

 

In (53), TB AmlLipsAn and TT,“, which are sensitive to oral articulators, are
ranked lower than *CCD/_{#, C} and higher than DENT-IO[G]. (53a), (53b) and (53c)
are tied in Max-IOscg because of segment deletion in order to fit into Korean syllable
structure. Violation of Max-IOseg is tolerated to a certain degree because Korean syllable

structure allows at least one coda consonant. Thus candidates with [l], [k], or [k’] in

coda position are evaluated against *CCD/_{#, C} in (53). (53b) violates a relatively
higher constraint within *CCD/__{#, C} because of the [mid] CD at the glottis. (53c)
violates the lowest constraint within *CCD/_{#, C} but (53c) is less optimal than (53a).
Candidates with sonorants in coda position always satisfy *CCD/_{#, C} most optimally

because /n/ or /1/ is not an obstruent.

Note that *CCD/_{#, C} affects obstruents only and TB An, LipsAn and TTAn
affect sonorants and obstruents together. Therefore, when *CCD/_{#, C} is ranked

higher than TBAn/LipsAn >> TTArt, as in (53), all candidates with obstruents in coda

130

 

position, as in (53b) and (53c), are excluded from being the most optimal in a sonorant-

obstruent cluster.

When TB An/LipsAn is evaluated, candidates with sonorants satisfy *CCD/_{#, C}
most optimally, as shown in (53a). Therefore, in KSK, when sonorants and obstruents
compete for the coda position, output forms always have either [I] or [n] in coda position
because [1] and [n] satisfy *CCD/_{#, C}. Candidates (53b) and (53c) satisfy TBA” but
satisfaction of *CCD/_{#, C} takes precedence over satisfaction of TBA", and LipsAn in
KSK.

However, when /l/ and /m/ compete for the coda position, candidates with [l] or

[m] in coda position satisfy *CCD/_{#, C} most optimally, regardless of the ranking of

*CCD/_{#, C} in the constraint hierarchy. Therefore, in KSK and SK, constraints TBA",

LipsAn and TTAn, and the ranking of TBAn/LipsAn >> TTArt, play crucial roles in

selecting the output, as shown in (54).

(54) /sa]m/ ‘life’

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/m/: GLO [critical], SL [open], Oral [open], Lips [closed]

 

 

 

 

/sa1m/ Max- *CCD/ {#, C} TBAr-t/ TTArt IDENT-
IO$68 *[open] *[mid] *[critical] *[closed] Lips A rt IO[G]
a? a_ sari; * at: 1:?
(KSK, SK)
b. sal * *!

 

 

 

 

 

 

 

 

 

 

When /1/ and /m/ compete for the coda position, KSK and SK have output with

[m]. Candidates with [l] or [m] in coda position tie in satisfaction of the constraint

131

*CCD/_{#, C}. Therefore, as shown above, in a sonorant-sonorant cluster the most
optimal candidate is selected by the constraints TBArf/LIPSAn in both KSK and SK.

A previous study (Lee, 1999) claimed that the preference of nasals over the lateral
/l/ is responsible for [sam] from /salm/ in a sonorant-sonorant cluster. However, there are
no other sonorant-sonorant clusters other than the /l/-/m/ cluster in Korean. Therefore,

there is no evidence for the preference of nasals over the lateral /l/ in a sonorant-sonorant

cluster. However, there is abundant evidence for the preference of TBAn/LipsAn over

TTAn in both KSK and SK.

(55) examines a morpheme ending with a /p/-/s/ cluster, an obstruent-obstruent
cluster, to see whether the constraint hierarchy suggested in (53) for KSK selects correct

output forms.

(55) /kaps/: ‘price’

/p/: GLO [mid], SL [closed], Oral [closed], Lips [closed]
/s/ 2 GLO mid], SL [critical], Oral [critical], TT [critical]

 

 

 

 

 

 

 

 

/kaps/ Max- *CCD/ {#, C} TB Ami TTArt IDENT-
10$,g *[open] *[mid] *[critical] *[closed] Lips A; n IO[G]
a. kap * a”(GLO) *(SL)?*
6" b. .kap.‘ * ‘99 ‘iy‘<G¥40?§£
c. kas’ * *(GLO) “(SL) 19s}:Tg‘ifii-ri‘T:fiEig‘iji'i}?
d. kat’ * I *(SL) " '*!A " 55*TTCD-i:
. ,;:.#(GL~Q)_1.

 

 

 

 

 

 

 

 

 

(55a) and (55c) are not the most optimal candidates because they violate relatively
higher ranking constraints within *CCD/_{#. C}. (55b) and (55d) also violate

[closed]/_{#, C}. If [t’] in (55d) corresponds to /s/ in the input, it violates [DENT-IO[G]

because of the [closed] CD of TT gestures from the [critical] CD of TT gestures for /s/.

132

However, (55b) is preferred over (55d) because (55b) has the lip articulator in the coda
position, instead of the TT articulator (55d). Thus the constraints and their hierarchy
suggested in (53) for KSK correctly select the output given different consonant clusters.

Therefore, the constraint ranking *CCD/_{#, C} >> TBAn/LipsAn >> TTAn is responsible

for patterns of CCCS in KSK.

2.2.3 Influences of non-surfaced consonants

In this section, I discuss two points with respect to consonants that failed in
competition for the coda position in a consonant cluster. First, when CM in VlCn-2Cn-
1C,,V2 does not surface in the output due to Korean syllable structure or phonetic
implementational conditions in a given context, gestures of non-surfaced CM influence

phonetic realizations of C“, as shown in (56).

(56) Phonetic realizations of Cn from VlCn-2Cn-1CnV2

a. Tensification /ilk_+ ta/ [ilt’a] ‘to read’(KSK)
/ilp_h+ ta/ [ilfla] ‘to recite’ (KSK)
/nolp_ + ta/ [nolfi a] ‘be wide’ (KSK)
b. Aspiration /olh + W [olfla] ‘ to be right’
lanh + ta/ [anfio] ‘don’t do’
lanh + ta/ [anfla] ‘not do + declarative’
c. /manh + 80/ [mango] ’be plenty + statement’
/ol_h_ + 50/ [olL’o] ‘be right + statement’

133

The comparison of (56a) to (56b) illustrates that phonetic realizations of CH are
influenced by gestures belonging to non-surfaced CM. In (56a), when gestures of non-
surfaced consonants are specified with occlusion (/k, ph, pl), Cn is tensed in the output, as
in [ilfi a] from /ilk + ta/ ‘to read’. In (56b), when gestures of non-surfaced consonants are
specified with the glottal fricative /11/, CH is aspirated in the output, as in [olfia] from loll;
+ ta/ ‘to be right’.

The gestural interactions between non-surfaced CM and Cu occur while

maintaining Korean syllable structure, as shown in (57).

(57) A /ol\
X X X
(C) V Cn-2 Cn-l Cn V (C)
Gl Gl G1
G2 G2 G2
G3 G3 G3

I suggest that Korean CCCS be viewed as the overlapping of timing as assumed in
Browman and Goldstein (1989 and 1990a), Byrd (1996) and Cho (1998a and 1998b),
instead of as deletion the entire timing and gestures associated with CH. The
overlapping of timing does not incur doubly linked timing slots in onset and coda
positions, thus, satisfying *COMPLEX (*0'[XX__] or *O'LXXD. Although CM in the
output does not occupy a timing slot, some gestures associated with CH are reassociated
to C“, resulting intensification of Cn, as in [halt’ a] from /halflta/ or aspiration of C“, as in

[anfla] from /anh_+ta/. Which gestures of Cm] can be reassociated to Cn follows phasing

134

principles of gestures that will be discussed in detail in chapter 4 (see the last page of
section 1.2.4.2 for a brief summary of phasing principles).

Data in (56a) presents another argument for the influence of gestures associated
with non-surfaced CM. For example, in Korean verbal morphology, lax stops are tensed
after obstruents and nasals, as in [ipt’_ a] from /ip + W ‘to wear’ and [anti a] from /an + ta/
‘to hug’. Lax stops are not tensed after the lateral /l/, as in [kalda] from /kal +ta/ ‘to
cultivate’. Instead, lax stops are voiced between two [+voice] features, as in [kalda] from
/kal + ta/ ‘to cultivate’. However, the lax stop /t/ is tensed in the environment where the
lax stop /t/ should be voiced in (56a). Ifwe assume influences of non-surfaced consonant
gestures on /t/ in (56a), the tense [t’] after [1] in (56a) is not an exception to obstruent
tensification in Korean Verbal Morphology. /t/ is tensed in the output because of non-
surfaced consonant gestures related to /k, ph/.

Second, gestures of Cn themselves play a role in phonetic realizations of Cn in
VlCn.2Cn-1CnV2. In other words, gestures of Cn and gestures of non-surfaced consonants,
namely CM, interact together in forming phonetic realizations of C“.

For example, when the glottal fricative /h/ is followed by lax stops, lax stops are
aspirated, as in [ol.kho] from /olh+ko/. However, in (56c), when the glottal fricative /h/ is
followed by /s/, /s/ is tensed, as in [ol.s’o] from /olhso/, instead of aspirated, as in (56b).
The reason that /s/ is not aspirated after /h/ as in (56b) stems from gestural characteristics
of /s/, which forms a natural class with /h/ as frication at the VT. As mentioned in the
analysis of fricatives in coda position, having the [critical] CD of TT gestures is one of
the important gestures for /s/. Considering that targeted gestures of onset consonants are

more likely to be maintained than targeted gestures of coda consonants (Byrd 1996 and

135

Cho 1998b), gestures of /s/ are likely to be maintained in (56c). However, the large
amount of airflow escaping from the g1 ottis for /h/ increases pharyngeal pressure behind
the [critical] CD of TT gestures and increases turbulent frication, resulting in tense /s’/
after /h/. Therefore, gestures of Cn and gestures of non-surfaced CM interact together in
phonetic realizations of Cu.

According to gestural characteristics of Cm], phonetic realizations of Cn are

divided into two patterns: tensification and aspiration of Cu:

(58) Tensification Aspiration
a. fhalt_ht_a/ —)[ha[.t’a] b. lama —-> [ant’a] c. ama/ —> [anfia]

Cn—l Cn Cn-l Cn Cn-l Cn

a“, ph, p/ /t/ /c, s/ m /h/ _ f /t/
SL [closed] [closed] [critical] [closed] [open] [closed]
Oral [closed] [closed] [critical] [closed] [open] [closed]

Tensification of Cn occurs when C..-) does not have the [open] CD at both the SL
and Oral levels, as shown in (58a) and (58b). The aspiration of Cn occurs when CM has
the [open] CD at the SL and Oral levels, as shown in (58c). Therefore, I conclude that
there are influences from gestural characteristics of non-surfaced CM to C“. In addition,
phonetic realizations of Cu depend on gestural characteristics of C“, as shown in the

comparison of (57a) and (57c), and gestural characteristics of C“, as shown in (58).

136

2.2.4 Summary

In the analysis of this study, I have first shown that there is a close relationship
between CN and CCCS. Primary constraints that are responsible for CN such as
*CCD/__{#, C} and IDENT-IO[G], and their basic rankings (*[open]>>*[mid]
>>*[critical] >>*[closed]>>...>> [DENT-IO[G]) are still maintained in CCCS of SK and
KSK. In addition, *CCD/_{#, C} and DENT-IO[G] also play a role in CCCS of SK and
KSK, especially when two obstruents compete for the coda position, as shown in (41) and
(5 5).

Second, the ranking of TBAn/Lipsm >> TTArt with respect to *CCD/__{#. C} is
responsible for different patterns in CCCS between SK and KSK, as shown in (47) and
(53). Therefore, there is no need for separate rules or constraints for SK and KSK. In
SK, constraints sensitive to articulators such as TB An, Lips“, TTArt are ranked higher than
the constraint *CCD/_{#. C}, which is sensitive to manner of articulation, as shown in
(47). In KSK, constraints sensitive to articulators are ranked lower than the constraint
*CCD/_{#. C}, as shown in (53).

Third, I have demonstrated influences of non-surfaced consonants on the
following consonants and predicted phonetic realizations of the following consonants
based on gestural characteristics of non-surfaced consonants, as shown in (58).

Fourth, in (48), output forms that appear to be exceptions in CCCS are found not

to be exceptional. For example, in SK, [hal.t’a] from /hal_tf+ta/ (37d) appears to be an

exception to [ik1 .t’a] from /il_lg_+ta/ and [nap’ .t’a] from /noLp+ta/ (37b). In the former,

the lateral /1/ surfaces in the output. In the latter examples, the lateral /1/ does not surface

137

in the output. Traditionally, (3 7d) is considered as an exception but (48) illustrates that
(3 7d) results from satisfying output constraints most optimally. Therefore, in this study,
CC CS in Korean is explained in a systematic and unified way, regardless of dialect.

Therefore, I conclude that CN and CCCS are related because both phonological
phenomena are motivated to fit into Korean syllable structure and phonetic
implementational conditions in a given context.

Previous analyses (Ahn 1985, Cho 1990, and Kim-Renaud 1974) fail to connect
CN and CCCS and to explain dialect differences between SK and KSK in a systematic
and unified way. Separate rules or principles are suggested for CN and CCCS. For

example, Ahn (1985) proposes CN and CCC S as follows:

(59) A. CN (Ahn 19852171) B. CCCS
Ahn (1985: 168)
A a. [-continuant, -coronal] > [+lateral]
f b. C1>C2 (a) is applied before (b)
+obstruen —-> -tense
[ or coronal [aspiratefl
0t coronal

CN in (59A) is proposed to prevent tense and aspirated obstruents from occurring
in coda position but obstruents maintain their place of articulation by maintaining [+/-
coronal]. However, (59A) cannot explain non-occurrence of the affricate [c] in coda
position because /c/ is specified with [+coronal, -anterior]. According to (59A), the
affricate [c] is allowed in coda position.

In (598), preference laws are proposed for CCCS. Non-continuant and non-

coronal segments such as [p, k, m] are preferred over the lateral /1/ because of (59Ba),

138

when two consonants compete for the coda position. Otherwise, the first of the two
consonants appears in coda position according to (59Bb). However, there is no
connection between CN and CCCS in Ahn’s (1985) proposals although both (59A) and

(59B) are motivated to fit into Korean syllable structure.

139

Chapter 3 Post-Obstruent Tensification

3. 1. Introduction

Before discussing Post-Obstruent Tensification (POT), the term ‘Obstruent
Tensification’ (ObsT) is first introduced for subsequent discussions on the phonetic
realization of 02 in the sequences VlOlOZVz (O: obstruents), V1N02V2 (N 2 nasals), and
V1L02V2 (L: the lateral /l/). In this study, ObsT refers to a phonological phenomena in
Korean in which a syllable initial lax obstruent /p, t, k, s, c/ is tensed after an obstruent, a
nasal, or the lateral /l/, resulting in [p’, t’, k’, s’, c’] respectively. ObsT is subdivided into
POT, Post-Nasal Tensification (PNT), and Post-Lateral Tensification (PLT) according to
the consonant before the lax obstruent. Instead of ‘Obstruent Tensification’, often the
term ‘Obstruent Glottalization’ is used in the Korean linguistics literature. In this study,
the term ObsT is used because the term ‘Obstruent Glottalization’ is likely to be
associated with glottal gestures only. However, ObsT is characterized by minimal glottal
opening, long oral-closure duration, heightened subglottal pressure, stiffness, and
muscular tension of the vocal tract walls and the pharynx (Kim-Renaud 1974, Lee 1994
and Rhee 1997).

This chapter mainly investigates cases in which a lax obstruent /p, t, k, s, c/ is

tensed to [p’, t’, k’, s’, c’] respectively after an obstruent, as shown in (1).

140

(l) a. Algal/noun ‘funny act or sayings’ -—> [il_<’.§’al]

b. /p"ut/prer1x + /sakwa/noun —> [phut’.§’a.kwa]
‘unripe’ ‘apple’ ‘unripe apple’
/ COLI/ verb stern + /§0/ declarative marker -) [001’.§’0]
‘be good’ ‘It is good’

C. / Sa-p/ noun + / Qil/ bound noun morpheme _) [33-91-911]
‘shovel’ ‘repeated action of something’ ‘shoveling’

d- / kuk/ noun “Ht 0r (”sub-compound morpheme + /.paP/noun—) [101131 42,3131]
‘soup’ ‘rice’ ‘a kind of food’
POT refers to tensification of lax obstruents after obstruents, as in (l). POT
appears within non-derived words, as in (1a), within words with bound morphemes, as in

(lb) and (1c), and across two nouns that form sub-compound nouns, as in (1d).

Whenever two nouns form sub-compound nouns as in (2a), [d] or [t’] appears

between two nouns in the output.

(2) a. /u/ ‘upper’ +/elin/ ‘elders’
/c"o/ ‘candle’ +/pul/ ‘fire’
/ko/ ‘nose’ +/tir)/ ‘bridege’

‘ ’ 9

b. /pom/ ‘spring’ +/pi/ rarn

[uderin] ‘seniors or elders’
[chot’p’ul] ‘candle light’
[kot’ t’irj] ‘nose bridge’
[pomp’i] ‘spring rain’

llll

In (2a) and (2b), POT occurs obligatorily even though the first nouns end with a
vowel or a nasal. Therefore, it has been assumed that there is a lexical sub-compound

morpheme, known as [t] or ['2], between two nouns that forms sub-compound nouns

(Sohn 1987 and Kim-Renaud 1974). Whether the sub-compound morpheme is [t] or [Y]

141

is not relevant to this study. What is important in this study is that the characteristic of
the sub-compound morpheme is occlusion, which is specified with either the [closed] CD
at the glottis or the [closed] CD at the SL level.

The Korean linguistic literature (Jun 1995 and Lee 1996) suggests that ObsT

occurs within an accentual phrase but not across accentual phrases, as shown in (3 a).

(3) a. /c”oka/noun + /ci,p/noun + lkochi/verb stem +/ki/
‘straw’ ‘house’ ‘to fix’ nominal suffix
-> [ChogaCip’IAeeenmar Phrase [kochigilAeeemuar Phrase ‘to fix a straw thatched root”

b. /ci,p/noun ‘house’+ lcip/noun ‘house’ —> [cip’. cip’] ‘every house’
/kak"/noun ‘each’+ /l_(ak’/noun ‘each’ —> [kal_<’. l_<ak"] ‘each person’
laph lnoun ‘front’ ltwie/noun ‘back’ —-> [ap’. twi] ‘front and back’

In (3a), the lax stop [k] in the third word is not tensed after the unreleased stop
[p’], as shown in [chogacip’][l_(ochigi]. This is because [k] in [l_<oc"igi] is not located
within the same accentual phrase with [p’] in [chogacip’].

In addition, ObsT does not occur when two nouns form co-compound nouns, as

shown in (3b). The [c, k, t] in the second word appears without being tensed after the

unreleased stop [p’, k", p’] respectively. Lax obstruents [c, k, t] are in a condition where

POT occurs but they are not tensed in (3b) after unreleased stops [p’, k’, p’]. Thus,

POT is under-applied in (3b).

142

Kang (1992) and Sohn (1987) attribute the under-application of POT in (3b) to
the different morpho-syntactic structures of sub-compound and co-compound nouns, as

shown in (4).

(4) a. sub-compound nouns (1d) b. co-compound nouns (3b)

NP b. NP
/l\
N//l\ N N g N
modifier head head i head
/c"o/ /t/ /pul/ Icip/ (arid) /cip/
‘candle’ ‘fire’ ‘house’ ‘house’

(4a) and (4b) represent morpho-syntactic structures of the sub-compound noun (1d) and
the co-compound noun (3b) respectively. When the sub-compound morpheme /t/

appears before a noun beginning with a lax obstruent as in (4a), the lax obstruent in the

second noun is tensed due to the sub-compound morpheme /t/, as shown in [chot’.p’ul]

‘candle light’ from /c"o/ ‘candle’ + /t/3ub-compound morpheme +/pul/ ‘fire’, even though the first
noun does not have /t/. The morpho-syntactic structure of co-compound noun (4b) has
an invisible ‘and’ branch, as represented by a broken line, which does not cause
tensification.

Data (5) shows cases in which POT does not occur.

143

(5) a. /nak_t"a/noun —> [nak1 .fla] ‘camel’
hes/prefix + (phan/noun —> [tet’phan]
‘ added’ ‘layer’ ‘added layer’
b. /t’el_(/noun + /§’al/noun -> [t’el_<’._s_’al]
‘rice cake’ ‘pattern’ ‘a mold for rice cake’
/pa.lS/noun + Eli/mun -> [pak’.s’il
‘pumpkin’ ‘seed’ ‘seed of pumpkin’

(5a) shows that aspirate obstruents are not tensed after unreleased stops (8’). It

is not clear in (5b) whether tense obstruents are tensed after unreleased stops (8 1) or not.

A comparison of (6a) to (6b) indicates that unreleased stops [8’] in syllable-final

position trigger tensification of the following lax obstruents.

(6) a. /maktae/ —> [mal_(’._t’ae] ‘stick’
/mac+§arj/ -—) [mat’.s’arj] ‘a kind of table setting for two people’

b. /pin_ta/ —) [pingae] ‘bedbug’
Admag/ —-> [kimban] ‘soon’
c. /polki/ -—> [polgi] ‘hip’

In (6a), /t/ and /s/ are tensed to [t’] and [s’] afier unreleased stops [k’] and [t’]

respectively. In general, as shown in (6b) and (6c), /t/, /p/, and /k/ are not tensed after
sonorants such as nasals or the lateral /l/, except for two cases shown in (7) and (8).
Instead, /t/, /p/ and /k/ are voiced between two [+voice] sounds, resulting in [d], [b] and
[g] respectively. Thus, having an unreleased stop in syllable-final position is a critical

condition for ObsT.

144

However, there are two cases in which obstruents are tensed after nasals or the

lateral /l/.
(7) 3- AIS/verb stern+ Ad/lgertmdial suffix '9 [Uf-ls’i] ‘laughing,
/ US/ verb stern + / ta/ declarative marker ‘9 [Ur-1’3] ‘laugh,
b- mil/verb stern + /.Ki/gerundial suffix '9 [all-Ki] ’tO hug,
/ 311/ verb stern + /1_(0/ conjunction marker "9 [333,0] ‘tO hug and,

/ atfl/ verb stern + / 93/ verb ending for indirect command —) [an $23] ‘let’s hug,
/ an/ verb stem + /§0/ verb ending for indirect command ‘9 [3:05.13] ‘Please hug,

c. /mU_l/verbstem +/l_d/noun derivational suffix —> [Innigil ‘biting’
/n0.l/verb stern + Ala/declarative marker —> [110148] ‘play’
First, obstruents are tensed after nasals if nasals end verb stems, as shown in (7b).
(7c) indicates that within the verbal morphology, obstruents are not tensed after the
lateral /l/. Thus, in the verbal morphology, triggers of ObsT are obstruents, which are
unreleased in syllable-final position, and nasals, as shown in (7a) and (7b) respectively.

Second, in Sino-Korean words, /t, c, s/ after the lateral /l/ are tensed, as shown in

(8)-

(8) Sino-Korean words

a. /kal_cirj/ —> [kalc’irj] ‘thirsty’
/paltal/ —> [pa]. t’al] ‘development’
[115a I]/ —> [i1.5’a13] ‘everyday’

b. /malki/ —9 [malgi] ‘the end period’

/pulp_ok/ —> [pulbok] ‘disobedience’

 

‘ If lki/ is a passive suffix, instead of a gerundial suffix, [k] is not tensed after nasals within the verbal
morphology, as shown in [an.gi.ta] from /an+ki+ta/.

145

In data (8), /t, s, c/ are tensed after the lateral /l/ in Sino-Korean words, presumably
because /t, s, c/ share the tongue tip articulator with /l/. The phonetic motivations behind

PNT in (7b) and PLT in (8a) are discussed in detail in chapter 4.

146

3.2 The gesture-based account of Post-Obstruent Tensification

3.2.1 Introduction

In section 2.1, I have shown that CN in Korean is motivated by the lack of
contrastive CD gestures, which distinguish different types of obstruents in syllable-final
position. For example, three types of Korean stops are neutralized in syllable-final
position due to the lack of contrastive glottal gestures.

This section argues for the following three points: First, occurrence of ObsT is
predictable based on CDs of C2 in the sequence of VlOlC2V2 (O: obstruents hereafter).
Second, contrary to CN, POT results from enriched tense gestures at the oral-closure
release of 02 in the sequence of VlOlOsz, where 02 is a lax obstruent. Third, there is
a close relationship between distribution of tense gestures in the sequence of V1C1C2V2

and morphological conditions of ObsT.

3.2.2 Occurrence of Obstruent Tensification

The gesture-based account of POT begins with what gestural characteristics of 01
are in the sequence of VlOlOsz. As will be illustrated in section 3.2.3, ObsT is
sensitive to CDs at the SL level for C1 and C2 in the sequence of VC1C2V.

In the discussion of CN, the output CDH for stops, affricates, and fricatives in

syllable-final position is represented as in (9).

147

(9) The output CDH for stops, affricates, and fricatives in _{#, C}.

VT (occlusion)
7" SL
lfl°§s<31____
’I/ ,” Oral
'1’ ’1’ [closed]
’I I, , v ’ as
I,” ,’ Tongue ’\
’i’ ’I' [closed] \\
Glottal Nasal Lateraf Ce‘ntral Lips
[closed] [closed] [closed] [closed]
I I I,” ~ ~ ‘ s ~ [p 1 ]
VEL TB TB TT
[closed] [closed] [closed]
[k1] [t’]

01 in VlOlOZVZ is always an unreleased stop, which is characterized by [closed]
CDs at the SL and all levels below the SL level, as shown in (9). In addition, recall that
the CD of glottal gestures of an unreleased stop is as narrow as that of voicing by default
(Browman and Goldstein 1989). Sawashima et al. (1980) also characterize unreleased
stops in word-final position as having a small degree of glottal opening.

The following table displays what happens to C2 after an unreleased stop when C2

ranges from stop to the lateral /l/ in the sequence of VlOlC2V2.

148

(10) Phonetic realizations of C2 after unreleased stops.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

01 C2
Output Input Output
GLO SL ORAL Stops
[closed]5L mid] [closed] [closed] lax C2 Tensification
[closed] Ora] [closed] lclosed] [closed] tense tense
bpen] [closed] [closed] aspirate aspirate
Affricates
[mid] [closed]/ [closed]/ lax C2 Tensification
[critical] [critical]
[closed] [closed]/ [closed]/ tense tense
Eritical] Eritical]
[open] [closed]/ [closed]/ aspirate aspirate
[critical] [critical]
Fricatives
mild] [critical] [critical] lax C2 Tensification
[closed] [critical] [critical] tense tense
open] [open] [open] /h/ C1 Aspiration
[critical] [open] [closed] nasals C1 N asalization
[critical] f1:[openllfffi: [open] /1/ C1 N asalization

 

 

 

 

 

(10) illustrates two important points about POT in Korean. First, it reveals a clear
gestural condition for POT. In VOIC2V2, when C2 does not have the [open] CD at the
SL level, C2 participates in POT. When C2 has the [open] CD at the SL level (shaded
areas in (10)), C2 does not participate in POT except for certain cases that will be

discussed later in chapter 4. Therefore, POT is sensitive to CDs at the SL level, which

are predicted from CDs of the nasal and oral levels.

Based on (10), I suggest that one of the most important gestural conditions for
tensification of lax-C2 is that the CD of C2 should not be the [open] CD at the SL level.

Therefore, ObsT is a phonological phenomenon captured at the SL level, as illustrated in

(11).

149

 

(1 l) Tensification of C2 01 Nasalization

 

 

a. 01 : C2 b. 01 : C2
Input /ip E ku/ ‘entrance’ lak 5 ma/ ‘devil’
SL closed : closed closed : open
Oral closed l closed closed i closed
Output [ika’u] [ag‘ ma]

In (11a) and (11b), 015 have identical CDs at the SL and oral levels. However,
(11a) and (11b) differ from each other at the SL levels. C2 of (1 1a) has the [closed] CD
at the SL level and C2 of (11b) has the [open] CD at the SL level. When the CD of C2 is
[open] at the SL level, 01 is assimilated to C2 according to gestural characteristics of C2,
as shown in (11b). When CD of C2 is [closed] at the SL level, C2 is tensed. Therefore, in
Korean, ObsT occurs when CD of C2 is not [open] at the SL level, as shown in (10).

The second important point that (10) displays about POT is that only lax-02s are
tensed in the sequence of VOlOsz. Therefore, lax and tense 025 are neutralized to
tense obstruents while aspirate 023 are maintained. I suggest that neutralization of lax
and tense 028 in the sequence of VOlOsz is closely associated with distribution of

tense gestures at the oral-closure release of 02.

150

3. 2. 3 Distribution of tense gestures and Post-Obstruent Tensification

As the condition for occurrence of ObsT has been introduced, it is now necessary
to find out what tense gestures are distributed in the sequence of V10102V2, where 02 is
a lax obstruent.

To further understand POT, it would be helpful to look at some experimental
studies that examine the sequence of V10102V2 in Korean. Silverman and Jun (1994)
measure oral flow, SL pressure, and pharyngeal pressure in order to investigate co-
articulation effects in the sequence of V10102V2. Their study provides some answers to

what happens to 0102 in the sequence of V10102V2, as shown in (12a) and (12b).

(12) a. /ipku/ (Silverman and Jun 1994:215)

Tongue body
Backward Dorsum

Forward Lips

 

Pressure High

 

 

Low
Time >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

 

Tongue position Closure Period ...——
Oral Pressure —— Pharyngeal Pressure ......

151

b. /ikpu/ (Silverman and Jun 1994:217)

 

 

Tongue body
Backward Dorsum B /k/ D
_h - . - .
Forward Lips A' ' ' "' ' "—73—- C
Pressure __________________
High ," -‘|‘
I I
’ ‘i
Low “-
Time >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Tongue position ' ‘ ‘ ' Closure Period =—
Oral Pressure -—- Pharyngeal Pressure ------

In (12a), point A represents the beginning of the oral closure for /p/. Point B
represents the beginning of the oral closure for /k/ in /ipku/. As the closure gesture of /p/
starts, the oral pressure and the pharyngeal pressure are increased. After the [closed] CD
of the TB for /k/ is formed, the increased pharyngeal pressure is still maintained because
of the [closed] CD of the TB. Contrary to the increased pharyngeal pressure, the oral
pressure is rarefied after the [closed] CD of the TB for /k/, due to the oral cavity
expansion. The TB is moved from the front vowel /i/ to the back vowel /u/ while keeping
[closed] gestures for /p/ and /k/.

In (12b), as the [closed] gesture of /k/ starts, the pharyngeal pressure is increased
but the oral pressure is rarefied due to the expanded oral cavity. After the [closed] CD of
the lips for /p/ is formed, the oral pressure stays low until the end of the oral burst for /k/.

However, the pharyngeal pressure is sustained because of the [closed] CD of the lips for

/p/.

152

Note that the increased pharyngeal pressure is maintained for an extensive period
as long as the oral closure exists in the oral cavity, as shown in both (12a) and (12b).
However, the oral pressure varies according to oral gestures, as shown in the comparison
of(12a) and (12b).

(12) illustrates two important points relevant to POT. The first important point is
the long closure duration between two vowels in V10102V2, as shown in the first
diagrams in (12a) and (12b). I assume that long closure duration between two vowels in
V10102V2 acoustically and articulatorily promote tensification of lax-025.

According to Han (1996), closure duration is a primary perceptual cue that
distinguishes tense consonants from lax consonants between two vowels. In addition,
Oh’s study (1998) illustrates that closure duration of tense obstruents between two
vowels is always longer than that of lax consonants regardless of speech style, as shown

in (13). Oh’s observation is also consistent with Kim (1965 and 1970).

(13) The overall closure duration of consonant clusters between two vowels.

(Oh 1998: 319 and 320)

Careful Speech Normal Speech

VIC 1V2 shortest/r VIC 1V2

V1C’1V2 V1C’1V2/ V1C1C’2V2
V1RC’2V2 / V1C1C’2V2

V1X1C’2V2/ longestslr V1RC’2V2 / V1X1C’2V2

R=sonorant. C1 and Xl= homorganic and non-homorganic with C’2 respectively.

Note that C’s in V1RC’2V2, V1C1C’2V2, and V1X1C’2V2 are from lax or tense
obstruents. Closure durations of consonant clusters in V1RC’2V2, V1C1C’2V2, and

V1X1C’2V2 are either longer than that of V1C’1V2 or are similar to that of V1C’1V2

153

regardless of styles of speech. Therefore, I assume that lax-02s in the sequence of
V10102V2 are tensed because one of important gestures of tense obstruents, namely
long closure duration, is distributed between two vowels.

The second important point that (12) illustrates related to POT is the increased
pharyngeal pressure between two vowels in the sequence of V10102V2. The increased
pharyngeal pressure is maintained in (12) until the end of the oral-closure release of 02 in
V10102V2, regardless of Oland 02’s places of articulation. The heightened pharyngeal
pressure, which is maintained for a long period, is closely associated with stiffness in the
vocal tract, according to Dart (1987).

Dart (1987) states the heightened vocal tract pressure results from the stiffened
vocal tract wall tension, as shown in (14). Namely, as the vocal tract tension is increased,

the vocal tract pressure is increased as well.

(14) From Dart (19842143)
The pressure increase can be understood as a result of stiffening the walls, which
is effectively the same as decreasing supra-glottal cavity volume, since it calls for
a reduction in the possibility of passive vocal tract expansion. This decrease in
elasticity of the cavity wall also contributes to a lower peak flow by decreasing
the amount of elastic recoil of the walls and thereby slowing down the initial flow
velocity at release.
According to Dart (1987), Kim (1965, 1970) and Hardcastle (1973), the
heightened pharyngeal pressure and the greater muscular tension in the vocal tract wall
are important properties of tense obstruents. Therefore, I assume that the heightened

pharyngeal pressure and stiffness in the vocal tract during oral closure of 0102 in

V10102V2 are also responsible for tensification of lax 025.

154

Sawashima et al. (1980)’s study also provides insight for glottal gestures between
two vowels in the sequence of V10102V2. According to their fiberscopic observation of
laryngeal gestures, glottal gestures of 01 are assimilated by glottal gestures of 02
regardless of 02’s places of articulation. In Sawashima et al.’s study, glottal gestures
between two vowels are divided into three groups.

First, when 025 are lax or tense stops, Sawashima et al. (19802129) observe that
“the degree of glottal opening for the consonant sequences of the syllable-final applosives
followed by the lax and forced stops is almost the same as that of the syllable-initial
forced stop.” Second, when 025 are lax fricatives, “there is a fairly large glottal opening
for both the lax and forced [tense] fricatives” (Sawashima et al. 19802129). However,
there is no significant difference between tense and lax 02 fricatives in terms of glottal
opening. Third, when 025 are aspirate stops, the glottal opening at the oral-closure
release of 02 is as wide as that of syllable-initial aspirate stops.

I provide two reasons for the fact that lax and tense stops of 02 pattern together in
terms of glottal opening in V10102V2. The first reason may be closely related to the fact
that glottal gestures of 01 are assimilated by those of 02 in V10102V2. This means that
glottal gestures of lax-O2 stops may begin while oral closure of unreleased stops of 01 is
maintained, as schematically shown in (15a) and (15b). In (15) and hereafter, the curved-
broken line on the glottal tier represents the actual glottal gestures in V10102V2. The
curved-solid line on the glottal tier represents the targeted glottal gestures for lax stops in

the input, which are contrastive in Korean.

155

(15) a. b.

 

 

 

 

 

 

 

 

oral-closure oral-closure
release relexse
Oral open 01 lax-02 stop Oral open 01 lax-O2 stop
close close
GLO open GLO open
close v"’—7‘§\ close m\
—> Time —> Time

As shown in (15a), the glottal Opening for lax-02 stops may have reached and
completed its targeted CD significantly earlier than that of lax stops in the input.
Therefore, by the time lax-02 stops are released, the CD at the glottis is already as closed
as that of tense stops. Recall that the glottal opening of tense obstruents comes to nearly
complete closure before oral-closure release, as mentioned in section 1.3.2.].

The second reason why lax and tense stops of 02 pattern together in V10102V2
may be related to the fact that there are no two glottal gestures relevant to 01 and 02
respectively. This may mean that the CD of glottal gestures for lax-02 stops begins with
the glottal opening value of 01, rather than from a point where lax stops in the input
would normally start. Because the CD of glottal gestures for 01 is very narrow, the
glottal gestures for lax-02 stops do not reach their target gestures (the broken line in
(15b)) within a given time and come to near closure before the oral-closure release.
Therefore, glottal gestures for lax-02 stops undershoot their target gestures, resulting in
the [closed] CD at the glottis before oral-closure release for 02 (Browman and Goldstein
1990)

In summary, when 025 are lax stops in V10102V2, the phonetic

implementational conditions promote tensification of lax-O2 stops because of long

156

closure duration and the minimal glottal opening at oral—closure release for 02, which are
coupled with heightened oral pressure and tension in the vocal tract. The similarities
between lax-025 in V10102V2 and tense obstruents are summarized in (16) based on
Han (1997), Dart (1987), Kim (1965 and 1970), Hardcastle (1973), Oh (1998), and

Sawashima et al. (1980).

(16) Phonetic implementational conditions between two vowels in V10102V2.

 

 

 

 

 

 

 

 

 

phonetic V10102V2 VO’V
implementational

conditions

oral-closure duration as long or longer than C’s long

pharyngeal heightened heightened

pressure

vocal tract high muscular tension in high muscular tension
tension the vocal tract

glottal opening at oral- as narrow as that of tense narrow

closure release

 

Sawashima et al. (1980) observed that lax-O2 fricatives in the sequence of
V10102V2 are phonetically realized as tensed fricatives although the glottal opening for
the lax fricative /5/ tends to be slightly large. This observation indicates that in addition
to the glottal opening at oral-closure release of 02, other factors such as the heightened
pharyngeal pressure and tension in the vocal tract are also responsible for tensification of
lax-02 as long as the CD at the SL level is not [open], as discussed in Rhee (1997), Kim-
Renaud (1974), and Sohn (1987).

If the CD at the SL level is [open], there is no heightened pharyngeal pressure and
tension in the vocal tract because air flows freely through either the oral or nasal cavity.

When 025 are fricatives in V10102V2, although the glottal opening is slightly large, the

157

heightened pharyngeal pressure and tension in the vocal tract are expected because CD5
at the SL level are [closed] for 01 and [critical] for 02. Therefore, narrow CDs at the SL
level for 01 and 02 in V10102V2 promote tensification of the lax fricative /s/.

Now is an appropriate time to explain why lax-025 in V10102V2 are not
aspirated. One of the answers to this question is that contrary to lax and tense obstruents,
the glottal opening for aspirate 02 stays wide open at the time of oral-closure release, as

shown in (17).

(17) Gestural score for aspirate 02 in V10102V2

oral-closure

 

 

 

 

 

release
Oral open A
01 aspirate 02
close
GLO open [fl ‘
close I”
->Time

When 025 are aspirate obstruents in V10102V2, no phonetic implementational
conditions for aspirate obstruents are altered. Relative long oral-closure duration may
cause heightened pharyngeal pressure and tension in the vocal tract. As mentioned in
1.3.2.2, the heightened pharyngeal pressure and tension in the vocal tract are also
characteristics of aspirate obstruents. However, the wide glottal opening at oral-closure
release distinguishes aspirate obstruents from lax and tense obstruents.

As mentioned above, when 025 are lax obstruents in V10102V2, phonetic

implementational conditions promote tensification of lax 025 because of the relatively

158

long oral-closure duration and [closed] glottal opening before or at oral-closure release,
which increases the heightened oral pressure and tension in the vocal tract. Therefore, to
have the [open] CD at the glottis at the time of oral-closure release for lax-025 in
V10102V2 would require an extra effort from speakers. According to Flemming (1995)
and Kirchner (2000), speakers tend to minimize articulatory effort. This means that
speakers tend to use gestures that are distributed in V10102V2. Therefore, lax-02s in
V10102V2 are not aspirated because there is no [open] CD at the glottis at the time of
oral-closure release for lax 025.

Based on (16) and Steriade’s (199727) claim that relative poverty or opulence of
cues induces neutralization or contrasts of a relevant feature, I propose a constraint
sensitive to distribution of tense gestures in V10102V2, as in (18). As shown in (11) and
(16), properties of tense are abundant when CDs of the SL level are not [open] for 01 and
02. In other words, gestures for lax obstruents are absent or not abundant in V10102V2.

This fact is reflected in ( 18).

(18) a. . b.
_-_91____3 92; ____________ *[midleLo/V101_V2
SL 519.899- -1.019§*~:4/99.ti.c.a.l- -.
Oral closed 1 closed/critical
(18a) identifies conditions where gestures to tense are abundant. The constraint
*[mid]GLo/V101__V2 (18b) prevents lax obstruents from appearing after obstruents
because there are no gestures for lax-02 obstruents in V10102V2. Examples of such

gestures include relatively short oral-closure duration, low air pressure and tension in the

vocal tract, and a relatively wider CD at the glottis than that of tense obstruents. The

159

constraint *[mid]GL0/V101__V2 is based on a suggestion from Steriade (1997) that if

there are no cues to a contrast in a given context, the contrast is not maintained, as shown

in (19):

(l 9) *[tense]/ in positions lacking contextual cues
(abbreviated: Context Cues (tense)) (Steriade 1997:70)

Stops cannot be tense or lax in positions where contextual cues to this
contrast are necessarily absent.

3.2.4 Morphological conditions and distribution of tense gestures

As CD5 at the SL level are predicted from the nasal and oral levels, any changes
in CD5 of the velum or oral level in C1 change CD5 of the SL level. Because POT is
sensitive to CD5 of the SL level, as shown in (10) and (11), it would be useful to examine
what happens when Cl is a nasal or the lateral /l/ in the sequence of VC102V. Nasals are
specified with the [open] CD at the SL level because of the velum opening, but have the
[closed] CD at the oral level, as shown in (20a). The lateral /l/ is specified with the

[open] CD at the SL level and the [open] CD at the oral level, as shown in (20b):

 

 

 

 

 

 

(20) a.
C1 02 Phonological alteration
Nasals SL oral and conditions
[open]SL stops [closed]§L [closed]%_ . Post-Nasal Tensification
[Closed]oral affricates [closed]/ [closed]/ - after the verb stem
[critical]§1, [critical]oL
fricatives [critical]st [critical]or11

 

 

 

 

 

 

160

 

 

 

 

 

 

 

 

 

 

 

Cl 02 Phonological alternations
/1/ SL oral and conditions
[open]SL stqas [closedJSL [closed]o,al Post-Lateral Tensification
[Open]oral affricates [closed]/ [closed]/ - when 02 is only /t, s, c/
[critical]§L [critical]o,al_ - in Sino-Korean
fricatives [critical]SL [critical]o,§

 

A comparison of (10) and (20) shows that the more [open] the CD in C 1, the more
conditions are attached to ObsT. For example, in (10), where Cl is specified with the
[closed] CD at the SL and Oral levels, ObsT occurs obligatorily across all lexical
categories within an accentual phrase. When C1 is the lateral /l/ specified with [open]
CD at the SL and Oral levels, as shown in (20b), ObsT is limited to Sino-Korean nouns
and cases where 02 is /t, s, c/.

I assume that the most favorable or strongest phonetic implementational condition
for ObsT is when ObsT occurs obligatorily across all lexical categories. Favorable
conditions for tensification of lax 025 in VC 102V are summarized as harmonic scales

based on CD of the SL and Oral levels.

(21) Favorable conditions for tensification of lax 025 in VC102V

 

 

 

 

21- -91----392 ..... b- -.C.1 ...... 1.9.2 ....... 9- -91---.:-9.2 ..... .
SL closed I closed 3 open 1 closed 3 gen I closed
Oral closed 1 closed closed : closed open : closed

A 3 B: A is more favorable than B for tensification of lax 028.

When 025 are lax obstruents, the most favorable condition for tensification is

when C1 has the [closed] CD at the SL and Oral levels, as shown in (21a). In other

161

words, there are no lax obstruents after unreleased stops. The least favorable condition
for tensification of lax 025 is when CD for C1 is [open] at the SL and Oral levels, as
shown in (21c). If the SL and Oral levels have [open] CD as in the lateral /l/, ObsT
occurs only in limited cases. In general, when C1 has the [open] CD at the SL level or
the [open] CD at both SL and Oral levels, lax 025 are voiced rather than tensed, as shown
in data (6b) and (6c).

The patterns in ObsT in (10) and (21) also exhibit the close relationship between

distribution of tense gestures and patterns of ObsT, as shown in (22).

 

 

 

 

 

 

 

 

 

 

 

 

 

(22)
a. 01+02 b. nasal+02 c. /1/+02

oral-closure duration long long short

oral tract closure of close close open

coda

CD5 at the SL level [closed][close/critical] open] [openL

airflow during the no yes through nasal yes, through the

oral articulator cavity oral cavity

closure

air pressure behind strong ¢ > weak

the oral closure

tension in the vocal high ¢ ’ weak

tract low

tense properties more ¢ ’ less

conditions for ObsT across all lexical after verb stem only when 02 is
categories except co- /t, s, c/ and in
compounds Sino-Korean

 

 

 

 

 

(22) illustrates that patterns of ObsT in Korean are closely related to CD at the SL
and Oral levels. CD at the SL and Oral levels influences pharyngeal pressure and tension
in the vocal tract according to degrees of airflow escaping through the nasal or oral

cavity. Therefore, ObsT occurs obligatorily after obstruents across all lexical categories

162

except co-compounds. This is because favorable phonetic implementational conditions
for ObsT are rich in the sequence of V10102V2.

However, lexical and/or phonological exceptions appear in V1N02V2 (22b) and
V1L02V2 (22c) (N : nasals and L: the lateral /l/). Especially, the pharyngeal pressure and
the vocal tract tension are weak in the sequence of V1L02V2, due to [open] CD at the SL
and Oral levels for the lateral /l/. Phonetic implementational conditions for ObsT are
weak in (22c). Therefore, lax 025 in V1L02V2 (22c) are tensed only when
morphological and additional phonological conditions are attached.

The relationship between conditions for ObsT and distribution of tense gestures in

(22) supports Anttila’s claim (2000: 28) for the emergence of morphology in Phonology.

(23) The Emergence of Morphology: Extra—phonological (morphological, lexical)
conditions emerge in environments where the phonological conditions are at their
weakest.

According to Anttila (2000), morphological or lexical conditions are attached to a
phonological process when phonological conditions to the process are not strong.
Therefore, the less tense gestures are distributed in a given context, the more
morphological and/or additional phonological conditions are likely to be attached to

ObsT as claimed by Anttila. (24) summarizes basic perspectives discussed in this

section.

163

(24) 0‘1 + 02 >>N + 02» L + 02>>h + 02

Gestures Gesturel Gesturel Gesturel 0
Of Cl Gesture2 Gesture2 0

that promote Gesture3 o -

tensification. 0 0

of lax 025 0

 

Influence strongest ‘ F weakest
Morphonological NO Yes Yes No ObsT

conditions

Gestural characteristics of consonants in syllable-final position influence the
following consonants to different degrees, as shown in (24). Thus, unreleased stops in
syllable-final position strongly influence the following lax obstruents. As a consequence,
POT occurs obligatorily across morphological categories within an accentual phrase.

The lateral [l] in syllable-final position weakly influences the following lax
obstruent. As a consequence PLT appears with the morphological and additional
phonological conditions, as shown in (20). Thus, different degrees of influence on the
following lax obstruents are reflected in terms of morphological and/or additional
phonological conditions.

When /h/ specified with the [open] CD in the vocal tract is C1, as shown in
V1h02V2, ObsT does not occur in 02, instead aspiration occurs. Therefore, ObsT is

sensitive to CD in the vocal tract, especially CD at the SL and Oral levels.

164

3.2.5 The gesture-based account of POT

In this section, based on the harmonic scales in (21), I propose a set of constraints

that ban lax obstruents in 02 position in the sequence of V1C102V2, as in (25).

(25) Distribution of tense gestures and ObsT

*[mid]GL0/V101_V2 >> *[mid]GL0/V1N_V2->> *[mid]GL0/ V1L_V2

.......... 01029a8a1021atera1+02
SL .- - 5195993919524 ............ 9 299,; 9195.94 .......... 0 net [9128551. - - -
Oral closed j closed closed I closed open 3 closed

A>>B = A is higher than B in a constraint hierarchy.

The enriched tense gestures at oral-closure release of 02 in V10102V2 are closely
related to 01 having [closed] CD5 at the SL and Oral levels. Thus, *[mid]GLo/

V101__V2 states that 02 cannot be lax in sequences of V10102V2 because properties of
tense, such as long oral-closure duration, [closed] CD at the glottis at the oral-closure

release, and heightened air pressure behind the oral closure are abundant in V10102V2,
as shown in (22a). *[mid]GL0/ V1N_V2 states that lax obstruents are prevented from

occurring after nasals due to the [open] CD at the SL level and the [closed] CD at the oral
level. Due to the [critical] CD at the glottis and the [open] CD at the SL level of nasals,
the amount of air accumulated behind oral closure is relatively weaker in the sequence of
V1N02V2 than that of V10102V2. Therefore, voice gestures, rather than tense gestures,
are increased in V1N02V2 (Kager 1999 and Ito et al. 1995). As a consequence, if ObsT
occurs after nasals, it does so with a morphological condition, as will be shown in chapter

4.

165

*[mid]GL0/V1L_V2 prevents lax 025 from occurring after the lateral /l/. The

sequence of V1L02V2 has the least favorable conditions for occurrence of ObsT due to
the [open] CD at the SL and Oral levels of the lateral /1/. In other words, ObsT would
occur in a more limited way after the lateral [1] than after nasals or obstruents.

Whenever a segment or a gesture in the input is banned from occurring in the
output, as in (25), a faithfulness constraint is at risk of being violated. Recall that IDENT-

IO[G] requires that gestures in the input be preserved in the output, as shown in (26a).

(26) a. IDENT-IO[G]:
Gestures in the input are maintained in the output.

b. *[mid]GLo/V101_V2>>*[mid]GL0/V1N_V2 >>
*[mid]GLo/V1L__V2 >> DENT-IO[G]

Tense or aspirate obstruents appear after obstruents, as shown in data (5). Tense,

aspirate or voice obstruents appear after nasals or the lateral /l/ as in [palt'al]

'development, [kan.thon] 'adultery', and [kulbi] 'a kind of fish'. Therefore, actual outputs

never have GLO [mid] after obstruents, nasals or the lateral /1/ except in co-compounds

as in [cip’. gip’]co-compound ‘every house’ from /cip/noun ‘house’+ /_c_ip/noun ‘house’. As a

result, CD of glottal gestures is not maintained in the output, as shown in data (1), (5),

and (6) (e.g. [ik’s’al] from ligal/noun 'funny act or saying', /pin_tae/ -> [pingae] 'bed bug'

and [pol.gi] from /po]ki/ 'hip'). Therefore constraints against GLO [mid] afier obstruents,

nasals or the lateral [l] are ranked higher than [DENT-IO[G], as shown in (26b).

166

POT occurs obligatorily when an obstruent is preceded by another obstruent
within an accentual phrase regardless of morphological category, as shown in (27a),

(27b), (27c), and (27d).

(27) a. Nouns
/iksal/ —-> [ik’ .sLal] ‘funny act or sayings’
/ma_lgtee/ -—> [mak’ .t’aa] ‘stick’
b. Verbs (verb stern +inflectional suffix)
mAK +1, adeclaritive -> [mAk 1 .t’ a] ‘ eat’
pis +koconjmcfion —-> [pit’ ._lg’o] ‘to comb + and’
cf. an + Kidajvation +ta —> [an.gi.da] ‘be hugged’
c. Modifier +noun
yep modifier 'next' + Qip 'house'—) [yep1 .g’ip] ‘next house’
d. Prefix + noun
pus ‘first’+ _s_akwa ‘apple’ —> [put’s’akwa] ‘fresh apple’
e. Nouns with a sonorant in C1 position
/kog_lgi/ —> [koggi] ‘air’ or 'rice bowel'
lolkull -—) [91. gul] 'face'

Obstruents that follow obstruents are never voiced, as shown in (27a), (27b),
(27c), and (27d). However, in general, obstruents after nasals or the lateral [l] are voiced,
as shown in (27c) but tensed in the verbal morphology or in some Sino-Korean nouns, as
shown in (7b) and (8a).

In (28), output candidates from lexical entries with different types of obstruents in
C2 position are evaluated against constraints and the rankings of constraints suggested in

(26).

167

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IDENT-10[G]
A. /apc;ok/ 'front side' /ap/ ‘front’ + /_c’ok/ ‘side’
/c’/: GLO closed], SL [close/critical], Oral [close/critical], TT [close/critical]
“1139-, Ok/ * [midloLo/ * [midloLo/ * [midloLo/ IDENT'
V101 V2 VIN V2 VlL V2 IO[G]
a. an“ .on *1 ' " ' ’ NNNNNNNNN *(GLO)
><?"b.ap1 .c’ok
0. ap ’ .ghok * !(GLO)
B. /pakflal/ ‘deprive’
/t”/: GLO [open], SL [closed], Oralklosed], TT closed]
makf‘al/ *[midIGLo/ *[midloLo/ *[midIGLo/ IDENT-
V101 V2 VIN V2 VIL‘V2 10191
b. pak’ .t’al *l(GLO)
G” c. pak’ .thal
C. /makt_a:/ ‘stick’
/t/: GLO [mid], SL [closed’ , Oral [closed], TT closed]
“flaky“ *[midloLo/ *[midloLo/ *[midloro/ IDENT'
V101_V2 VIN-l2 . V1L V2 .. . [9191 ..
a-mak‘tae *1 ......
> CF‘bmak’ .t’ze *(GLO)
> c.mak’ .f'ze *(GLO)

 

 

 

 

 

(28A) and (288) illustrate that when the second consonants are aspirate or tense
obstruents, constraints and their ranking suggested in (26) select candidates with aspirate
or tense obstruents in C2 position as the most optimal candidate. However, as shown in
(28C), when the second consonant is a lax obstruent, constraints and their ranking in (26)

select both (28Cb) and (28Cc) as the most optimal candidates.

168

This is because both

 

 

 

(28Cb) and (28Cc) optimally satisfy output constraints in (26). Therefore, there must be
a constraint which selects (28Cb) over (28Cc) as the most optimal candidate.

I suggest that the selection of (28Cb) over (28Cc) results from interaction between
[DENT-IO[G] and a constraint sensitive to articulatory effort in V10102V2. As
mentioned above, when 025 are lax obstruents in the sequence of V10102V2, phonetic
implementational conditions promote tensification of lax 025. This is because gestures of
tense are distributed between two vowels in the sequence of V10102V2.

In particular, at the time of oral-closure release for lax 025, the CD at the glottis
for lax 025 is already as closed as that of tense obstruents (Sawashima et al. 1980).
Therefore, having the [open] CD at the glottis for lax 025 requires an extra effort from
speakers. However, speakers tend to minimize articulatory effort as discussed in
Flemming (1995) and Kirchner (1998 and 2000). In other words, speakers tend to use the
gestures distributed in the sequence of V10102V2.

Recall that there are no two separate glottal gestures for 01 and 02 respectively in
the sequence of V10102V2 according to Sawashima et al. (1980)’s observation.
Sawashima et al.'s observation is similar to Browman and Goldstein's (1986) observation
of glottal gestures in s+stop clusters such as found in English words 'spin', 'star', and
’skin'.

According to Browman and Goldstein (1986), there are no separate glottal
gestures for /s/ and stops respectively. Instead, glottal gestures of /s/ influence glottal
gestures of /p/, resulting in unaspirate stops (see 2.1.3.2.2). According to Sawashima et
a1. (1980)’s observation, glottal gestures of 01 are assimilated by glottal gestures of 02 in

V10102V2. This means that there is an effort to minimize articulatory gestures between

169

two vowels with respect to glottal gestures. According to Flemming (1995) and Kirchner
(2000: 529), the principle of minimization of effort is universal and is computed for each
candidate representation.

Based on the principle of minimization of articulatory effort, and the observation
of Sawashima et al. (1980) and Browman and Goldstein (1986), I suggest a constraint
sensitive to articulatory effort of glottal gestures between two vowels, as shown in (29).
Note that constraints sensitive to gestural efforts may be sensitive to articulators involved
in a given context (Kirchner 1998, Cho 1998a and 1998b and Byrd 1996).

Constraint (29) promotes reduction of glottal gestures between C 1 and C2.

(29) Minimization of articulatory effort at the glottis (Mini Eff [GLO]=0):

CD5 of glottal gestures for C1 and C2 between two vowels should be identical or
minimal.

[closed] I critical] | mid [open]

\ minimal minimaj minimal ->one violation

 

more than mmrmaI —-> two violations

 

maximum —) three violations

Mini Eff [GLO]=0(hereafter Mini Eff=0) demands no glottal CD changes for C1
and 02 between two vowels. Therefore, Mini Eff=0 is satisfied when CDs at the glottis
for C1 and C2 between two vowels are identical, whether they are [closed], [critical],
[mid], or [open].

CD5 of glottal gestures for C1 and C2 between two vowels are minimal when CD5

at the glottis for C1 and C2 are one degree narrower or wider with respect to each other,

170

as shown in (29). Therefore, GLO [critical] occurs with either GLO [closed] or GLO
[mid]. When difference of CD for C1 and C2 is minimal, as shown in (3 5), Mini Eff=0 is
violated only once.

CDs of glottal gestures for C1 and C2 between two vowels are more than minimal
when these are more than one degree narrower or wider with respect to each other, as
shown in (29). Therefore, GLO [critical] occurs with GLO [open] or GLO [mid] occurs
with GLO [open]. When the CD difference for C1 and C2 is more than minimal, as
shown in (29), Mini Eff=0 is violated twice.

CD5 of glottal gestures for C1 and C2 are maximum when opposite ([open] vs.
[closed]). Therefore, GLO [closed] occurs with GLO [open]. When CD difference for
C1 and C2 is maximum, as shown in (29), Mini Eff=0 is violated three times.

A comparison of (28A) and (28C) to (28B) indicates that Mini Eff=0 is lower

than IDENT-IO[G], as shown in (30). Especially (28Bc) illustrates that although the CD

difference at the glottis between [k’] and [th] in [pak’ ._t_hal] is more than minimal, the

input [open] CD at the glottis for /th/ is preserved.

(30) *[mid]GL0/ V101_V2 >> *[mid]GL0/ V1N_V2 >>
*[mid]GLo/ V1L_V2 >> [DENT-IO[G] >> Mini Eff=0

A. /apc’ok/'front side' lap/ ‘front’ + /c’ok/ ‘side’

/p/: GLO [mid], SL [closed], Oral[closed], Lips [closed]
/c’/: GLO closed], sgclose/cfifich Oral[close/critical], TT [close/critical]

 

 

 

 

 

”139,014 *[midloLo/ *[midloLo/ *[midloLo/ IDENT Mini
V101_V2 V1N_V2 V1L_V "0““ Effzo
a~ap’.90k *1 <99
@“bap’c’ok *
c. ap’ .chok *!(GLO) **

 

 

 

 

 

 

 

171

B. /pakfla1/ ‘deprive’

/k/ 2 GLO [mid], SL [closed], Oral[closed], TB [closed]
/t"/: GLO [open], SLElosed], Oral[closed], TT [closed]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0’3th *[midloLo/ *[midloLo/ *[midloLo/ 13521 $310
V101_V2 V1N_V2 V1L_V2 ' l 1 ‘
a.pak‘-ta1 *1 111mm)“
b. pak‘ red ”(0140)
c1"”c. pak1.thal **
C. /mak12e/ ‘stick’
/k/ 2 GLO [mid], SL [closed], Oral [closed]
/t/: GLO ’midLSL [closed], Oral [closed]
/maktae/ * [midloLo/ * [mileLo/ * [midloLo/ IDENT Mini
V101 V2 VrN V2 V1L V2 'IOK’] Eff—0
a- mak‘tae *1 1:531?:*-:i~i"fiii*1i-.f:i:é‘-iii'ifzsif.
C"b. mak1.t’2e "(GL0) *
c. mak1.thae *(GLO) **!

 

 

 

 

 

 

 

Candidate (3 OAa) violates *[mid]GL0/V101_V2, IDENT-IO[G], and Mini Eff=0.
Mini Eff is violated once because CDs at the glottis for 01 and 02 are [critical] and
[mid] respectively. [c] in [ap 1 .cok] has the [mid] CD at the glottis while [p1] has default

[critical] CD or narrower than [critical] CD in syllable-final position. Therefore, effort is

required from speakers to change the CD of the glottis from [critical] to [mid] in
V1C1C2V2. Candidate (30Aa) violates *[mid]GL0/V101_V2 because the lax obstruent
[c] appears in the position where lax gestures are absent, as mentioned in 3.2.3. (30Aa)

also violates DENT-IO[G] because [0] in [ap1.cok] does not maintain the input glottal

gestures.

172

(30Ac) violates [DENT-IO[G] because [ch] in [ap1.chok] does not maintain the

input [closed] CD at the glottis. (3 OAc) violates Mini Eff=0 twice because CD difference
at the glottis is more than minimal, namely from [critical] to [open].
(30Ab) also violates Mini Eff=0 once because the CDs at the glottis for Oland 02

are [critical] and [closed] respectively. However, violation of Mini Eff=0 is minimal
compared to (30Ac). Although (30Ab) violates Mini Eff=0 once, it is selected as the

most optimal candidate because it does not violate higher ranked constraints
*[mid]GLo/V101_V2 and [DENT-IO[G].

Candidate (3OBa) violates *[mid]GLo/V101__V2 because the lax obstruent [t]
appears in the position where lax gestures are absent. Candidate (3 03a) violates IDENT-

IO[G] because [t] in [pak1.tal] does not maintain the input glottal gestures for /t”/.

(3 08a) violates Mini Eff=0 once because CD5 at the glottis are [critical] and [mid]. [t] in

[pak1.tal] has the [mid] CD at the glottis while [k1] in [pak1.tal] has the [critical] CD or

narrower than [critical] at the glottis.

In (3OBb), IDENT-IO[G] is violated because the [open] CD at the glottis for /t”/ is
not maintained. Mini Eff=0 is violated in (30Bb) once because CD5 at the glottis are
[critical] and [closed].

(30Bc) violates Mini Eff=0 twice because CD difference between [k1] and [th] in

[pak1.f‘a1] is more than minimal. However, (30Bc) is the most optimal candidate

because it satisfies higher constraints, *[mid]GL0N101__V2 and IDENT-IO[G].

In (30C), candidate (30Ca) violates *[mid]GLO/V101__V2 because the lax

obstruent [t] appears in the position where lax gestures are absent. (30Ca) satisfies

173

IDENT-IO[G]. However, satisfaction of *[mid]GL0/V101__V2 takes precedence over

satisfaction of IDENT-IO[G] because *[mid]GLo/V101~V2 is ranked higher than IDENT-
IO[G] in the constraint ranking hierarchy.
Candidates (30Cb) and (30Cc) are tied in *[mid]GL0/V101_V2 and IDENT-

IO[G]. Therefore, Mini Eff=0 plays a.role in selecting the most optimal candidate.

Candidate (30Cb) violates Mini Eff=0 once because CD difference between [k1] and [t']
in [mak1.t_'2e] is minimal. Candidate (30Cc) violates Mini Eff=0 twice because CD
difference between [k1] and [th] in [mak1.tha:] is more than minimal. Thus, [k1] and [th]
in (30Cc) demand more articulatory effort from speakers than [k1] and [t'] in (30Bb) do.

Therefore, (3 OCb) is the most optimal candidate.

As constraints related to ObsT after obstruents, nasals or the lateral /l/ have been
introduced, it is now necessary to argue for the ranking of constraints against GLO [mid]
with respect to *CCD/_{#, C} and other constraints proposed in chapter 2. (31)
illustrates constraints responsible for CN and CCCS.

(31) a. *COMPLEX >> *Max-IOSeg >> TB An/Lips An >> TT,,n >>
*CCD/_{#, C} >> IDENT-IO[G]
b. *Contrastive Constriction Degrees (CCD)/ _{#, C} (From chapter 2 (21))

Contrastive CD at the glottis and at the SL level for obstruents is not allowed
when obstruents appear in word-final position or before a consonant.

*[open] /_{#, C} >>*[mid] /_{#, C} >>*[critical] /_{#, C} >> *[closed] /_{#, C}

174

Because constraints against GLO [mid] after obstruents, nasals, or the lateral /l/
are ranked above IDENT-IO[G], as shown in (30), it is first necessary to examine the
constraint ranking between *CCD/_{#, C} and constraints against GLO [mid] after
obstruents, nasals, or the lateral /l/. As seen in chapter 2, *CCD/_{#, C} is violable but
constraints against GLO [mid] after obstruents, nasals or the lateral /l/ are never violated,
as shown in (30). Therefore, *CCD/_{#, C}, which is responsible for Korean CN, is
ranked lower than constraints against GLO [mid] after obstruents, nasals, or the lateral

[l], as shown in (32).

(32) *[mid]GL0/ V101_V2 >> *[mid]GL0/ V1N_V2 >>
*[mid]GLo/ V1L_V2 >>*CCD/_{#, C}>>IDENT-10[G] >> Mini Eff=0

/saskas/ 'a type of hat'

/5/ 2 GLO [mid], SL [critical], Oral [critical], 'I'I’ [critical]
/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]

 

/sa_s_kas/ *[mid]GLO/ *CCD/_{#, C} IDENT- Mini
V101 V2 IOlGl Eff=0

 

a. saskat’ *1 :1}:iffii‘l’TGLO?115L133?iii. ‘ " '

 

b. sattéat *1 1;;2155j}:.'i‘{"(Glifo‘tS’Igi'is33$:szva *(TTCD)

 

c. saver *1 (812m)

’ *(TTCD)

 

>rrd. sat1.k'at1 “elem _. -.. **<GL0>"‘
1 _ *(TTCD))

 

 

 

 

 

 

 

> 6' ”5"" “61292-1!

 

In (32), because consonants between two vowels are obstruents, only
*[mid]GL0/V101__V2 is displayed in candidate evaluation. Hereafter, in order to

conserve space when evaluating candidates, if a lexical entry does not have obstruents,
nasals, or the lateral /1/ in C1 position, constraints relevant to absent segments will not be

illustrated.

175

Because candidate (32a) maintains target gestures of the input, it satisfies IDENT-

IO[G]. However, it violates higher constraints such as *[mid]GL0/V101__V2 and

*CCD/_{#, C}. (32a) violates *[mid]GL0/V101_V2 because lax [k] appears after [5].

(32a) violates *CCD/_{#, C} twice because of GLO [mid] and TT [critical] for [s] in
syllable-final position. (32a) satisfies Mini Eff=0 because [5] and [k] have the identical
[mid] CD at the glottis. Although (32a) satisfies [DENT-IO[G] and Mini Eff=0 most

optimally, it is not the most optimal candidate because of violations of
*[mid]GL0/V101_V2 and *CCD/_{#, C}, which are ranked higher than DENT-IO[G] in
the constraint hierarchy.

Assuming [t] and [t1] correspond to /s/, (32b) and (32c) violate

*[mid]c,L0/V101_V2 because lax [k] appears after [t] and [t1] respectively. (32b)

violates *CCD/_{#, C} twice because of GLO [mid] and SL [closed] for [t]. (32c)
violates *CCD/_{#, C} once because of SL [closed] for [t1]. (32b) violates [DENT-IO[G]
because the input CD at the TT is changed from [critical] to [closed]. (32c) violates
IDENT-IO[G] twice because the input CD5 for /s/ at the glottis and at the TT are changed.
(32b) satisfies Mini Eff=0 because [t] and [k] have the identical [mid] CD at the glottis.

(32c) violates Mini Eff=0 because CDs at the glottis for [t1] and [k] in [sat1.kat1] are

[critical] and [mid] respectively. Both (32b) and (32c) violate *[mid]GL0/V101_V2 and
*CCD/_{#, C}. Therefore, both (32b) and (32c) are eliminated from being the most
optimal candidate because they violate relatively higher constraints.

Candidates (32d?) and (32e‘>) are tied in *[mid]GLo/V101__V2. Therefore,

*CCD/_{#, C} may play a role in selecting the most optimal output. Candidate (32d) has

176

the [closed] CD at the SL level in syllable-final position, resulting in violation of
*CCD/_{#, C}. Candidate (32e‘>) satisfies *CCD/_{#, C} because there is no obstruent
in syllable-final position due to deletion of /s/ in the input. Therefore, according to
constraints and their rankings in (32), (32e>) should be the most optimal candidate.
However, the actual output is (32d¢“), rather than (32e). The difference between (32d)

and (32e>) is in the alteration or deletion of /s/.

Recall that Max-105,,g is responsible for maintaining of input segments. As
shown in the analysis of CCCS, Max-IOsflg is ranked higher than *CCD/_{#, C}. In

addition, Max-IOscg is violated when satisfaction of *COMPLEX, which bans a consonant

cluster in syllable-final position, is at risk. Otherwise, segments in the input are not

deleted. *COMPLEX and constraints against GLO [mid] after obstruents, nasals or the
lateral /1/ are never violated in output forms in Korean. However, Max-IOscg is violated

in some output forms. Therefore, constraints against GLO [mid] after obstruents, nasals

or the lateral /l/ are unranked with respect to *COMPLEX and ranked higher than Max-

IOSCg, as shown in (33).

177

(33) *COMPLEX, *[mid]GL0/ V101_V2 >> Max-IOscg >> *CCD/_{#, C} >>
IDENT-IO[G] >> Mini Eff=0

/koskan/ 'barn'

/s/ 2 GLO [mid], SL [critical], Oral [critical], TT [critical]

 

 

 

 

 

 

 

 

 

/k/ 2 GLO [mid], SL [closed], Oral [closed], TB 'closed]
/ko_s]<_an/ *COMP *[midloLo/ Max- *CCD/_{#, C} IDENT- Mini
E V101 V2 103% IO[G] Eff=0
b. kosk'an *(GLO) [(GLO) * .
i .. “(SL). , w ,
C- kolek—an i *! if?"3"‘(GL0123-‘1 . *CI’TCID)‘ ’
d. kotk'an E *(GLO) *(TTCD) *
g . *!(SL)! ‘ ,2 w
e. kot1.kan i *l f " "(SLI , *(TTCD) ""
Crf. kot1._k'an 5 *(SL) *(TTCD) , "‘
j . 7 «(3140) ' . {a
g. kok'Lan 5 *1 g ' *

 

 

 

 

 

 

 

Candidates (33b), (33d), and (33f) are tied in *COMPLEX and *[mid]GL0/

V101_V2. However, (33f) satisfies *CCD/_{#, C} most optimally. The comparison of

(331) to (33g) shows that constraints against GLO [mid] after obstruents, nasals, or the
lateral /l/ are ranked higher than *Max-IOseg. (33g) satisfies *[mid]GLo/V101_V2, as

does (33f), by deleting /s/. (33g) does not violate *CCD/_{#, C}, DENT-IO[G], or Mini

Eff=0, whereas (33f) violates all three constraints. However, (33 g) is not selected as the
most optimal candidate because it violates Max-10mg.
Constraints against GLO [mid] after obstruents, nasals, or the lateral /l/ are ranked

higher than Max-IOSCg. TB A", Lips An, and TTAn, are ranked lower than Max-IOscg. By

transitivity, TBAn, Lips An, and TTAn, are ranked lower than constraints against GLO

178

[mid] after obstruents, nasals, or the lateral /l/, as shown in (34).

candidates from a lexical entry with an obstruent-obstruent cluster.

(34)

*CCD/ _{#, C} >> IDENT-IO[G] >> Mini Eff=0

/kapskwa/

/p/: GLO [mid], SL [closed], Oral [closed], Lips [closed]
/s/: GLO [mid], SL [critical], Oral [critical], TT [critical]
/k/ 2 GLO [mid], SL [closed], Oral filosedLTBklosed]

(3 4) evaluates

*COMPLEX, *[mid]GL0/ V101_V2 >> Max-IOscg >> TB AmlLipsAn,» TTAn >>

 

/kapsl_<wa/

‘COMPE *[mileLo/

i V101

V2

Max-

TBAml

TTAi-t

*CCD/
_{#. C}

IDENT-
IO[G]

Mini
Eff=0

 

a. kapskwa

*l

Lips A“,

* *i'v*f*'(G‘LO)f ::"~'<:*i:i'~é;ii7'57;
,215,,,SL)_.., A. ,.

 

b. kaps'k'wa

*l

—:'=i:*s*<9w>vs
“(SL) -

(GL0)

 

c. katkwa

*l

2115110120)?!
L "’(SL I 152.5

 

d. kaskwa

*l

Vi‘~~"‘(SL)“ '

 

e. kat1.kwa

*l

<89

 

f. kapkwa

*1

 

g. kap_1._kwa

*l

;;‘i3}‘<5,.L)-'}<> 2

 

h. kakwa

 

i. kaklwa

 

j. kask'wa

*l

 

k. kat.l_<1wa

*l

"71'? 1.:‘(GL01
2 1*(SLTJ’L' 1

*(rrcn)

 

l. kat1.k_'wa

*l

(SL)

(GL0)
*(TTCD)

 

1n. kapkwa

" ' iii‘(o‘Lo)’

*(SL)

(GL0)

 

 

G'n. kap1 k‘wa

 

 

 

 

 

‘(SL)

 

 

 

In (34), candidates which violate *COMPLEX and *[mid]GLo/ V101_V2 are

eliminated as possibilities to be the most optimal candidate because both *COMPLEX and

*[mid]GL0/V101_V2 are the highest ranked constrants. Therefore, candidates from

(34a) to (34g) are eliminated as possibilities to be the most optimal candidate.

179

 

Candidates (34h) and (34i) are not the most optimal candidate because two input

segments are deleted, violating Max-IOscg twice. Candidates from (34j) to (34n) are tied
in *COMPLEX, *[mid]GL0/V101_V2, and Max-103,2. Therefore, TB An, Lips An, and
TTAn play a role in selecting the most optimal candidate. Because TB An, and Lips An, are

ranked higher than TTAn, candidates violating TB An, and Lips A“ are not possibilities to

be the most optimal candidate. Therefore, candidates from (34j) to (341) are not the most

optimal candidates because the Lip articulators in the input are not maintained.

Candidates (34m) and (34n) are tied in *COMPLEX, *[mid]GL0/ V101_V2, Max-

10368, and TTAn. Therefore, *CCD/_{#, C} plays a role in selecting the most optimal

candidate. (34m) violates *CCD/_{#, C} twice because of GLO [mid] and SL [closed] in
syllable-final position. (34n) violates *CCD/_{#, C} once because of Lips [closed].
Therefore, (34n) is selected as the most optimal candidate.

Constraints and their ranking in (34) correctly select the actual output as the most
optimal candidate regardless of whether lexical items finish with a consonant or a
consonant cluster, as shown in (3 3) and (34). In addition, (34) displays that CN and POT
result from interaction of output constraints sensitive to phonetic implementational
conditions in V1C1C2V2.

Up to this point, discussion has concerned occurrence of ObsT after obstruents.
As mentioned above, occurrence or non-occurrence of ObsT is sensitive to gestural
characteristics of C 1. When C1 is an obstruent, a lax 02 is always tensed in Korean

within an accentual phrase. However, when C1 is a nasal or the lateral /l/, a lax-02 stop is

180

voiced within an accentual phrase except in Korean verbal morphology and some Sino-
Korean nouns (Jun 1993 and 1994).

If ObsT results from interaction of output constraints, non-occurrence of ObsT
after nasals or the lateral /l/ in (6b) and (6c) should result from interaction of output
constraints suggested in (34). In order to see if when 01 is a nasal or the lateral /l/,
constraints and their hierarchy in (34) correctly select actual output forms as the most
optimal candidates, (35A) evaluates candidates from a lexical entry with N+a lax stop /t/.

(35B) evaluates candidates from a lexical entry with L + a lax stop /k/.

(3 5) *COMPLEX, *[mid]GLo/ V1N_V2 >> Max-103,8 >> TB AmlLipsAn,» TTAn >>
*CCD/ _{#, C} >> IDENT-IO[G] >> Mini Eff=0

A. /bint2e/ ‘bed bug’

/n/: GLO [critical], SL [open], Oral [closed], TT [closed]

 

 

 

 

 

 

/t/ 2 GLO Enid], SL [closed], Oral [closecfl, TT 'closed
/bi_n_tze/ I"COMP j *[mileLO/ Max- TB Ana/ TT A [1 *CCD/ IDENT Mini
? V‘N—W ”*2 92w 4‘3"? "’"3"? "‘1’
3- bin._t_2e i 1‘! .1 [I 1‘ ...... 3. 2:51535,1?1,";71312f:* 2‘: ’
b. bint’a: ”1010} *l
c. binf'ae 1 *[GLOI “‘1
ed. biggie i *[GLOI

 

 

 

 

 

 

 

 

 

181

B. *COMPLEX, *[mid]GL0/ V1L_V2 >> Max-IOscg >> TB Am/LipsAn,» TTAn >>
*CCD/ _{#, C} >> [DENT-IO[G] >> Mini Eff=0

lelkul/

/l/ 2 GLO [critical], SL [open], Oral [open], TB [open],TT [closed]
/k/: GLO mid], SL [closed], Oral[closed], TT [closed

 

 

 

 

 

 

/e&ul/ 1COMP[*[mid]GL0/ Max- TB An/ ”A“ *CCD/ IDENI‘ Mini

[V1L_V2 103,8 Lipsin _{#,C} -10[o] Eff=0

a. elkul *1 ...

b. elk’ul *[GLOI *1

c. etkhul E *[GLO] m
9*” d. e[.gul *[GLO]

 

 

 

 

 

 

 

 

 

As shown in (35A) and (3 5B), constraints *COMPLEX, Max-10568, TBA“, LipsAn,

and TTAn are not relevant to candidates from /bintae/ and /elkul/. This is because both

/bint2e/ and /elkul/ do not have more than two consonants between two vowels. The
constraint *CCD/ _{#, C} is not relevant to candidates in (35A) and (35B) because
*CCD/ _{#, C} pertains to obstruents in syllable-final position.

Candidates (35Aa) and (3 5Ba) violate *[mid]GLo/ V1C1_V2 because lax
obstruents are in the position where gestures to lax obstruents, such as relatively short
oral-closure duration, are absent. Both (3 SAa) and (3 SBa) satisfy [DENT-IO[G] because
input target gestures are preserved. However, satisfaction of *[mid]GLo/V1C1_V2 is
more important than satisfaction of [DENT-IO[G] because the former'is ranked higher
than the latter in the constraint hierarchy. Therefore (3 SAa) and (3 53a) are not desirable
as most optimal candidates.

All the candidates in (35A) and (353) except (3 5Aa) and (35Ba) violate IDENT-

IO[G] once because input target gestures at the glottis are not maintained while

182

maintaining oral gestures. All the candidates in (35A) and (3 5B) except (35Aa) and
(3 53a) are tied with regard to IDENT-IO[G] and higher constraints. Therefore constraints
lower than [DENT-IO[G] play a role in selecting the most optimal candidate.

Note that Mini Eff=0 is a gradient constraint, as shown in (29). Identical glottal
CDs of 01 and 02 do not violate Mini Eff=0. There is no displacement of CD5 at the
glottis for 01 and 02. Therefore, the wider the CDs at the glottis for the flanking
segments, the greater the displacement (hence effort) required to achieve a given degree
of consonantal constriction according to Kirchner (1998222).

(3 5Ac) and (35Bc) violate Mini Eff=0 to the greatest degree because glottal
gestures move from the [critical] CD for sonorants to the [open] CD for aspirate
obstruents, thus violating Mini Eff=0 twice. Candidates (3 5Ab), and (3 5Bb) violate Mini
Eff=0 once because CDs of glottal gestures for 01 and 02 are one degree narrower with
respect to each other.

Candidates (3 SAd) and (3 53d) satisfy Mini Eff=0 most optimally because CDs of
glottal gestures are identical, being [critical]. Therefore, according to constraints and
their ranking in (34), (35Ad) and (3 5Bd) are the most optimal candidates and they are
actual output forms in Korean.

Note that after an obstruent, the lax fricative /s/ and the lax affricative /c/ pattern
together with lax stops. However, /5/ and /c/ separate themselves from lax stops in
V1_V2, V1N_V2, and V1L_V2. A lax stop is voiced in V1_V2, V1N_V2, and
V1L_V2 unless constraints related to morphological conditions are active, as will be

shown in section 4.2.2. However, /5/ and /c/ are never voiced in Korean in V1_V2,

183

V1N_V2, and V1L_V2. In addition, /5/ and /c/ are never tensed unless constraints
related to morphological conditions are active, as will be shown in section 4.2.2.

Therefore, a thorough investigation related to phonetic constraints on voicing and
tensing in fricatives and affricates is needed. Because the focus of this study is not about
phonetic constraints on voicing and tensing in fricatives and affricates, the syllable-initial
/s/ or /c/ after the [+voice] feature is not discussed in this study except when
morphological conditions are active. A discussion on the syllable-initial /s/ or /c/ after
the [+voice] feature would seem to be premature because there are not enough studies
about phonetic constraints on voicing and tensing in fricatives and affricates to set up
phonetically motivated constraints in a given context (see Ohala 1983:194-01 for
phonetic constraints on voicing in fricatives and stops).

As shown in (33), (34), and (3 5), occurrence of ObsT after obstruents and non-
occurrence of ObsT afier nasals or the lateral [l] are explained with the same constraints
and their rankings. In addition, I have shown that contrast and neutralization of features
in a given context result from interaction between constraints sensitive to phonetic
implementational conditions, faithfulness constraints, and minimization of articulatory
effort. One remaining issue relevant to ObsT is to account for ObsT after nasals in the
verbal morphology and after the lateral [l] in some Sino-Korean nouns. The first part of

chapter 4 investigates ObsT after nasals or the lateral /l/ with morphological conditions.

184

3. 3 Discussion

As mentioned in Rhee (1997), there are two types of approaches to POT: the
geminate analyses by Han (1992) and Lee (1994) and the non-geminate analyses by Cho
and Inkelas (1994), Kang (1992), Kim-Renaud (1974), and others. The main assumption
in the geminate analyses is that tense obstruents have two timing slots, as shown in (42a).
Lax obstruents have one timing slot, as shown in (42b). POT is represented as insertion

of [CG] (constricted glottis) when two timing slots are associated as shown in (42c).

(42) a. Tense obstruents b. Lax obstruents c. POT
X X X X X
V l \/
[CG] 0 o
|
[CG]

According to Han’s (1992) geminate account of POT, there is a geminate rule that
associates two timing slots, as in (43b). After applying the geminate rule to input, [CG]
is inserted and lax consonants are tensed, as in (43c). A delinking rule deletes the line

associated with the first obstruent, as shown in (43d).

(43) a. Input b. Geminate c. POT d. Delinking
/p/ /k/. /p/ /k/. /p/ /k/. /p/ /k/.
X X —>X X —> X X -—> X X
| | \ / \ / 12‘ |
[-son][-son] [-son] [CG] [CG]

son: sonorant

185

However, Cho and Inkelas (1994) point out that deletion of [CG] in a geminate structure,
as in (43d), violates geminate inalterability (Hayes 1986), which prevents a rule from
being applied to only one part of a geminate.

In this study, geminate inalterability problems do not occur because tense and
aspirate obstruents are associated with one timing slot. Relative long closure duration is
one of the properties of tense and aspirate obstruents.

In addition, the delinking rule (43d) results in shortening of oral—closure duration
and deletion of [CG] in C1. The shortening of oral-closure duration occurs due to the
delinking of the line associated with the first obstruent. This means that there are two
types of tense obstruents in the output: one associated with two timing slots as in (42a)
and the other associated with one timing slot as indicated in (43d). However, there is no
conclusive evidence that there are two types of tense obstruents distinguished by different
timing slots (Oh 1998).

In the gesture-based account of POT, characteristics of tense gestures, such as
minimal glottal opening in C1C2, which is closely associated with [CG], and relative long
oral-closure duration between two vowels, motivate ObsT. These gestures promote
tensification of lax C25. Therefore, the minimal glottal opening and long oral-closure
duration in C1C2 should not be deleted, as implied by (43 d).

Based on Steriade (1991b), Lee’s (1994) geminate analysis represents lax, tense,

and aspirate obstruents, as in (44). A0 stands for zero aperture and represents closure.

Am stands for maximal aperture and represents approximant release. In addition, Af

stands for aperture-cum-friction and represents fricative release.

186

(44)

a. Lax O b. Tense O c. Aspirated 0

Root Tier ° ° °
I /\ /\
X- tier X X X X X
/\ V V
A0 Am A0 Am AO Am
/\ /\
L SL L SL
| l
[SG] [CG]

SG: spread glottis

d. Unreleased O

a—x—°

Based on the above representations, Lee (1994) explains POT as follows:

(45)

a circle. The delinked Am is recovered by an extra timing slot X in (45c). The inserted
timing slot X is incorporated into the following syllable-initial position, resulting in an

onset with two timing slots as in (45d). After incorporating the inserted timing slot X

a. Input b. Delinking of Am
/p/ /k/ /p/ /k/
Xle [X Xle [X
/\ /\ | /\
AOAm AoAm A0 AoAm

d. Regrouping of X e. Insertion of [CG]

/p/ /p/ /k/
X]. X]. [X X
| V l V
A0 AOAm A0 AoAm
l
[CG]

(45b) shows that Am is delinked from the lax stop timing slot /p/, as indicated with

187

c. Insertion of X

/p/ /k/
X]. Q) [X
| /\

A0 Am ADA-m

into the following syllable, [CG] is inserted because of the two timing slots in syllable-
initial position in (45c).

As pointed out by Rhee (1997), Lee’s (1994) analysis of POT fails to provide
phonetic or phonological reasons for the following: the inserted timing slot X,
incorporation of the inserted X into the following syllable position, and the insertion of
[CG] instead of [SG] (spread glottis). If both aspirate and tense consonants are
associated with two timing slots, as shown in (44), there is no reason why [SG] could not
be inserted in (45e).

In addition, Lee’s (1994) analysis of POT needs to be modified when C15 are

fricatives.
(46) A. /s/ /s’/ /h/
Root Tier ° ° °
1 | l
X- tier X X X
| | |
Ar Ar Am
/\ /\
L SL L SL
l |
[CO] [SG]

188

B. /koskan/ ‘barn’

a. Input b. Delinking of Am c. Insertion of X

/s/ /k/ /s/ /k/ /t/ /k/
X]. [X X]. [x X]. [x
l /\ I /\ I | /\
A. AoAm AOO AoAm A0 A... AA...
(1. Regrouping of X e. Insertion of [CG] *[kotkan]

/t/ / . /t/ /k/

X]. (18:) X]. [X

I V | v

A0 AOAm A0 AOAm

As shown in (46Ba), there is no Am to be delinked in fricatives. Af in syllable-
final position becomes A0, as shown in (46Bb). This is supported by data such as

[kot1.k’an] from /ko_skan/ ‘barn’ and [kat1.k’o] from /kas_’k_o/ ‘went and’. As a

consequence, there is no delinked Am to be recovered by an extra-timing slot X, as shown
in (46Bc). Thus there is no extra timing slot X to be incorporated into the following
syllable-initial position, as shown in (46Bd), and to trigger insertion of [CG], as shown in
(46Be). Therefore, obstruents preceded by fricatives should not participate in POT

according to Lee (1994). However, obstruents preceded by fricatives participate in POT,
as shown in [kot1k’an] from /koskan/ ‘barn’ and [kat1k’o] from /kas’ko/ ‘went and’.
Therefore Lee’s (1994) analysis of POT needs to be modified further.

As shown in (32), in the gesture-based account of POT, ObsT after fricatives is

treated the same way as ObsT after stops or affricates is treated. Therefore, there is no

reason to treat ObsT after fricatives differently from ObsT after stops or affricates.

189

In the non-geminate analyses, tense and aspirate obstruents are represented with
one timing slot. Long closure duration is an inherent property of tense and aspirate

obstruents. In the non-geminates analyses, POT is generally represented as follows (47).

(47) POT
a. [-sonorant] —>[+tense] / [-sonorant] (Ahn 1985)
b. [-sonorant] ——>[+constrict glottis]/ ([-sonorant] )5 (Kang 1992)

8 : An accentual phrase

Compared to the geminate analysis of POT, the non-geminates analyses of POT
do not face geminate-inalterability problems. However, regardless of geminate or non-
geminate analyses of POT, Han (1992), Lee (1994), and others using rule-based
approaches do not explain why the inserted feature is [CG], instead of [SG] or some other
feature(s).

The gesture-based account of POT in this study provides phonetic reasons for
tensification, rather than aspiration, of lax 025 in the sequence of V10102V2. As
discussed in section 3.2.3, tense properties, such as long oral-closure duration, minimal
glottal opening at the oral-closure release of 02, and heightened air pressure behind the
oral closure, are abundant in V10102V2.

When tense gestures are abundant between two vowels in V10102V2, having
[open] CD at the glottis at the oral-closure release of lax 025, as shown in (17), would
require gestural effort from speakers. However, speakers tend to use gestures distributed
in a given context in order to minimize articulatory effort. This principle of minimization
of articulatory effort plays a role to a degree, in the tensification, rather than aspiration, of

lax 025, as discussed in (29) and (30). Therefore, interaction of constraints sensitive to

190

phonetic implementational conditions in a given context, a constraint sensitive to
articulatory effort, and a faithfulness constraint are responsible for tensification of lax
025 in the sequence of VOlOzV, as illustrated in (30).

Many Korean linguists either directly or indirectly accept the fact that there is
close relationship between POT and CN (Park 1990, Sohn 1987, Kim-Renaud 1974, Lee
1994, Rhee 1997, and Cho 1990). However, few studies show the close relationship
between POT and CN in a systematic and unified way.

Kim-Renaud (19742130) states that phonological descriptions such as (47a) and
(47b) obscure the fact that “‘fortition”(0bsT in this study) actually occurs always and
only when a lenis obstruent (a lax obstruent in this study) is preceded by an unreleased
stop’ even though (47a) and (47b) are descriptively correct. Kim-Renaud describes POT
in terms of the [-release] feature, as shown in (48a). Therefore, [+tense] is inserted when

an obstruent is preceded by an unreleased stop.

(48) a. [+release, -sonorant]—> [+tense]/[-release]_ (Kim-Renaud 1974: 131) - POT

b. [-sonorant] —> [+tense]/ [+nasal] & _ (& : verb stem boundary) - PNT
(Kim-Renaud 1974: 175)

c. [-sonorant, +cororanl] —> [+tense] /[lateral] l __

(! Chinese loan word boundary) (Kim-Renaud 1974: 173) - PLT

Kim-Renaud (1974)’s relating of CN and POT is correct. However, ObsT occurs
not only after an unreleased stop, but also after a nasal in the verbal morphology, or after
the lateral /l/ in Sino-Korean nouns, as shown in (48). However, phonological
descriptions such as (48b) and (48c) seem to be as obscure as (47a) and (47b) are. (48b)

and (48c) do not show any mechanism behind them as (48a) does.

191

 

There are three questions raised related to (48). First, if the [-release] feature of
obstruents is closely related to ObsT as shown in (48a), why are [nasal] and [lateral]
related to ObsT in (48b) and (48c)? Second, why does morphological information play a
role in (48b) and (48c) but not in (48a)? Third, why does C2 have to be [-sonorant,
+coronal] in Sino-Korean nouns if C1 is the lateral /l/?

Section 3.2.2 and 3.2.4 in this study provide answers to these questions. I have
shown that CDs at the SL and Oral levels are closely related to occurrence or non-
occurrence of ObsT and morphological or phonological conditions of ObsT, as shown in

(49).

(49) Favorable conditions for tensification of lax 025

 

 

 

 

a. -91---- L92 ..... b. -91 ...... . 9.2 ....... c. -91---.:-<.32 ......
SL closed : closed 3 open : closed 3 open : closed
Oral closed [closed closed [closed open 1 closed

A 3 B: A is more favorable than B for tensification of lax 025.

As shown in (49), the narrower the SL and Oral levels are, the more C2 is likely be
tensed. This is because gestural and aerodynamic properties of tense are distributed more
in (49a) than in (49c), as discussed in 3.2.3 and 3.2.4.

Gestural characteristics of obstruents in syllable-final position influence the
following consonants to different degrees. Unreleased obstruents in syllable-final
position most strongly influence lax obstruents that follow due to the [closed] CD at the
SL and Oral levels. Therefore, POT (48a) occurs obligatorily across morphological
categories within an accentual phrase. The lateral /l/ in syllable-final position most

weakly influences the following lax obstruents due to the [open] CD at the SL and Oral

192

i -. av - _,W

levels. As a consequence, PLT appears with the most restricted morphological and
phonological conditions, as shown in.(48c). The morphological condition for PLT is
Sino-Korean nouns and the additional phonological condition is that C2 must be /t/, /s/ or
/c/. The influence of nasals on the following consonants falls between that of obstruents
and the lateral /1/. Therefore, PNT requires the morphological condition. PNT occurs
only after a verb stem within an accentual phrase, as shown in (48b). Therefore,
morphological conditions for ObsT are sensitive to phonological conditions of ObsT.
Interestingly, when phonological conditions for ObsT are weak, as shown in (49c),
morphological and additional phonological conditions are attached to ObsT, as shown in
(480).

So far, I have compared the analysis of this study to the analyses of POT within
rule-based approaches. It is now necessary to compare the analysis of this study to an
analysis within constraint-based approaches because this study adopts the basic concepts
of constraint-based approaches, as shown in chapter 1.

Rhee (1997) investigates POT within OT and constraints in Rhee are based on
Steriade’s (1993 and 1994) aperture geometry. Three types of Korean obstruents are

represented with one timing slot in Rhee’s study, as shown in (50):

(50) a. Lax C b. Aspirated C c. Tensed C
X X X
/ \ / \ / \
A0 Am A0 Am A0 Am
I |
[SG] [CG]

193

Rhee (1997) states that interaction of constraints and their hierarchy in (51) is
responsible for occurrence of tensification of lax 025 after an obstruent and non-

occurrence of tensification of lax-025 after nasals or the lateral /l/.

(51) a. CODA CONDITION b. SHARE—VOICE
(CODA COND) (SHA-[voc])
X] N C
| \/
A0 [voice]

0. Attract-[CG]
(ATT-[CGD
LO <—-—>[CG] (LO : Long duration of Obstruency, that is A0 A0 or A0 Af)
X] [X
| |
A0 A0

|
[CG]

d. PARSE- Am
(PAR- Am)

*<Am>

e. CODA COND, SHA-[voc] >> ATT-[CG] >> PAR- Am

CODA CONDITION (51a) prevents the maximal aperture (Am) from occurring in
syllable-final position and allows A0, which represents closure. SHARE—VOICE (51b)

demands the sharing of the [voice] feature in the sequence of NC. Attract-[CG] (51c)
suggests that when two Aos are next to each other, [CG] is attracted to the second
segment, as shown in (51c). PARSE-Am (51d) requires the release phase of an obstruent

to be preserved. (5 1e) shows the ranking of constraints (51a), (51b), (51c), and (51d).

(52) illustrates non-occurrence of ObsT after nasals (Rhee 1997:271).

194

(52) CODA COND, SHA-[voc] >> ATT-[CG] >> PAR- Am

/pantze/ ‘opposition’

/n/: GLO [critical], SL [open], Oral [closed], TT [closed]
/t/: GLO [mid], SL [closed], Oral [closed], TT [closed]

 

/p a _n 126/
AOAm A0 Am
/ \

inasal] [voice]

CODA COND iSHA
-[voc]

ATT~[CG]

PAR-Am

*INS—
[CG]

 

a. pa n1 t’ are
/ \ l

*l

 

[nasal] [voice] [CG]
b

. pa n1 t as
A0<Am> A0 Am

/ \
| nasal] [voice]

*1

 

3°C. pa n1 d a?
A0<Am> A0 Am

/ \ /

[nasal] [voice]

 

d. pa nL t’ a:
AoAm A0 Am
/ \

l
[nasal] [voice] [CG]

*1 II-‘

 

e. pa 11L (1 as

AOIAXrn A0 Am
/ \

 

nasal] [voice]

*1

 

 

 

 

 

 

Candidates (52a), (52b), and (52c) satisfy CODA COND because there is no Am in
syllable-final position. Candidates (52d) and (52e) violate CODA COND due to Am in

syllable-final position. Therefore, (52d) and (52e) are not the most optimal because

CODA COND is the highest ranked constraint.

Among (52a), (52b), and (52c), (52c) satisfies SHA-[voc]. (520) violates ATT-

[CG] because there is no insertion of [CG] when two A05 are next to each other.

195

 

However, violation of ATT-[CG] in (52c) is not important in selecting the most optimal

candidate because (52c) satisfies SHA-[voc], which is a higher ranked constraint than
ATT-[CG]. Therefore, (52c), which does not attract [CG] when two A05 are next to each

other, is selected as the most optimal candidate.

Rhee (1997)’s analysis of POT contributes to the understanding of POT and the
connection between CN and POT. However, there are some issues that still need to be
discussed. Constraints and their ranking suggested in (51) fall short in explaining the
non-occurrence of ObsT after the lateral /1/. Constraints and their ranking in (51) predict
the wrong output when a lax obstruent is preceded by the lateral /l/, as shown in (53).

Take as an example /salku/ ‘apricot’:

196

(53) CODA COND, SHA-[voc] >> ATT-[CG] >> PAR- Am
/salku/ ‘apricot’

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/k/: GLO [mid], SL [closed], Oral [closed],'TB [closed]

 

 

 

 

 

 

 

 

 

 

 

/sa 1 k u/ CODA :SHA ATT- PAR-Am *INS-
Am A0 Am COND -[VOC] [CG] [CG]
1
[voice]
> a. s a l1 . k’ u *l *
A0 A0 Am
I |
[voice] [CG]
b. sa 11 . k u *1 1’
Ao AoAm
|
[voice]
3?” c. s a l1 . g u 1" 1‘ *
A0 A0 Am
\ /
[voice]
(1. s a 1‘1 . k’ u *l * *
Am An Am
I l
bloice] [CG]

 

 

* [L] indicates released consonants.

SHA-[voc] is satisfied vacuously in (53a) because SHA-[voc] demands that the
[voice] feature be shared in the NC sequence, rather than the LC sequence. Candidate
(53a) violates PAR-Am because Am for /l/ is not maintained. According to constraints and
their hierarchy in (51), (53a>) is the most optimal candidate but (53c) is the actual
output, even though it violates SHA-[voc].

Regardless of the ranking of constraints in (51), the actual output (53c) is not

selected as the most optimal candidate. However, in the gesture-based account of POT in

197

this study, constraints and their hierarchy in (34), which are suggested for tensification of
lax 025 after an obstruent, correctly predict non-occurrence of ObsT in lax 025 after
nasals or the lateral /l/, as shown in (3 5A) and (3 5B).

One may modify SHA-[voc] so that the [voice] feature is shared by NC and LC
sequences. However, modification of SHA-[voc] to include a nasal and the lateral /l/
poses problems in the verbal morphology, as will be discussed in chapter 4.

In the verbal morphology, an obstruent preceded by either an obstruent or a nasal

iS tensed: as in [11111, a] from /u§ verb stem '1' La inflectional suffix/ ‘tO smile, and [all-fa] from

/anv,,,b mm + ta inflectiona] sumx/ ‘to hug’. However, an obstruent preceded by the lateral /l/

is voiced, instead of tensed, as in [5a]. da] from /sal verb stem + ta inflectional suffix/ ‘to live’.

Therefore, if SHA-[voc] is modified to include nasals and the lateral /l/, they should

pattern together with respect to ObsT in the verbal morphology. However, nasals and the
lateral /1/ exhibit different patterns with respect to ObsT, as shown in [ant’ a] from /anvc,b
stem + ta inflectional sul’lix/ ‘IO hug, and [58LQ3] from / 531 verb stem '1' la inllectional suffix/ ‘IO llVC’.

Therefore, modification of SHA-[voc] to include NC and LC sequences is unable to

resolve the problem in (53).

198

 

3.4 Summary

The gesture-based account of POT in this study offers several advantages that
previous studies on POT (Han 1992 and 1997, Lee 1994, Kim—Renaudl974, and Rhee
1997) do not offer. The following generalizations and predictions are made through the
gesture-based account of POT:

First, the necessary condition for ObsT is that C2 must not have the [open] CD at
the SL level. When C2 has the [open] CD at the SL level, lax obstruents in C2 position
do not participate in ObsT, as shown in (10). This generalization is possible through
constriction degree hierarchies predicted from gestures of active articulators for segments
at a given time, as proposed by Browman and Goldstein (1989 and 1990a and 1992).

Second, the systematic relationship between C1 and C2 with respect to ObsT is
illustrated in terms of CDs at the SL and Oral Levels. When CDs at the SL and Oral
levels are [closed] for C2, the narrower the opening of C1 is, the more C2 is likely to be
tensed, as shown in (20) and (21). Therefore, the morphological condition in the verbal
morphology and the additional phonological condition in Sino-Korean nouns emerge
when gestural conditions of ObsT are relatively weak, as summarized in (24).

Third, the gesture-based account of POT illustrates the close relationship between
phonetic implementational conditions for a feature and neutralization of the feature in a
given context. As shown in 3.2.3, when C1 is an obstruent and C2 is a lax obstruent in
the sequence of V1C1C2V2, relatively long oral-closure duration and minimal glottal
opening, which are coupled by heightened oral pressure and tension, promote

tensification of the lax-C2 obstruent. This is because long oral-closure duration, minimal

199

glottal opening, heightened oral pressure, and heightened vocal tract tension between two
vowels in V1C1C2V2 are properties of tense obstruents, rather than properties of aspirate
or lax obstruents.

In addition, the gesture-based account of POT in this study connects an
unreleased stop in syllable-final position to POT as Kim-Renaud (1974), Lee (1996) and
Rhee (1997) also do. Unreleased stops in syllable-final position increase tense gestures
in V1C1C2V2. This implies that when C1 and C2 are obstruents, candidates with an
unreleased stop in syllable-final position are preferred over candidates without an

unreleased stop. This implication is illustrated in (54).

(54) *[mid]GL0/ V101_V2» Max-10,,g >> *CCD/_{#, C} >>
IDENT-IO[G] >> Mini Eff=0

/koskan/ ‘barn’

/s/: GLO [mid], SL [critical], Oral [critical], TT [critical]
/k/ 2 GLO [mid], SL [closed], Oral [closed], TT [closed]

 

 

 

 

 

 

 

 

lkogan/ *[midlor-o/ Max- *CCD/_{#, C} IDENT- Mini
V101 V2 1086% IGIG] Eff=0
a. koskan *1 i . f2if‘.“(GL0). i " V ' ‘ ‘ "
b. kotkan *! 5;..'?*(GLO). 1 , *(TTCD) 1.
C. kot1._lgan *l » . ?-__,**_(S.L) ' *(TTCD); . . :1:
1 " 2' '3*(GLO)~ l
d. kosk’an *(GLO) *(GL01 *
e. kotlg’an *(GLO) *(TTCD) ."9
*!(SL) g  ;:
7f. kot’.1_<_’an *(SL) *(TTCD) Vi ‘3
*(GLO). ” , ‘

 

 

 

 

 

 

 

Candidates from (54a), (54b), and (54c) are eliminated as possibilities to be the

most optimal candidate because these candidates violate the highest constraint

200

*[mid]GL0/ V101_V2. Candidates from (54d), (54e), and (54f) are tied in satisfying

*[mid]GL0/ V101__V2 and Max-IOseg. Therefore, *CCD/_{#, C} plays a role in

selecting the most optimal candidate. (54f) violates *CCD/_{#, C} once while (54d) and
(54e) violate *CCD/_{#, C} twice. Thus, (541) is the most optimal candidate. Therefore,
*CCD/_{#, C}, which is responsible for CN, plays an active role in selecting the most

optimal candidate for ObsT.

201

Chapter 4 Obstruent patterns after sonorants and lh/

4.1 Introduction

In chapter 3, it was suggested that patterns of ObsT in Korean are closely related
to CD5 at the SL and Oral levels for C1 and C2 in V1C1C2V2. Because gestures and
aerodynamic properties of tense, such as minimal glottal opening, and long oral-closure
duration, heightened pharyngeal pressure, and vocal tract tension, are sufficient in
V10102V2, lax 025 are tensed obligatorily within an accentual phrase. In other words,
when CDs of C15 are [closed] CDs at the SL and Oral levels in V10102V2, lax 025 are
tensed obligatorily, as shown in section 2.2.2 (1 1).

This chapter investigates ObsT after nasals or the lateral /t/(Post-lateral
Tensification (PLT)), and coalescence of /h/ and a lax stop (Aspiration). Note that nasals,
the lateral /1/ and /h/ have the [open] CD at the SL level, as shown in (1b), (1c) and (1d).
Chapter4 adds more examples that illustrate the close relationship between phonetic

realization and CD in the vocal tact.

202

(1) CDsofCt in V1C102V2

a. Obstruents b. Nasals c. Lateral /l/ d. /h/
GLO [closed], [mid], [open] [critical] [critical] [open]
SL [closed]/[critical] [open] [open] [open]
Nasal [closed] [open] [closed] [open]
Oral [closed]/[critical] [closed] [open] [open]
Lips/ [closed] [closed] [open] [open]
TB/ [closed] [closed] [open] [Open]
TT [closed]/[critical] [closed] [closed] [Open]
Phonetic ObsT ObsT ObsT Aspiration
realization
of lax 02
Conditions No conditions Yes Yes
for ObsT After a verb stem Sino—Korean nouns

02 must be /t, s, c/

(1) shows a close relationship between CD5 of C 1 and (non)-occurrence of ObsT.
When Cl does not have any [closed] or [critical] CD at the vocal tract as in /h/ (shaded
area in 1d), Cl's influence to the following 02 is the weakest with respect to ObsT. This
is because there is no long oral-closure duration and/or minimal glottal opening.
Therefore, in general, aspiration, instead of tensification, of lax O25 occurs when C 1 is /h/
in Korean.

The first goal of this chapter is to show that there are phonetic motivations behind
the conditions for PNT and PLT. The conditions for ObsT are understood best when the
articulatory gestures of nasals or the lateral /l/, and principles of co-articulation between
two adjacent gestures are taken into account (Browman and Goldstein 1990a and 1990b,
Byrd and Tan 1996, J. Jun 1995, and Cho 1998a and 1998b). The second goal of this
chapter is to demonstrate that aspiration of lax stops (Aspiration) to a degree results from

the optimal satisfaction of constraints suggested for ObsT. The third goal of this chapter

is to show that Post-Nasal Tensification (PNT), Post-Lateral Tensification (PLT), and
Aspiration result from interaction of the same constraints sensitive to phonetic
implementational conditions in a given context, faithfulness constraints, and
minimization of articulatory effort as POT does. Therefore, PNT, PLT, and Aspiration,
which are treated as unrelated phenomena, are treated as related phenomena and are
explained in a unified way in this chapter.

Contrary to POT, there are few studies that discuss ObsT after nasals (PNT lb), or
after the lateral /l/ (PLT 1c) in a unified way, as in this study. In addition, POT, PNT,
PLT, and Aspiration of lax C25 are treated with separate rules without any connection to
each other in previous studies (Kim-Renaud 1974, Ahn 1985, and Park 1990).

In section 4.3, I show that patterns of [h] are predictable and regular like patterns
of other obstruents in Korean such as stops, fricatives and affricates. The regular patterns
of /h/ are also motivated by phonetic implementational conditions in a given context.
The phonetic realization of /h/ and lax obstruent sequences are considered to be irregular
and exceptions in previous Korean studies (Ahn 1985, Cho 1990, Kim-Renaud 1974,

Park 1990 and others).

204

4.2 Obstruent Tensification after sonorants

4.2.1 Introduction

Syllable-initial lax stops after unreleased stops are never voiced across all lexical
categories, as shown in (2). Instead tensification of lax 025 occurs due to the distribution
of tense gestures in V10102V2, as discussed in chapter 3. Note that data in this chapter
is mainly from verbs or Sino-Korean nouns for subsequent discussions on PNT in the

verbal morphology and PLT in Sino-Korean nouns.

(2) Verb Stem (VS) Inflectional suffix(IF S)

/US/ + [kl/gerundial suffix —) [ut1.k’i] ‘laughing’
/US/ + / ta] declarative marker —) [1111 -I1 a] ‘laugh,

hnek/ + / ta] declarative marker _) [1118131 £3] ‘eat,

/mek/ + /§o/com,m,ndmyIx —-> [mak1; 0].. ‘do eat’
Sino-Korean nouns

/kukki/ —> [kuk1. k’i] ‘flag’

/kukce/ —> [kuk._c’ 8] ‘ international ’

In general, lax-02 stops are voiced between two vowels (3a), after the lateral /1/

(3b) or nasals (3c) across lexical categories except for certain cases such as in (4) and (5).

205

(3) a. Voicing of syllable-initial lax stops after a vowel
Verb Stem Inflectional suffix

/mek/ + /9/command suffix —) [mega] ‘dO eat’

Sino-Korean nouns

/kaku/ 'furniture' —> [kagu]
/sopi/ ‘consume‘ -—> [sgbi]

b. Voicing of syllable-initial lax stops after the lateral /l/

Verb Stem Inflectional suffix

/mu1/ + /kl/ noun derivational suffix —) [mulgi] ‘bifing’
/ 1101/ '1' / Ia/ declarative suffix ‘_) [nolgi] ‘play,

Sino-Korean nouns

/ma&i/ —> [maIgi] ‘the end period’
/pulpok/ —> [puLbok] ‘disobedience’

c. Voicing of stops between a nasal and a vowel.

Verb Stern Inflectional suffix

/e;1ki/ + /nta/ * —> [egginda] ‘harden’
/o/ + /gt_a/ —> [onda] ‘come’

*/nta/ is used after an action verb stem ending in a vowel and used for writing
essays and reports as simple or plain style statements.

Sino-Korean nouns

/tanki/ 'short period' ——) [tag g i]

/kamtok/ 'inspector' —> [kar_n_.dok]

There are certain cases where lax stops are not voiced after nasals or the lateral /l/

in the verbal morphology and Sino-Korean nouns. Instead, obstruents after verb-stem-

206

ending nasals or /t, s, c/ after the lateral /l/ in Sino-Korean nouns are tensed, as shown in
(4) and (5).
The difference between (4a) and (4b) is that nasals in (4a) end verb stems while

nasals in (4b) do not end verb stems.

(4) a. ObsT after verb-stem-ending nasals

Verb Stern Inflectional suffix
/kam/ + /§o/c(,mandsun-Ix —> [kamvss’o] ‘do wind’
/afl/ + ki/gerundial suffix ' _) [aflVS-K,i] ‘hugging,
/an/ '1' / ta] declarative marker ‘9 [aflVS-fa] ‘hug,

b. Voicing after nasals
Verb Stem Inflectional suffix
/o/ + Lina! —-> [ovsnda] ‘come’

/O_I:Lkl/ + 43/ declarative marker “'3 [ell-givs-da] ‘harden’

A comparison of (4a) and (4b) reveals that syllable-initial lax-02 stops are tensed
after nasals that end verb stems. Otherwise, syllable-initial lax-02 stops are voiced in the
verbal morphology.

In addition to (4), /t, s, c/ after the lateral /l/ are tensed in Sino-Korean nouns, as
shown in (5a). However, /k, p/ after the lateral /l/ are not tensed, instead they are voiced

in Sino—Korean nouns.

(5) a. Tensification of /t, s, c/ after the lateral /l/ in Sino-Korean nouns

/kal_cir)/ —) [kal_.c_’irj] ‘thirsty’
/pal;al/ —> [pa1.t’al] ‘development’
/il_sa I]/ ——> [iLs’arJ] ‘everyday’

207

b. Voicing afier the lateral /l/ or nasals in Sino—Korean nouns

/mal_ki/ —-> [ma]_.gi] ‘the ending period’
/pulp_ok/ —> [pulbok] ‘disobedience’
/kamtok/ ——> [kani dok] ‘inspector’

4. 2. 2 The gesture-based account of tensification after sonorants

Constraints suggested for POT in chapter 3 play a role to a degree in selecting

output forms in (2). (6) evaluates candidates from /mek/vc,b stcm + /ta/dnc1mtive marker ‘to

eat’ against constraints suggested in chapter 3 (43). Note that /mek+ta/ has the sequence

V10102V2.

(6) *COMPLEX, *[mid]GLO/ V101__V2 >> *[mid]GLO/ V1N__V2 >> *[mid]GLo/
V1L_V2 >> Max-108.3g >> *CCD/_{#, C} >> IDENT-IO[G] >> Mini Eff=0

/mek/VS + /ta/11:s ’tO eat’

/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]
/t/: GLO [mid], SL [critical], Oral [critical]J T [critical

 

/mek/+/ta/ *Cowg *[mid]GLO/ Max- *CCD/_{#, C} IDENT- Mini

 

iVlOl v2 10,,g , IO[G] Eff=0

 

b. met‘ta ? *1 iii:-Figli;(SL>:i/*(GL91

 

0- met-fa = ‘*!*<GL0> (SL) *(GLO):

 

d. mekt'a ; *l*(GLO) (SL) *(or-o) {gggérji214:]:

 

e. mek1.tha *(SL) ** "* l
i (GLO]

 

 

 

 

 

 

 

 

err. mék1.£a *(SL) ** *
= (GLO)

 

208

Candidates (6a) and (6b) violate tlie highest-ranked constraint *[mid]GL0/
V101_V2, which prevents GLO [mid] from occurring in a position where gestures
related to lax obstruents are lacking or absent.’ Therefore, (6a) and (6b) are eliminated by
*[mid]GL0/ V101_V2.

Candidates (6c), (6d), (6e), and (6f) violate *CCD/_{#, C}, which does not allow
contrastive CD5 in syllable-final position. However, violations in (6e) and (61) are less
critical than violations in (6c) and (6d) because (6e) and (6f) violate *CCD/_{#, C} only
once. (6e) and (6f) are tied in violations of [DENT-IO[G] because the input glottal

gestures for /k/ and /t/ are not preserved in [mak1 .t’ a] (6e) and [mak1.tha] (6f).

Therefore, the next lower-ranked Constraint Mini Eff=0 plays a role in selecting
the most optimal candidate. (6e) violates Mini Eff=0 twice because CDs at the glottis for
C1 and C2 are [critical] and [open] respectively. (61) violates Mini Eff=0 once because
CD5 at the glottis for C1 and C2 are [critical] and [mid] respectively. Therefore, (6f)
satisfies constraints in (6) best, and is the most optimal candidate and the actual output.

However, in certain cases actual output forms are not selected as the most optimal

candidates according to constraints in (6), as shown in (7) and (8). (7A) and (7B)

evaluate candidates from /0/Vs + Ana/{pg ‘to oom’ and lan/vs+/ta/1ps ‘hug’ respectively.
Note that in /o/vg + /n_ta/1ps ‘come’, the nasal /n/ does not end the verb stem, as shown in

data (4b), while in /an/Vs + /ta/11:g ‘hug’, the nasal /n/ ends the verb stem, as shown in data

(4a).

209

(7) *COMPLEX. *[midkn-o/ V1N_V2 >> Max-Ioscg >> *CCD/_{#, C} >>
Imam-IO[G] » Mini Eff=0

A. /o/vs + /nta/n:s‘come’

/n/ 2 GLO [critical], SL [open], Oral [closed], TT [closed]

 

/t/: GLO rmicflLSIchlosed], Oral [closed], TT[closed]
0vs+nta1ps *COMPE *[mid]GL0/ Max- *CCD/ IDENT Mini
E V1N_V2 Ioscg _{#,C} -IO[G] Eff=0

 

 

 

 

 

 

 

 

 

 

 

a. on. ta 5 *1 F?"$3,153" 7:5:217.5375}. f7 f7ffi‘fj’fff'j

b. on.t’a *[GLOI *!

c. onf'a 1 *[GLOI **l
@‘d. on.__da *[GLOI

 

 

B. /an/vs+/ta/n:s ‘hug’

/n/: GLO [critical], SL [open], Oral [closed], TT [closed]
/t/ 2 GLO [mid], SL [closed], Oral [closed], TT [closed]

 

aflvsflarrs 1COMP§ *[mid]GL0/ Max- *CCD/ IDENT Mini
‘ V1N_V2 Ioscg _{#, C} -10[G] Err=o

 

 

 

 

 

 

 

 

 

 

 

a. antze *! ‘5":7~77'37_f;.7 if.,1”if;{Iglffiifi’ff1:. 731(13)::lift”if: 1:575?.1'gliflfii
G" b. t 51 *

c. ant a: * “‘1
> (1. da '

 

Candidates (7Aa) and (78a) violate *[mid]GLo/V1N_V2 because [t] is in the

position where gestures to lax obstruents are not abundant. As mentioned in Rhee (1997)
and Kim-Renaud (1974), in VlN 01V2, heightened pharyngeal pressure and tension may
be relatively low due to the [open] CD at the SL level for nasals. However, there is a
long oral-closure duration in V1N01V2, which is one of the important properties of tense
obstruents, as shown in Han (1997). According to Ohala (1983), longer oral-closure

duration is likely to create unfavorable phonetic implementational conditions for voicing.

210

Candidates in (7A) and (7B) except (7Aa) and (7Ba) are tied in [DENT-IO[G]
because they do not maintain the input glottal gestures. Therefore constraints lower than
IDENT-IO[G] play a role in selecting the most optimal candidate.

(7Ac) and (7Bc) violate Mini Eff=0 to the greatest degree because glottal gestures
move from the [critical] CD for sonorants to the [open] CD for aspirated obstruents, thus
violating Mini Eff=0 twice. Candidates (7Ab), and (7Bb) violate Mini Eff=0 once
because CDs of glottal gestures for C1 and 02 are one degree narrower with respect to
each other. Candidates (7Ad) and (7Bd) satisfy Mini Eff=0 most optimally because CDs
of glottal gestures are identical, being [critical]. Therefore, according to constraints and
their ranking in (6), (7Ad<?°) and (7Bd>) are the most optimal candidates.

However in (7B), the actual output is (7BbG'), not the most optimal candidate
(7Bd>) selected by (6). Candidate evaluation from /paltal/ in Sino—Korean nouns reaches
the same conclusion. The most optimal candidate (7Bd‘>) selected by constraints in (6)

is not the actual output, as shown in (8).

(8) *COMPLEX, *[mid]GL0/ V1L_V2 >> Max-IOscg >> *CCD/_{#, C} >>
IDENT-IO[G] >> Mini Eff=0

/palta1/smoK‘development’ (SinoKI Sino-Korean nouns)

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]

 

/t/ 2 GLO [mid], SL [closed], Oral [closed], TT [closed]

 

 

 

 

 

/palt31/Sino-K *COMPE *[mid]GLo/ Max- *CCD/_{#.C} IDENT- Mini
' V1L_V2 10,,8 IO[G] Eff=0
a. paltal i *1 . :, f f *
we b. palt’al i * *
c. palf‘al j a: #11:!
> d. palgal ' :1:

 

 

 

 

 

 

 

211

(8a) is not the most optimal candidate because it violates *[mid]GLo/ V1L_V2,

which is ranked relatively high in the constraint hierarchy. Candidates (8b), (8c) and (8d)
are tied in violation of IDENT-IO[G]. Therefore, Mini Eff=0 selects the most optimal
candidate. Candidate (8d >) is the most optimal candidate according to (6) because [1]
and [d] have identical [critical] CD5 at the glottis, and therefore no displacement of the
glottis. Recall that the CD value of the glottis for [+voice] feature is [critical] (Browman
and Goldstein 1989: 225). The most optimal candidate (8d‘>) selected by constraints in
(6) is not the actual output (8b?)

(9) compares the most optimal candidates selected by (6) and actual outputs,

using (88b?) and (8Bd>).

(9) /palta1/3inoK‘development’ (Sinoxi Sino-Korean nouns)

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/t/ 2 GLO [mid], SL [closed], Oral flosed], TTfilosed]

 

 

 

 

 

 

/Palta1/sin01( *COMP § *[mid]GLO/ V1N_V2 Max- *{gcg 131111;]- [$12,
: IO~ _ , =
4. >23
a. palt’al j at: :1:
(88b?) 2 ‘
‘3’ b palgal at:
(SBd>) :

 

 

 

 

 

The lax stop /t/ is voiced after the lateral /l/, satisfying Mini Eff=0 in (9b), which
is the most optimal candidate according (6). However, /t. s, c/ after the lateral /l/ as in
(5a) are tensed in Sino-Korean nouns, as shown in candidate (9a). In addition, lax stops
except /t/ as in (5b), are voiced after the lateral /l/, as shown in (9b).

Therefore, constraints and their rankings responsible for ObsT should be sensitive

to morphological information. I suggest (10a) and (10b), which prevent voicing of lax-

212

02 stops after verb-stem-ending nasals or after the lateral /1/ when 02 is /t, s, c/ in Sino-

Korean nouns respectively.

(10) a. *[critical]GL0/ VIN] vs _V2: (VS: verb stem)
GLO [critical], namely voicing, is not allowed after nasals that end verb stems.

b. *[critical]GL0/ V1L_V2 when 02 is /t, s, c/ 2 (Sino-Korean nouns)
GLO [critical], namely voicing, is not allowed after the lateral /1/ when 02 is /t, s,

c/ in Sino-Korean nouns. (hereafter *[d, z, d3151nok/ V1L_V2)

Before evaluating candidates against (10), because in OT constraints are supposed
to be universal, I will provide phonetic motivation for the morphological or additional
phonological conditions for PNT and PLT. Note that in Korean, lax obstruents are never
voiced after unreleased stops, as shown in (3), due to the abundant tense gestures between
the two vowels in V10102V2.

In addition, Ohala (1983: 199) suggests that ‘a voiceless one (stop) becoming
voiced depends in a major way on the duration of the stop closure, a shorter duration
allowing voicing to be maintained for all of, or a majority of, the stop duration’. Based
on (3) and Ohala’s suggestion, I propose the following harmonic scales for voicing of lax

stops in V1C102V2.

(1 1) Favorable conditions for voicing of lax-02 stops

 

 

 

a- -Lateral---i,9.2 ...... b- fies-2115-192 ...... C- .91 ..... 3-9-2 .....
SL open 1 closed 3 open J closed 3 closed j closed
Oral open 1 closed closed 1 closed closed 1 closed

A 3 B: A is more favorable than B for Voicing of lax-O2 stops.

213

(11) illustrates that the more C] has [open] CDs in the vocal tract, the more lax-
02 is likely to be voiced in V1C102V2. Interestingly, the harmonic scales for voicing of
lax-02 stops in (10) are the reversed harmonic scales for tensification of lax-025 in
V1C102V2 (see chapter 3 (31)). Ohala (1983: 295) suggests that ‘the longer the stop
closure is, the greater is the likelihood that voicing will be extinguished.’

S. A. Jun (19952250) shows that because of the overlapping of consonant and
vowel gestures, word-medial lax voiceless stops in (3a) have the smallest and shortest
glottal opening that facilitates voicing of lax stops. In addition, Browman and Goldstein
(1990a: 370) describe voicing assimilation as ‘a reduction in the magnitude of the glottal
opening-and-closing gestures responsible for the voicelessness’.

Based on Ohala (1983), S. A. Jun (1995), and Browman and Goldstein (1990a), I
assume that due to [open] CD5 at the SL and Oral levels of the lateral /l/, gestures related
to voicing, such as the [critical] glottal opening and sufficient airflow through the
[critical] CD of the glottis, are distributed more in V1L02V2 (11a) than in V10102V2
(11c). Therefore, voicing of lax-02 stops after the lateral /l/ is more phonetically natural
than voicing of lax-02 stops after nasals or unreleased stops. In addition, due to the
[open] CD at the SL level and the [closed] CD at the Oral level for nasals, voicing of lax-
02 stops after nasals is less phonetically natural than voicing of lax-02 stops after the
lateral /l/. However, voicing of lax-02 stops after nasals is more phonetically natural than
for voicing of lax-02 stops after unreleased stops.

However, the phonetically grounded tendency of voicing can be interrupted in
languages because of two reasons: a faithfulness constraint and/or morphology-sensitive

constraints. If an input-output faithfulness constraint such as [DENT-IO[G] is ranked

214

relatively high in a language, the input glottal gestures are maintained in the output.
However, as seen in chapters 2 and 3, in Korean IDENT-IO[G] is ranked relatively low in
the constraint hierarchy. Therefore, I assume that in Korean, morphological conditions,
such as verb-stem-ending or Sino-Korean nouns, play a role in interrupting the voicing
tendencies of lax voiceless stops in a given context.

In addition, I suggest that the stronger phonetic conditions for a phonological
process are, the more conditions may be attached to interrupt the process. Therefore,
because the L-lax 02 sequence has the strongest phonetic conditions for voicing of lax
025, disallowing voicing of lax-02 stops after the lateral /1/ requires both morphological
and additional phonological conditions, as shown in (1%). Because the N-lax 02
sequence has weaker phonetic conditions for voicing of lax 025 than that of the L-lax 02
sequence, disallowing voicing of lax-02 stops after nasals only requires the
morphological condition, as shown in (10a). Therefore, there is a close relationship
between distribution of voicing gestures and morphological conditions.

(10a) and (10b) are ranked higher than IDENT-IO[G] because the output forms in
(4) and (5) do not preserved the input gestures. Violation of [DENT-IO[G] is taken for

granted in the output froms of (4) and (5). As there is no conclusive evidence that (10a)
and (10b) are ranked higher than *CCD/_{#, C} or Max-[0368, constraints (10a) and

(10b) are unranked with respect to *CCD/_{#, C} as in (12) until proven otherwise.

215

(12)

*COMPLEX, *[mid]GLo/ V1N_V2 >> Max-103,8 >> *CCD/_{#, C},

*[CfificalloLo/ Vlvas _V2, *Id, 2, d3ISinoK/V1L_V2>>
IDENT-IO[G] >> Mini Eff=0

A. /o/vs + /nta/1Fs ‘come’

/n/: GLO [critical], SL [open], Oral [closed], TT [closed]
'mid], SL [closed], Oral Elosed], TT [closed]

/t/: GLO

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ovs 111131135 *COMP] *[midloLo/ Max‘ *CCD/1 *[criticalloLo/ IDENT Mini
E V1N_V2 IOseg —{#’ C} E V1N]VS _V2 -IO[G] Eff=0
b. ogt’a 5 *[GLOI *1
c. onf’a *[GLOI **1
G’d. (giga *[GLO]
B. /an/Vg+/ta/ IFS ‘hug’
/n/: GLO [critical], SL [open], Oral [closed], TT [closed]
/t/: GLO [mid], SL [closed], Oral [closed], 11' [closed]
anvs+ia 11:5 1C0MP§ *[midloLo/ Max‘ *CCD/ *[critical]GL0/ IDENT Mini
E V1N_V2 IOseg —{#’ C} E V1N]VS _V2 'IOIG] Eff=0
21.29111 ' *g ‘ ' ' ., " éxixvifii-i‘;:~*:.=i::*g:*'.*é<3::I-i‘*~.:;*'.*i** ;~ a:
*3" b. agt’a * *
c. antha * **1
d. anda *1 ’1‘

 

 

 

 

 

 

 

(12Aa) is eliminated by *[mid]GLo/ V1N__V2, as GLO [mid] appears in the

position where phonetic implementational conditions promote voicing rather than lax

stops due to the [critical] CD of the glottis and the [open] CD at the SL Level for nasals.

The constraint *[critical]c,1,o/ VlN]vs _V2 is irrelevant to candidates in (12A) because

the vowel /0/ ends the verb stem, instead of the nasal /n/, as in /o/Vs + /nta/1ps ‘come’.

(12Ab), (12Ac), and (12Ad) violate IDENT-IO[G] once because the input [mid] glottal

CD for /t/ is not preserved in each output candidate. Therefore, the next lower-ranked

216

 

constraint Mini Eff=0 evaluates (12Ab), (12Ac), and (12Ad). Candidate (12Ad) satisfies
Mini Eff=0 most optimally because both [n] and [d] have [critical] CDs at the glottis.
Thus, there is no displacement of the glottis between [n] and [(1]. Therefore, candidate

(12Ad) is the most optimal candidate and the actual output form.

In the tableau (12B), the constraint *[critical]GL0/ V1N]Vs _V2 plays a role in

candidate evaluation because the nasal /n/ ends the verb stern, as in /an/vs+/ta/IFS ‘hug’.

Candidate (12Bd) is eliminated by *[critical]GL0/ V1N]vs _V2. Candidates (12Bb) and

(12Bc) are tied in violation of IDENT-IO[G]. Therefore, both candidates (12Bb) and
(12Bc) are passed on to the next-lower ranked constraint Mini Eff=0. Although both
candidates violate Mini Eff=0, (12Bb) is selected as the most optimal candidate because

in (12Bb), the displacement of the glottis between two consonants is less than that of

(12Bc).

Tableau (l3) evaluates /pal_tal/s,noK ‘development’ and /pulpok/SinoK

‘disobedience’. Note that /pal_t,al/smoK has /t/ after the lateral /l/ and /pulpok/SinoK has /p/

after the lateral /l/.

217

(13) *COMPLEX, *[mid]GL0/ V1N_V2 >> Max-103,,g >> *CCD/_{#, C},
*[critical]GL0/V1N]vs _V2, *[d, z, d3]g,noK/V1L_V2>>
DENT-IO[G] >> Mini Eff=0

A. /paltal/smoK ‘development’ (Sinoxz Sino-Korean nouns)

/l/ 2 GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/t/: GLO [mid], SL [closed], Oral mosed], TT [closed]

 

 

 

 

 

 

 

 

 

 

 

/paltal/ Sino—K *COMP i l.‘imidiGLO/ Max. *CCD/ j *[d, Z, d3]SinoK IDENT Mini
E V1L_V2 10,,g _{#, C} gNlL V2 -IO[G] Eff=0
apaltal 1 *1 ‘ ’ *f 1
‘7” b. palt’al i i at at:
c. palfal * **1
d. palgal E E *! ,3 3*; 2. ,-'jii.1‘v"f,il‘f:.;

 

 

B. /pulpok/smoK‘disobedience’ (ginoK: Sino-Korean nouns)

/l/: GLO [critical], SL [open], Oral [open], TB [open], TT [closed]
/p/: GLO [mid], SL [closed], Oral [closed], Lips [closed]

 

 

 

 

 

 

/pUIp0k/Sino-K *COMPE *[mid]GL0/ Max- *CCD/ [*[d, z, [13],,“0K IDENT Mini
§ V1L_V2 Ioscg _{#, C} §N1L_V2 -10[G] Eff=0
bpul. ’ok in n
c. prfihok j * **1
a" d. pul._b_ol( i ' 1k

 

 

 

 

 

 

 

The constraint *[d, z, d3]5inoK/ V1L_V2 is active in (13A) but inactive in (13B).

Therefore, candidate (13Ab) is the most optimal candidate because it does not have the

voiced [d] after the lateral [l] and articulatory effort is minimal. However output

candidates in (13B) vacuously satisfy *[d, z, d3]s,noK/ V1L_V2. Therefore, (13Bd) is

selected as the most optimal candidate because there is no displacement of the glottis
between [1] and [b]. Tableaux ( 12) and (13) show that constraints suggested in (10) and

constraints suggested for POT in chapter 3 (33) correctly predict actual output forms in

218

the verbal morphology and Sino-Korean nouns. Therefore, PNT in the verbal
morphology and PLT in the Sino—Korean nouns result from interaction of the same output
constraints suggested for POT. Thus, PNT, PLT, and POT are connected and related
together to the extent that they share the same constraints.

Up to this point, I have suggested constraints responsible for PNT in the verbal
morphology and PLT in Sino-Korean nouns in (10). In addition, I have also provided the
phonetic motivation related to distribution of voicing gestures and morphological and
additional phonological conditions in (10b).

Therefore, it is time to provide phonetic reasons why all lax 023 that follow nasals
are tensed in verbal morphology while only /t, s, c/ that follow the lateral /1/ are tensed in
Sino-Korean nouns. I suggest that lax 023 are tensed at the level where CDs of C1 and

C2 form [closed][closed] or [closed][critical] CD sequences, as shown in (14) and (15).

(14) N lax 02
SL [open] 3 q ‘ [closed/critical]
Oral [closed][[f [cloSed‘].
TB [closed]/13/ [closed] /k/
TT [closed] /n/ [closed/critical] /t, c, s/
Lips [closed] /m/ [closed] /p/

In N-lax 02 sequences, [closed][closed] or [closed][critical] CD sequences are
formed at the Oral level. Therefore, all lax 025 are tensed after nasals because lax 028
are specified with the [closed] CD at the Oral level for /p, t, k/, the [critical] CD at the
Oral level for /s/, and the [closed/critical] CD at the Oral level for /c/. In N-lax 02
sequences, due to the overlapping of nasals and lax 023, when the obstruents begin, there

is still a bit of air flowing out of the nose (Kager 1999:61). Therefore N-lax 02

219

sequences do not have ideal conditions for ObsT. Thus, ObsT occurs when morphology-
sensitive constraints such as (10a) are active, as shown in (12Bb).
In L-lax 02 sequences, [closed][closed] or [closed][critical] CD sequences are

formed only at the TT level (shaded area), as shown in (15a).

(15) a L lax 02 b. L lax 02

SL [open] [closed/critical] [open] [closed]

Oral [open] [closed] [open] [closed]
TB [Open] [closed] _ ‘n . [Open] [closed] /p/
TT [eloSed]f/l/ [closed/critiical]./t, c, s/ [closed] /1/ [open]
Lips [open] [open] [open] [closed] /k/

/t, s, c/ after the lateral /1/ form [closed][closed] or [closed][critical] CD sequences
at the TT level because /1/ and /t, s, c/ share the TT, as shown in (15a). However, /p, k/
after the lateral /1/ never form [closed][closed] or [closed][critical] CD sequences, as
shown in ( 15b). Therefore, /t, s, c/ after the lateral /1/ are tensed, resulting in [t’, s’, 0’]
respectively while /p, k/ after the lateral /l/ are voiced, resulting in [b] and [g], as shown
in [maLgi] from /mami/Sino-x ‘finishing period’ and [pul.bok]5ino-K from /pu1pok/
‘disobedience’.

In the sequence of /l/-/t, s, c/, there is a prolonged [closed] CD of the tongue tip as
shown in (15a). Distribution of tense gestures may be increased to a degree due to the
long closure duration of the tongue tip. However, in /1/-/p, kl, there are no consecutive
[closed] CDs of an oral articulator , as shown in (15b). When the oral articulator moves

from the TT to the TB, a leakage of air occurs. Therefore, only ObsT of /t, s, c/ afier the

lateral /1/ is allowed in Sino-Korean nouns.

220

4. 2. 3 Summary

Tableaux (6), (12), and (13) reveal that phonetically motivated constraints and

their hierarchies responsible for POT do not change according to lexical categories in

Korean. Only morphology-sensitive constraints such as *[critical]GL0/V1N]vs _V2 and
*[d, z, d3]smoK/V1L__V2 are active or inactive according to lexical items evaluated as

shown in (12) and (13). Therefore, I have shown that POT, PNT, PLT, and voicing of
voiceless lax-02 stops result from interaction of the same phonetically grounded
constraints. Therefore, there is no need to suggest different phonological rules as has
been the case in the Korean linguistic literature (Kim-Renaud 1974, Ahn 1985, Sohn
1987, and many others) (e. g. chapter 3 (48)).

In addition, I have suggested in the gesture-based account of PNT and PLT that
morphological and/or additional phonological conditions emerge to interrupt strong
phonological or phonetic conditions for voicing of lax-02 stops. Therefore, there is a
close relationship between morphological and/or additional phonological conditions and
distribution of gestures relevant to voicing or tense in a given context.

In addition, the gesture-based account of PNT and PLT in this study provides
phonetic reasons why all lax 02s are tensed after verb-stem-ending nasals in verbal
morphology while only /t, s, c/ after the /1/ are tensed in Sino-Korean nouns, as shown in
(14) and (15). Patterns of PNT and PLT are motivated by CDs in the vocal tract as is
POT. Therefore, ObsT occurs at the levels where CDs of Cl and C2 form

[closed][closed] or [closed][critical] CD sequences. This is because [closed][closed] or

221

 

[closed][critical] CD sequences create favorable conditions for tensification of lax 02s

rather than voicing of lax-02 stops (Ohala 1983).

222

4.3 Patterns of /h/ and obstruent sequences

4.3.1 Introduction

No surface forms end with [h] in Korean. Therefore /h/ cannot be in syllable and
word-final position. However, due to morphological concatenations such as addition of a
prefix or suffix, there are cases where /h/ appears before lax-02 stops, resulting in the

sequence of VlhOsz, as shown in (16).

(16) a. Lax-02 stop aspiration: fast or casual style of speech

Input Output
/SUh/prefix ‘male’ + /l_<_a/ ‘dog’ [swim] ‘male dog’
/suh/ prefix ‘male’ + lpom/ ‘tiger’ [suphom] ‘male tiger’
mob/VS ‘to place’ +/ko/1}:M [no.k_ho] ‘place and’
/noh/VS + /Q&/ [FM [no.gha], ‘let’s place something’

b. Lax-02 stop aspiration and CN of /h/:
slow or formal style of speech (Kim-Renaud 1974: 124)

Input Output
/SUh/prcfix ‘male’ + keel ‘dog’ [sut‘.k_hae] ‘male dog’
/suh/ prefix ‘male’ + lpom/ ‘tiger’ [guiding ‘male tiger’
/noh/vs ‘to place’ +/ko/1FM [not’.,k_"o] ‘place and’
mob/vs + /Qa/ [FM [[10303]] ‘let’s place something’

223

c. Lax-02 stop aspiration and regressive place assimilation and CN:
slow but casual style of speech (Park 1990: 119)

Input Output
mun/W,x ‘male’ + /kae/ ‘dog’ [suk’flae] ‘male dog’
/suh/ prefix ‘male’ + /p_9m/ ‘tiger’ [sup‘,ph9m] ‘male tiger’
/no_h/vs ‘to place’ +lko/1FM [nokfiflo] ‘place and’
/noh_ /vs ‘to place’ + /ga/H:M [not’da] ‘let’s place something’

If the underlying forms of the prefix ‘male’ and the verb ‘to place’ are /su/ and

/no/ respectively, as in the output form (16a), /k/ in /kze/ ‘dog’ and /k/ in /ko/1FM should

be voiced between two vowels, rather than aspirated. Therefore, the underlying forms of
the prefix ‘male’ and the verb ‘to place’ are /suh/ and /noh/ respectively, as shown in
(16a).

Recall that in V10102V2, 01 is neutralized as an unreleased stop in syllable-final
position due to *CCD/_ {#, C} and lax 02 is tensed obligatorily in Korean. However, in
the sequence of VlhOsz, although /h/ is considered as an obstruent at the VT level, as
shown in chapter 1 (27), a lax-Oz stop after /h/ is not tensed. Instead, /h/ and a lax-02
stop are obligatorily coalesced into an aspirate stop, as shown in (16).

Although coalescence of /h/ and a lax-02 stop in VlhOsz is obligatory,
neutralization of /h/ in syllable-final position depends on the rate or style of speech as
shown in (16b) and (16c). According to Park (1990) and Kim-Renaud (1974:124), output
forms such as (16b) and (16c) appear ‘in a slow, bookish, emphatic pronunciation’. I
assume that output forms in (16c) result from regressive place assimilation, which is
optional in the casual style of speech in Korean. Therefore, I analyze ( 16b) and ( 16c)

together in this study. However, constraints related to place assimilation are not

224

discussed here because this study mainly investigates manner of articulation (see J. Jun
1995 for place assimilation within OT)
In addition to (16), a lax-OI stop and /h/ in VlOth2 are obligatorily coalesced

into an aspirate stop, as shown in (17).

(17) a. Lax-OI stop aspiration: fast or casual style of speech

Input Output
/nakha/ [na.k_”a] ‘falling’
mim/ [kifii] ‘in a huny’
/anc+hi +da/ [an.g"i.da] ‘be seated’
/mas/ ‘taste’ + /hako/ ‘and’ [mathago] ‘and taste’

b. Lax-OI stop aspiration and CN of 01: slow or formal style of speech

Input Output
/nakh_a/ [nak’ki’a] 'falling'
kimi/ [kip’rfli] ‘in a hurry’(adverb)
/mas/ ‘taste’ + /hako/ ‘and’ [matthago] ‘and taste’

Although a lax-Ol stop and /h/ in VlOthz are obstruents, neutralization of lax-
01 stops in syllable-final position depends on the rate or style of speech while aspiration
of lax-OI stops is obligatory in Korean. In summary, data in (16) and (17) shows that lax
stops preceded or followed by /h/ are coalesced together obligatorily.

Up to this point, I have shown cases where lax O] and 02 are stops in sequences
of VlOth2 or V1hOzV2. When lax 01 and 02 are fricatives (F ), phonetic realization of

the fricatives depends on the location of the fricatives relative to /h/, as shown in (18).

225

(18) A. V1FhV2 (F: fricatives)

a. Aspiration: fast or casual style of speech

Input Output
/mas/ ‘taste’ + /hako/ ‘and’ [ma.tha. go] ‘and taste’
huh/prefix ‘male’ + lhama/ [su.f'_a.ma] ‘male hippo’

b. Aspiration and CN of 01: slow or formal style of speech

Input Output
/mas/ ‘taste’ + /hako/ ‘and’ [mat‘tha go] ‘and taste’
huh/prefix ‘male’ + /hama/ [sut’.t_h_a.ma] ‘male hippo’

B. V1hsV2

a. Tensification of /s/: fast or casual style of speech

Input Output
/SUI_1_/prcfix ‘male’ + /so/ ‘cow’ [SU.§’0] ‘OX’
Atom ‘to place’ + /so/ IFM [no.go] ‘place (command)’

b. Tensification of /s/ and CN of /h/: slow or formal style of speech

Input Output
huh/mm ‘male’ + /so/ ‘cow’ [sut’s’o] ‘ox’
/noh/_ ‘to place’ + /§0/IFM [not’io] ‘place (command)’

The F h cluster in V1FhV2 (F: fricatives) (18A) patterns similar to (17). Therefore,
Output forms always have aspiration of lax stops. However, the oral fricative /s/ after /h/
is tensed as shown in (18B). In (18Ab) and (18Bb), neutralization of 01 depends on the
rate or style of speech, as shown in (16b), (16c), and (17b). Hereafter, the term

‘ Aspiration’ covers coalescence of /h/ and lax obstruents, as shown in (l6), (l7) and

226

(18A). Because tensification of /s/ in the /s/-/h/ sequence has been discussed in section
2.2.3 and chapter 3, (18B) will not be discussed in the following sections.

Patterns of /h/ in (16), (17) and (18) are treated with separate phonological rules
in previous approaches (Kim-Renaud 1974, Park 1990 and others), as will be shown in

the following section.

4.3.2 Previous studies

Though different in details, Aspiration in (16a) and (17a) is described in the
Korean linguistic literature (Cho 1990, Park 1990, and Kim 1986) by using a mirror

image rule such as (l9).

(19) Mirror image rule (Park 1990: 113): Aspiration

r —tense h W
[continuant] 2
{ 1 —tense } —> [+asp] a
h -continuant 1
L 2 [I J J

 

 

(19) states that a lax stop and /h/ are coalesced into aspirate stops, resulting in [sukhae]
from /SUh_/prcfix ‘male’ + /_lga=:/ ‘dog’ or [nalgha] from makha/ ‘falling’, as in data (16a) and

(17a).
According to Kim-Renaud (1974) and Park (1990), the rule order of CN-related

rules (20a-c) after Aspiration (19) is responsible for output forms in (16b) and (17b), as

227

shown in (20). The default rules (20b) and (20c) are suggested for /h/ —> [t‘] in syllable-

final position, as shown in (20d).

(20) a. h —-> [-release]/ _$ ($: syllable boundary)

b. [h, -release] —> [+coronal] CN
c. [-release] —-> [-asp, -tense, -cont]
Aspiration CN
d. /suh/pm~Ix ‘male’ + /Lae/ ‘dog’—> [suhL'iae] -—> [sut’._k:ar]

In addition, Park (1990) proposes an optional rule which inserts a homorganic
consonant between a vowel and a tense or aspirate consonant in order to explain data in

(16c) and (17b). The insertion of a homorganic consonant is applied after aspiration as in

(21).

(21) Input Aspiration (18) Insertion of a homorganic consonant
/kip_h_i/ —+ [kiphi] —> [kipip i]

However, phonetic motivation for insertion of a homorganic consonant is not
provided in Park (1990). Contrary to the rule order suggested in (20), output forms in
(22) suggest CN before Aspiration, as shown in the comparison of (22b) and (22c). In

(22b), when Aspiration is first applied, the structural description for Aspiration (19) is not

met because there is no [-tense, -continuant] in /suh/p,efix‘male’ +/h_ama/ ‘hippo’.

Therefore, only CN is applied in the input, resulting in *[sut‘ .hama], as shown in (22b).

228

In (22c), CN is first applied to /suh+hama/ and the result of CN creates the

environment for the application of Aspiration, as shown in (220). Therefore, Aspiration

is applied to ‘suthama’, resulting in [suthama]. However CN before Aspiration as in

(22c) is the reverse of the rule order suggested in (20).

(22)

a. Input Output
mun/mm ‘male’ +/hama/ ‘hippo’ [sufiama]

/SU_h/prcfix ‘male’ +/holar)i/ ‘tiger’ [suflorarji]

Input

b. /SUh_/prcfix ‘male’ /hama/—-> Not Apply (NA)—>

Input

Aspiration

CN

c. /SUh/prcfix ‘male’ /hama/—> suthama ——>

or [sut’flama] ‘male

hippo’

or [suffioxagj] ‘male tiger’

CN Output
sut’.hama *[sut’hama]
Aspiration

suthama [suthama]

Park (1990) and Kim-Renaud (1974) describe the rule ordering paradox in (22)

with the syllabification principle or the following module of rule application.

(23)

Input
syllabification
CN
Aspiration
syllabification

Insertion of
a homorganic
consonant

a.
/SUI_'I_/prcfix ‘male’ +/_hama/ ‘hippo’
suhhama
suthama
suthama
NA
[suthama] (18Aa)
or

[sut‘thama] (18Ab)

229

b.

/noh/vs+ /ko/1FM ‘and place’
NA

NA

nokho

noE‘o

[nokho] (18a)

or

[nok‘Lho] (18b)

For example, according to Park (1990), syllabification occurs at the end of each
word cycle. In addition, in the presence of a syllable structure, syllable sensitive rules
such as CN are applied before syllable non-sensitive rules such as Aspiration.

CN is applied before aspiration in data (23a) because the syllabification has
occurred after prefixation. However, syllabification does not occur until the end of

derivation for data like (23b). Thus, Aspiration is applied before CN in (23b), resulting

in [no.kho] from /noh/VS+ /ko/1pM ‘and place’.
However, Park (1990) predicts incorrect output forms for data such as /SU_I_I_/prefix

‘male’ +/k&/ and /suh/ prefix ‘male’ + /p_om/ ‘tiger’, as shown in (24).

(24) a. /SUh/prefix ‘male’ +/k2e/
Syllabification suh . kee
CN sut’ . kee
Aspiration NA
Syllabification NA
*[sut’kee]
*[sut’.k’2e]
insertion of or
a homorganic *[suk’. k’m]
consonant

b. Park(1990 :117)

‘...the opposite order ASP before SFN... this is due to tighter cohesion between
compounding elements in frequently used terms. . . ’

Although the input in (24a) has the same prefix /suh/, meaning ‘male’ as in (23a),

output forms of /suh+ka-‘:/ are *[sut’km], *[sut’k’a] or *[suk’. k’ze], which are not correct.

230

Output forms of /suh+l<ae/ should be [sukhm], [sut’kha], or [suk’khae], as shown in (18a),

(18b), and (18c). Park (1990) claims that the rule ordering paradox in (24) is due to word
frequency as shown in (24b). However, in the following section, I will show that Park’s
explanation based on word frequency fails to recognize generalizations in data (22).

In fact, the rule ordering paradox, namely CN before Aspiration, is also found in
the sequence of V1FhV2, as shown in (22c) and (25a). In addition, the fricative /s/

inVlhsVZ is tensed, as shown in (25b).

(25)
Input Output
a. /phus+hopak/ [phuflobak] or [phuffiobak] ‘unripe pumpkin’
(compound noun)

‘ 3

b. /suh + 50/ [sus’o] or [sut’.s’o] ox

/noh +_s_o/ [no go] or [not’s’o] ‘place (command)’

In the following section I will argue that phonetic realizations of /h/ + obstruent
sequences in (16), (17), and (18) are consistent and predictable when phasing principles
of gestures in a given context and characteristics of individual gestures are considered
(Byrd 1996, Browman and Goldstein 1990a, Cho 1998a and 1998b). Therefore, output
forms in (16), (17), and (18) result from interaction of phonetically-grounded constraints

and faithfulness constraints, as do POT, PNT, and PLT.

231

4.3.3 The gesture-based account of /h/ and obstruent sequences

As a start, (26A) and (26B) evaluate candidates from /h/-/k/ and /k/-/h/ sequences
against constraints suggested for POT (26), because both /h/ and /k/ are obstruents.
Output constraints suggested for POT (26) correctly select output forms of (16a) and

(17a) as the most optimal candidates, as shown in (26). (26A) and (26B) evaluate
candidates from /I'th/Vs ‘to place’ + /ko/IFM ‘place and’ and /nokha/ ‘falling’

respectively. In 4.3.3, the sequence of two consonants is marked with numbers as in

/no_h1 + 1(20/ in order to relate input and output segments.

(26) *COMPLEX, *[mid]GL0/ V101_V2 >> Max-IOch >> *CCD/_{#, C},>>
[DENT-IO[G] >> Mini Eff=0

A. /noh/+/ko/ ‘and place (verb)’

/h/: GLO [open], SL [open], Oral [open]

 

 

 

 

 

 

 

/k/: GLO 'mid], SL[closed], Oral [closed], TB [closed]
/nOh_1+ kzo/ *COMP 5 *[mid]oLo/ Max- *CCD/ IDENT- Mini
E V101_V2 IOscg _{#, C} IO[G] Eff=0
b. no.kp_o *1 ‘ . f g
C. not]. 1(2th ”(GLO) 1 *(TT) . * .
' *(SL) * . 3»
d not’1.k2h’o *!(GLO) - *m). 4 _ ,****
O — *(SL) . . if}:
e. not1 1. k’zo *!(GLO) '_1-*(TT) 7 a:
“ *(GLOL' ~ v ''''''
© f. not“ I. kzhlo *!(GLO) am

am _

 

 

G’g. no. 1(2th

 

 

 

 

 

 

51> indicates free variation with the output depending on the style of speech.

232

 

B. /nakha/ ‘falling’

/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]

 

 

 

 

 

 

/h/: GLO "open], SL [open], Oral [open
/na_]_(_1+ hza/ *COMP *[mid]oLo/ Max- *CCD/_{#, C} IDENT- Mini
g V101__V2 IOscg IO[G] Eff=0
a. nakla ; *! :5:if"iii-:i'iiri; 1717;fi‘iiii‘iiifi;1{iffifgrffliff;i; 1.17:);2:xiii-r13;
b. nakr’hza 3 “(SL) (GL0) "
Q o. nakr’krma 3 ”(SL) mom)"
@‘d na klhza

 

 

 

 

 

 

 

Tableaux (26A) and (26B) illustrate that coalescence of a lax stop and /h/ results
from most optimal satisfaction of output constraints in (26). Therefore it is unnecessary
to suggest extra rules or constraints only for Aspiration, as have previous analyses (Cho
1990, Park 1990, and Kim 1986) does.

In addition, (26) illustrates how directions of Aspiration in data (16) and (17) are
produced in the course of satisfying output constraints, as opposed to directions that are
given or stipulated. Therefore Padgett’s (1996:20) claim that the direction of assimilation
is determined by output constraints is supported here. Therefore, the mirror image rule in
(19) is not necessary to understand direction of Aspiration. As has been seen in this
study thus far, constraints responsible for POT, PNT, PLT, and Voicing of lax stops after
sonorants are also responsible for Aspiration, as illustrated in (26).

As output forms in (16a) and (173) have been explained, it is now necessary to
investigate data (16b), (16c), and (17b). In (16b), (16c), and (17b), there is Aspiration of
lax stops and CN, as shown in candidates (26Af) and (26Bc). Note that candidates
(26Af'=>) and (26Bc=i>) are used freely with (26Ag‘9’) and (26Bd‘9’) respectively in the

slow or formal style of speech.

233

In (26), there are two main points to be explored. First, why is (26Bc=i>) preferred
over candidate (26Bb) even though (26Bb) and (26Bc©) are tied in constraint violations?
Second, why do (26AfC¢>) and (26Bc'i>), which are free variants of (26Ag9') and
(263d?) respectively, have unreleased stops in syllable-final position?

In order to explain the preference of (26BCC>) over candidate (26Bb), it is
necessary to introduce the phasing principles suggested by Browman and Goldstein
(1990a and 1992), Byrd (1996), and Cho (1998a and 1998b). The phasing principles deal
with how gestures are organized when they are overlapped in time in a given context.

The phasing principles are based on the idea that gestures have their own intrinsic
time and space. In addition, gestures do not move one after another but are phased with
respect to other gestures. The overlapping of gestures is sensitive to locations where
gestures appear in a syllable structure, the articulator involved, the location of a gesture,
prosody, and grammatical categories such as morphemes (Browman and Goldstein 1990,
Byrd 1996, Cho 1998b).

For example, Browman and Goldstein (1990a) suggest that gestures are likely to
be overlapping each other within a syllable structure. In addition, according to Byrd
(1996) and Cho (1998b), there are less overlap pings and less timing variables in an onset
consonant cluster (#CC) than in a hetero-syllabic consonant cluster (C#C) or a coda
cluster (CC#), as shown in (27a).

(27) a. #CC << C#C, CC# ,
(A<< B: A is less overlapped and less variable in its timing than B)

b. [S] . [g] << [(1] - [g]
C- [8] + [8] << [81+ [(1]

234

In addition, Byrd (1996) states that phasing principles should also reflect place,
manner, and sequence effects. For example, the alveolar fricative /s/ is less overlapped
by the following velar stop /g/ (27c) than is an alveolar stop in coda position. Byrd also
found that the fricative [s] is less overlapped than [(1] regardless of the syllable position
(27b and c).

Browman and Goldstein (1990a and 1992) propose how input gestures are
organized in the output when two gestures overlap in time. When two input gestures are
on different tiers, gestures are overlapped in time without disturbing each other’s tract.
When two input gestures are on the same tier, the two gestures disturb each other because
the same articulator moves toward different targets. Thus, gestures are likely to be
altered.

For example, according to Browman and Goldstein (1992), in /ada/ the consonant
and vowel gestures achieve their target invariably regardless of the overlapping of vowel
gestures. This is because input gestures for /d/ and /a/ are on different tiers, namely the
TB for /a/ and the TT for /d/. However, in /idi/, /d/ and /i/ are on the same tier (TB), thus
the consonant and vowel gestures cannot achieve their targets simultaneously because the
TB moves toward different positions, resulting in palatalization of /d/.

All gestures are overlapped in time to a certain degree. However, languages
differ in the length of overlapping between segments. Experimental studies have shown
that Korean has longer overlapping time between onset and coda consonants than English
does (J. Jun 1995, Cho 1998a and 1998b, Keating 1990, and Silverman and Jun 1994).
As Korean allows relatively longer overlapping time between segments, there are cases

when the phasing principles discussed so far play a crucial role in Korean grammar.

235

Based on the fact that Korean has longer overlapping time between onset and
coda consonants than English does, I adopt the constraint Overlap suggested by Cho

(1998b) as shown in (28).

(28) Overlap: Two adjacent gestures must be overlapped.

The constraint Overlap favors the overlapping of gestures in time. Therefore,
gestures do not move one after another. In other words, gestures are likely to be phased
with respect to other gestures. In the case of gestural overlapping, the phasing principles
introduced in this section play a role in shaping the phonetic realization of segments
involved.

According to Browman and Goldstein (1992), there are three distinct types of
temporal overlap: minimal overlap, partial overlap, and complete overlap. Based on
these three types of temporal overlap, Cho (1998a and 1998b) evaluates the constraint
Overlap in a gradient fashion. However, in this study, I will consider the constraint
Overlap in terms of the presence or absence of the overlapping of gestures, as will be
shown in (29).

The output forms in (16a) and (17a) and their free variants (16b), (16c), and (17b)
satisfy Overlap always, resulting in Aspiration. However, free variants of (16a) and (17a)
violate *CCD/_{#, C} due to unreleased stops in syllable-final position, as shown in data
(16b), (16c), and (17b). Therefore, satisfaction of Overlap takes precedence over
satisfaction of *CCD/_{#, C} in actual output forms. Therefore, Overlap is ranked

higher than *CCD/_{#, C}, as shown in (29).

. 236

 

(29) *COMPLEX, *[mid]GLo/ V101_V2 >> Max-10368, Overlap >>
*CCD/_{#, C}>> IDENT-IO[G] >> Mini Eff=0

/nakha/ ‘falling’
/k/: GLO [mid], SL [closed], Oral [closed], TB [closed]

 

 

 

 

 

/h/: GLO ropen], Sflopen], Oral[o en] -
/nak1+ hza/ *COMP *[midIGLo/ Max- [Overlap *CCD/ IDENT Mini
E V101 V2 Ioscg a _{#,C} -IO[G] Efi°=0
a. nakl a *i
b.nak1" ha i i *1 *2*<SL)5~"9199)I ;iif:j.?".‘fjr
=9 ’ h: 5 5 ”(SL) £175,755fi"."~j’§'..’f?{§;**
c' ”‘11" k‘ a f. g s-ayrietori;
C971.113 klhza : '

 

 

 

 

 

 

 

The ranking between Overlap and Max-IOseg is not known at this time as available data

was unable to determine the ranking between Overlap and Max-IOseg.

Up to this point, I have provided a reason why [nak‘kha] (26BcEi> or 29c) is
preferred over [nak’.ha] (26Bb or 29b) even though they are tied in other constraint

violations. [nak’. kha] (26Bc=>) is preferred over [nak’.ha] (26Bb) because Overlap (28)

plays a crucial role in selecting output forms, as shown in (29).

Note that free variants of (16a) and (17a) in the slow or formal style of speech

always have unreleased stops in syllable-final position, as shown in (30).

(30) a. slow/formal style of speech b. fast/casual style of speech
/nohko/ [not‘.kho] (18f) [no.kho] (18g)
/nakha/ [nak’kha] (180) [nakha] (18d)

237

I suggest that syllable-final unreleased stops in the slow or formal style of speech
are due to the decreased gestural overlap. The increase or decrease of the gestural
overlap is also viewed as a function of speech rate in Browman and Goldstein (1990a and
1990b), Byrd and Tan (1996), J. Jun (1995), and Cho (1998a and 1998b). Especially
Byrd and Tan’s experimental study shows that decreasing articulatory duration and
increasing co-production between successive articulations are the characteristics of fast
speech.

In order to investigate the phasing relationship between two adjacent gestures,
Cho (1998b) conducted an interesting experimental study using Korean. Cho observed
the consonant sequences in (31), using an Electro-magnetic midsagital articulometer
(EMA) for the approximation of the temporal movements of articulators and an
Electropalatography (EPG) for the temporal relationship between gestures.

Cho (1998b: 31) found that there is less variability in gestural timing and
intergestural overlap in the tauto-morphemic-C1C2 sequences than in the hetero-

morphemic-C1C2 sequences, as shown in (31a) and (31b).

(31) a. Tauto-morphemic sequence b. Hetero-morphemic sequence
kp k#p
&c &#c
p i c“ # c

# indicates a morpheme boundary.

I
:—’ morpheme boundaries ‘ 'l

I

I Gl|G
I I
I

  
 
   

 

less variability in gestural timing;
less overlap

238

more variability in gestural timing;
more overlap

rm.

Based on the above observation, Cho (1998b) concluded that the intergestural
timing is lexically specified. In addition, Cho suggested a constraint IDENT (Timing)
(3 2) that demands maintaining of intergestural timing belonging to the same lexical entry

in order to explain plaltalization in the sequence of ‘ti’ and ‘ni’.

(32) IDENT (timing): (Cho 1998b)

Intergestural timing belonging to the same lexical entry must be preserved

I suggest that in the slow or formal style of speech such as (16b), ( 16c), and (17b),
speakers make effort to maintain the lexically specified timing. Therefore, IDENT
(timing) plays an active role in the slow or formal style of speech.

Output forms in (3 O) are in the free-variation relationship because they can be
used freely without changing meaning. Note that output forms in ( 16b), (16c), and (17b)
violate *CCD/_ {#, C} and maintain the input timing due to unreleased stops in syllable-
final position. Therefore, IDENT (timing) is ranked higher than *CCD/_ {#, C} in the

slow or formal style of speech, as shown in (33a).

(33) a. *CCD/_ {#, C} >>IDENT (timing) (in the fast or casual style of speech)
b. IDENT (timing) >>*CCD/_ {#, C} (in the slow or formal style of speech)

However, in (16a) and (17a), *CCD/_ {#, C} is satisfied because there are no
syllable-final unreleased stops but IDENT(timing) is violated because the input timing
slot occupied by /h/ or lax stops is not maintained in the output. Therefore, *CCD/_ {#,

C} is ranked higher than IDENT (timing) in the fast or casual style of speech, as shown

239

:ngT-f‘*mm

in (33b). Note that (33) implies free ranking between *CCD/__ {#, C} and IDENT
(timing) depending on the rate or style of speech. Main constraints and their rankings
such as Max-IOscg and Overlap are maintained while sub-parts of constraints are freely

ranked.
(34) evaluates candidates from /suh + horagi/ ‘male tiger’ with the V1FhV2

sequence (data in (18A)), for the fast or casual style of speech.

(34) *COMPLEX, *[mid]GL0/ V101_V2 >> Max-10868, Overlap >>
*CCD/_{#, C} >> IDENT(timing) >> IDENT-IO[G] >> Mini Eff=0

/SUh/prefix ‘male’ + /hora13i/ ‘tiger’ ‘male tiger’

/h/: GLO [open], SL [open], Oral [open]

/h/: GLO [open], SL [open], Oral [open]
/suh1+h20ra13i/ *[midloLo/ Max- g Overlap *CCD/ IDENT IDENT Mini
V101 V2 108cg 5 _{#, C} (Tuning) -IO[G] Efi‘=0

21‘ SUhI-hzo-raIJ-i *’ *(GLO)

- i *2 azritGLO')? 13:15:32 *rrcDja; *

b' SUt1.h20.I'aI].l *fl)[.3.-1:»5.;3‘953131"..ig :g' 7;]?

*1 ..

 

 

 

 

C. 81111201211}i . , .

 

 

e.sut’1.t1hzo.ra13.i . .. . .
a . :
(ff su.t1h‘o.rarj.r : * *TTCD

 

 

 

 

 

 

 

 

 

 

Note that in the fast or casual style of speech, *CCD/_{#, C} is ranked higher
than IDENT(timing), as shown in (34). (34c) is eliminated by Max-10362. In fact, it is
hard to determine if [h] in (34c) is from the first or second /h/. Candidates (34a), (34b),
(34d) are eliminated by Overlap because there is no overlap between the first /h/ and the

second /h/. (34c) satisfied IDENT(timing) due to maintaining of the input timing slot for

the first /h/, resulting in CN. However satisfaction of *CCD/_{#, C} takes precedence

240

 

 

over satisfaction of IDENT(timing) in fast or casual speech. Therefore, (34ft?) is the
most optimal candidate because it satisfies *CCD/_{#, C}, which is ranked higher than

IDENT(timing).
(3 5) evaluates candidates from /suh + horarji/ ‘male tiger’ in the slow or formal

style of speech. In the slow or formal style of speech, IDENT(timing) is ranked higher
than *CCD/_{#, C}, as shown in (3 5), because, speakers put forth effort to maintain

lexically specified timing.

(3 5) *COMPLEX, *[mid]GL0/ V101__V2 >> Max-10,68 >> Overlap >>
IDENT(timing) >> *CCD/_{#, C} >> IDENT-IO[G] >> Mini Eff=0

/suh+horarji/ ‘male tiger’

/h/: GLO [open], SL [open], Oral [open]

/h/: GLO [openLSLio en], Oral [open]

/ suh1+hzorarji / *[mid]G Max- Overlap IDENT *CCD/ IDENT Mini

L0, 10s,,g (Timing) _{#, C} -IO[G] Eff=0

VlOl_
V2

a. suh1.hzo.rarJ.i *’ *(GLO)

- *2 ;.:T?~ir<GLor-_*'.*izrrcn *

b‘ S““'h2°'r“9" (SL)

 

 

 

 

d sutr’.hzoraIJ.i *! :rjzzrrsrtras:Trev
@esut’r .tih‘oraIJ-I *(SL) gi*TTC91 .j ** - i.
’7 , '31] 2:47.317. -’ . I . ‘3 3 7
f su.t1h‘o.rar].r 1” *TTCD '

 

 

 

 

 

 

 

 

 

 

 

 

 

A comparison of (35e) and (35f) clearly shows that in slow or formal speech,
IDENT(timing), which requires maintaining of the lexically specified timing, is active in
selecting the most optimal candidate. (35e) satisfies *CCD/_{#, C} but satisfaction of

IDENT(timing) takes precedence over satisfaction of IDENT(timing). Therefore, free

241

 

variants of (16a), (17a), and (18Aa) have unreleased stops in syllable-final position
because IDENT(timing) is ranked higher than *CCD/_{#, C} in the slow or formal style
of speech.

According to Kager (1999: 404-07), a free ranking approach is appropriate in
explaining free variation because explanations for degrees of dissimilarity between output
forms derive from maintaining most constraint hierarchies, while free ranking only a
portion of constraints, as shown in (34) and (3 5).

The gesture-based account of Aspiration in this study provides simplicity and
generalizations in explaining data (16), (17), and (18A), as shown in (26), (29), (34), and
(35). Previous studies of Aspiration (Cho 1990, Park 1990, and Kim 1986) could not
offer such explanations.

In the gesture-based account of Aspiration, output forms in (16), (17), and (18)
are the most optimal candidates that are selected in the course of satisfying output
constraints. Thus, there is no need of a mirror image rule for Aspiration, different kinds
of syllabification modules, or a rule ordering paradox for data in (22) as shown in section

4.3.2.

242

 

. -TT

Chapter 5 Conclusion

The analyses in this study have been based on two main theoretical assumptions:
Articulatory Phonology, which integrates phonetics and phonology (Browman and
Goldstein 1986, 1989, 1990a, 1990b, 1990c, 1991, 1992 and Browman 1994) and
Optimality Theory (Prince and Smolensky 1993 and McCarthy and Prince 1993a, 1994
and 1995). Articulatory Phonology is adopted for the units of representation and
Optimality Theory for relating an input to an output.

I adopted the articulatory gesture as the basic unit of phonetic and phonological
representation. In Articulatory Phonology, the presence or absence of a gesture, different
constriction degrees (CDs), and the timing or organization of gestures (Browman and
Goldstein 1986 and 1991) are contrastive. Therefore, any changes in gestures or
organization of gestures, in turn, are reflected in phonology. Especially CDs in
articulators and constriction degree hierarchies (CDHs) predicted from the CD of an
individual articulator were found to be useful in this study. This is because manner of
articulation is closely related to CDs of the vocal tract. Therefore, the gesture-based
accounts of consonant patterns in Korean have provided insight into understanding
various phonological phenomena in Korean.

In chapter 2 of this study, Coda Neutralization (CN) is considered as a by-product
of phonetic implementational conditions in a given context. The absence of active glottal
gestures and lack of air escaping through the oral cavity for obstruents in syllable-final
position create articulatorily and acoustically unfavorable conditions to maintain

contrastive CDs in the vocal tract. Thus CN was analyzed systematically in terms of

243

 

CDs, which are closely related to manner of articulation. I have shown in section 2.1 that
interaction of *Contrastive Construction Degree (CCD) /_{#, C} and IDENT-IO[Gesture]
is responsible for CN. Therefore, I conclude that CN is a neutralization in manner of
articulation that occurs systematically in syllable-final position in Korean. Previous
studies of CN fail to treat CN as a neutralization in manner of articulation because they
treat CN as delinking of laryngeal features, [+continuant] and/or [-anterior], as shown in
section 2.1.2 (3). Note that delinking of [~anterior] indicates neutralization in place of
articulation.

In addition, I have shown that constraints responsible for CN also play a crucial
role in Coda Consonant Cluster Simplification (CCCS) of Seoul Korean (SK) and
Kyung-Sang Korean (KSK), especially when two obstruents compete for the coda
position. In a fricative + stop cluster, the fricatives never appear in the output because
*CCD/_{#, C} is relatively ranked higher than IDENT-IO[G]. In addition, in KSK, when
competition for the coda position occurs in a sonorant-obstruent cluster, a candidate
ending with a sonorant is selected as the most optimal candidate regardless of articulators

of obstruents because *CCD/_{#, C} is ranked relatively higher than constraints that
require the Tongue Body, Lips, and Tongue Tip articulators (TBArt, LipsAn and TTAn

respectively) to be maintained, as shown in section 2.2.2.2 (53). Furthermore, in CCCS,
phonetic realizations of syllable-initial consonants are predicted based on gestural
characteristics of non-surfaced consonants, as shown in section 2.2.3 (58).

In chapter 3, I have argued that Post-Obstruent Tensification (POT) results from
enriched tense gestures at the oral-closure release of 02 in the sequence of V10102V2,

where 02 is a lax obstruent. In V10102V2, phonetic implementational conditions

244

 

promote tensification of lax 028 because of long closure duration between two vowels
and the minimal glottal opening at oral-closure release for 02, which are coupled with

heightened oral pressure and tension in the vocal tract. Therefore, the phonetically
grounded constraint *[mid]GL0/ V101_V2, which prevents lax 025 in V10102V2 due to

distribution of tense gestures, plays an active role.

The gesture-based account of POT in this study offers several advantages that
previous studies on POT (Han 1992 and 1997, Lee 1994, Kim—Renaudl974, and Rhee
1997) do not offer. The following generalizations and predictions are made through the
gesture-based account of POT: First, when C25 do not have the [open] CD at the
Supralaryngeal (SL) level, C23 participate in Obstruent Tensification (ObsT) obligatorily.
When C25 have the [open] CD at the SL level, C25 do not participate in ObsT except for
certain cases in the verbal morphology and Sino-Korean nouns. Therefore ObsT is
sensitive to CDs at the SL level. Second, the more [open] the CD in C1, the more
conditions are attached to ObsT because the [open] CD at the oral level is more likely to
create favorable phonetic implementational conditions for voicing, rather than
tensification (Ohala 1983).

Chapter 4 dealt with Post-Nasal Tensificatin (PNT) and Post-Lateral Tensification
(PLT). In previous studies, PNT and PLT were described with morphological conditions
and additional phonological rules, without providing any phonetic and phonological
motivations, as shown in section 3.3 (48). In addition, POT, PNT, and PLT were
unconnected and unrelated. I have shown that phonetically motivated constraints and
their hierarchies responsible for POT are also responsible for PNT and PLT to a degree.

Therefore, POT, PNT, PLT, and Voicing of lax-02 stops result from interaction of the

245

 

Elfin-r" ' 3:. ._ u

same phonetically grounded constraints and morphology-sensitive constraints such as

*[critical]GL0/V1N]vs _V2 and *[d, z, d3]SimK0,ean/V1L_V2, as shown in section 4.2.2

(12) and (13).

In addition, this study provided phonetic reasons why all lax 023 are tensed after
verb-stem-ending nasals in verbal morphology while only /t, s, c/ after the lateral /l/ are
tensed in Sino-Korean nouns, as shown in section 4.2.2 (14) and (15). I suggested that
lax 025 are tensed at the level where CDs of C1 and C2 form [closed][closed] or
[closed][critical] CD sequences. In N-lax 02 sequences, [closed][closed] or
[closed][critical] CD sequences are formed at the Oral level. Therefore, all lax 023 are
tensed after nasals. In L-lax 02 sequences, [closed][closed] or [closed][critical] CD
sequences are formed only at the TT level, and only when lax 025 are /t, s, c/ because /1/
and /t, s, c/ share the TT, as shown in section 4.2.2 (15a).

In addition, I have argued that Aspiration in /h/ + obstruent sequences as in (16),
( 17), and (18) are consistent and predictable when phasing principles of gestures in a
given context and characteristics of individual gestures are considered (Byrd 1996,
Browman and Goldstein 1990a, and Cho 1998a and 1998b).

The tableaux in section 4.3.3 (26A), (26B), (29), (34) and (35) illustrate that
Aspiration results from most optimal satisfaction of output constraints suggested for
ObsT. Therefore it was unnecessary to suggest extra rules or constraints only for
Aspiration, as previous Korean linguistic literature (Cho 1990, Park 1990, and Kim 1986)
does. Especially, section 4.3.3 (29) illustrates that the constraint Overlap is active and
ranked relatively high, resulting in Aspiration. I have also shown that directions of

Aspirations are determined in the course of satisfying output constraint hierarchies most

246

 

optimally, rather than by different rules or stipulations such as the mirror image rule in

section 4.3.2 (19) and (20). In addition, [sut’kfae], [nak‘.l_(h_a], and [mat‘t ago], which

are free variants of [nakfla], [suk’la], and [mathago], are explained based on the close

relationship between the rate of speech and intergestural timing of gestures. Output
forms with unreleased stops in syllable-final position result from the constraint ranking
order of IDENT(timing) over *CCD/_{#, C} in the slow or formal style of speech.
However, output forms without unreleased stops in syllable-final position result from the

constraint ranking order of *CCD/_{#, C} over IDENT(timing) in the fast or casual style

of speech. The explanation for free variants of [nafla], [su.k_hae], and [ma.tha.go] was

possible by adopting the basic concept of Articulatory Phonology that gestures have their
own time and space specified in the lexical entry. Therefore, the rule-ordering paradox
between (20) and (22) in section 4.3.2 is resolved by free ranking between *CCD/_ {#,
C} and [DENT (timing) according to the rate or style of speech while other constraints
and their rankings are maintained.

In conclusion, Korean consonant patterns (CN, POT, PNT, PLT, Voicing of lax
stops, and Aspiration), which are closely related to manner of articulation, have been
explained in terms of CD in the vocal tract in this study. Therefore, this study presented
a unified analysis of Korean consonant patterns within Articulatory Phonology
(Browman and Goldstein 1986 and 1990a) and Optimality Theory (Prince and
Smolensky 1993 and McCarthy and Prince 1993a, 1994 and 1995)

By adopting basic tenets of Articulatory Phonology and Optimality Theory, I have

shown that phonological phenomena treated as separate, as exceptions, or as unrelated

247

o'. {in Germ-1" In H”

phenomena can be combined and treated as regular and related phenomena within a
gestural approach.

Korean linguistic studies (Cho 1990, Kim-Renaud 1974, SA. Jun 1993 and 1995,
Rhee 1997, and Silva 1993) have suggested implicitly and explicitly that various aspects
of Korean consonant patterns investigated in this study would benefit greatly from an
approach that takes into account the interplay between phonetics and phonology.

However, few studies have been attempted that connect phonetic characteristics to
phonological explanations by using the same representational units (gestures in this
study) for both phonetics and phonology. Most studies based on Articulatory Phonology
deal with only one phonological phenomenon in Korean phonology such as place
assimilation or palatalization (J. Jun 1995 and Cho 1998a and 1998b), so they could not
identify generalizations that lie behind different phonological phenomena such as CN,
CCCS, PNT, POT, Aspiration, and others.

In addition, previous Korean studies based on Articulatory Phonology focus only
on explaining gestural movements in a certain context (S. A Jun. 1995). None of the
previous studies of which I am aware investigate various aspects of Korean consonantal
behaviors in terms of CDs, which are closely related to manner of articulation.

This study makes several contributions to promoting a better understanding of
Korean Phonology and Phonology in general. First, efforts made in this study to explain
Korean consonant patterns in terms of gestures provide a systematic and unified analysis.
Second, by integrating phonetics and phonology using phonetically grounded constraints,
some idiosyncratic phonological phenomena are better able to be explained and

understood.

248

 

I would like to propose a suggestion for future studies to enhance our
understanding of the interplay between phonetics and phonology as well as Korean
phonology. It would be helpful to know if phonological descriptions based on gestures,
especially CDs, could be extended to other languages and provide similar benefits for the

understanding of those languages.

249

' ‘A’r' .- 1
‘ .

’k~I.‘..

rid .Qh.... 42"?
"I

References

Ahn, Sang-Choel. 1985. The interplay of phonology and morphology in Korean.
Urbana-Champaign, IL: University of Illinois dissertation.

Alderete, John. 1997. Dissimilation as local conjunction. Proceedings of North East
Linguistics Society 27, ed. by Kiyomi Kusumoto, 17-31. Amherst, MA: p
Graduate Linguistic Student Association.

Anttila, Arto. 2000. Morphologically conditioned phonological alternations. RCA-425,
Rutgers Optimality Archive, http://roa.rutgers.edu/.

 

Archangeli, Diana. 1988. Aspects of underspecification theory. Phonology 5.183-208. ,-

tk- I

Archangeli, Diana and D. Terence Langendoen. 1997. Optimality Theory: An overview.
Malden, MA: Blackwell Publishers Inc.

Archangeli, Diana and Douglas Pulleyblank. 1993. Grounded Phonology.
Cambridge, MA: MIT Press

Benua, Laura. 1997. Transderivational identity: Phonological relations between words.
Amherst, MA: University of Massachusetts dissertation.

Bickley, Corine A. and Kenneth N. Stevens. 1986. Effects of a vocal-tract constriction
on the glottal source: Experimental and modeling studies. Journal of Phonetics
14.3 73-82.

Browman, Catherine P. 1994. Lip aperture and consonant release. Phonological
structure and phonetic form: Papers in laboratory phonology IH, ed. by Patricia
A. Keating, 331-53. Cambridge: Cambridge University Press.

Browman, Catherine P. and Louis Goldstein. 1986. Towards an articulatory phonology.
Phonology Yearbook 3.219-252.

Browman, Catherine P. and Louis Goldstein. 1989. Articulatory gestures as

phonological units. Phonology 6201-51.

250

Browman, Catherine P. and Louis Goldstein. 1990a. Gestural specification using
dynamically defined articulatory structures. Journal of Phonetics 18299-3 20.
Browman, Catherine P. and Louis Goldstein. 1990b. Representation and reality:
Physical systems and phonological structure. Journal of Phonetics 18.411-24.
Browman, Catherine P. and Louis Goldstein. 1991a. Tiers in articulatory phonology,
with some implications for casual speech. Papers in laboratory phonology 1:
Between the grammar and physics of speech, ed. by John Kingston and Mary E. I
Beckman, 341-76. Cambridge: Cambridge University Press.
Browman, Catherine P. and Louis Goldstein. 1991b. Gestural structures:

Distinctiveness, phonological processes, and historical change. Modularity and

—_—. ‘nfl x.-

 

the motor theory of speech perception, ed. by Mattingly Ignatius G and Michael

‘—_._—

Studdert-Kennedy, 313-38. Hillsdale, NJ: Lawrence Erlbaum Associate, Inc.

Browman, Catherine P. and Louis Goldstein. 1992. Articulatory Phonology: An
overview. Phonetica 49.155-80.

Byrd, Dani. 1996. Influences on articulatory timing in consonant sequences. Journal of
Phonetics 24.209-44.

Byrd, Dani and Cheng Cheng Tan. 1996. Saying consonant cluster quickly. Journal of
Phonetics 24.263-82.

Catford, J. C. 1977. Fundamental problems in phonetics. Bloomington, IN: Indiana
University Press.

Cho, Kyu Bin. 1995. High school grammar (in Korean). Seoul: Chi Hak Sa

Cho, TaeHong. 1998a. Intergestural overlap and timing in Korean palatalization: An
optimality theoretic approach. Japanese Korean Linguistics 8, ed. by David J.
Silva, 261-76. Stanford, CA: CSLI Publications.

Cho, TaeHong. 1998b. Specification of intergestural timing and gestural overlap: EMA
and EPG studies. Los Angeles: University of California thesis.

Cho, Young-Mee Yu. 1990. Parameters of consonantal assimilation. Stanford, CA:

Stanford University dissertation.

251

Cho, Young-Mee Yu and Sharon Inkelas. 1994. Post-obstruent tensification in Korean
and geminate inalterability. Theoretical issues in Korean linguistics, ed. by Young
Key Kim-Renaud, 45-60. Stanford, CA: Stanford Linguistic Society, CSLI.

Chomsky, Noam and Morris Halle. 1968. The sound pattern of English. New York:
Harper & Row.

Chung, Kook. 1980. Neutralization in Korean: A functional View. Austin, TX:
University of Texas dissertation.

Clements, George N. 1985. The geometry of phonological features. Phonology
Yearbook 2.225-52.

Clements, George N. 1991. The role of the sonority cycle in core syllabification. Papers
in laboratory phonology 1: Between the grammar and physics of speech, ed. by
John Kingston and Mary E. Beckman, 283-333. Cambridge: Cambridge
University Press.

Clements, George N. 1997. Berber syllabification: Derivations or constraints?
Derivations and constraints in phonology, ed. by Iggy Roca, 289-330. Oxford:
Oxford University Press.

Clements, George N., and Elizabeth V. Hume. 1995. The internal organization of Speech
Sounds. The handbook of phonological theory, ed. by John A. Goldsmith, 245-
306. Cambridge, MA: Blackwell publishers.

Clements, George N., and Samuel Jay Keyser. 1983. CV phonology: A generative
theory of the syllable. Cambridge, MA: MIT Press.

Daniloff, R. G., and R. E. Hammarberg. 1973. On defining coarticulation. Journal of
Phonetics 1.239-48.

Dart, Sarah N. 1987. An aerodynamic study of Korean stop consonants: Measurements
and modeling. Journal of Acoustical Society of America 81.1.138-47.

Dinnsen, Daniel. A. 1985. A re-examination of phonological neutralization. Journal of

Linguistics 21 .265-79.

252

Dinnsen, Daniel. A., and J. Charles-Luce. 1984. Phonological neutralization, phonetic
implementation, and individual differences. Journal of Phonetics 12.49-60.

Dorman, Michael F., Lawrence J. Raphael, and David Isenberg. 1980. Acoustic cues
for a fricative-affricate contrast in word-final position. Journal of Phonetics 8.
397-405.

F ant, Gunnar. 1977. The nature of distinctive features. Phonological theory: Evolution
and current practice, ed. by Valerie Becker Makkai, 360-65. Lake Bluff, IL:
Jupiter Press.

F errero, F. E., G. M. Pelamatti, and K. Vagges. 1982. Continuous and categorical
perception of a fricative-affricate continuum. Journal of Phonetics 10.231-44.

Flemming, Edward. 1995. Auditory representations in phonology. Los Angeles: UCLA

.‘a‘vuaJ-n r‘h assu- ‘hn'a JAOP-laz‘bfld-In‘vfi
! . __.,
a

dissertation.

F ourakis, Marios. 1984. Should neutralization be redefined? Journal of Phonetics 12.
291-96.

Fowler, Carol A. 1990. Comments on the contributions by Pierrehumbert and Nearey.
Journal of Phonetics 18.425-34.

Fromkin, Victoria A. 1975. The interface between phonetics and phonology. UCLA
Working Papers in Phonetics 31.104-7.

Fujimura, Osamu. 1972. Acoustics of speech. Speech and cortical functioning, ed. by
John Gilbert, 107-65. New York: Academic Press.

Fujimura, Osamu. 1991. Comment: Beyond the segment. Modularity and the motor
theory of speech perception, ed. by Ignatius G. Mattingly and Michael Studdert-
Kennedy, 25-31. New Jersey: Lawrence Erlbaum Assoiciate, Inc.

Fujimura, Osamu and Masayuki Sawashima. 1971. Consonant sequences and laryngeal
control. Annual Bulletin, Research Institute of Logopedics and Phoniatrics 5.1-6.
Tokyo: University of Tokyo.

Goldsmith, John A. 1976. Autosegmental Phonology. Cambridge, MA: MIT Ph. D.

dissertation.

253

Goldsmith, John A. 1990. Autosegmental and metrical phonology. Cambridge, MA:
Basil Blackwell Ltd.

Grijzenhout, Janet and Martin Kraemer. 1999. Final Devoicing and voicing assimilation
in Dutch derivation and cliticization. ROA-303-0399, Rutgers Optimality
Archive, http://roa.rutgers.edu/.

Gussenhoven, Carlos and Haike Jacobs. 1998. Understanding phonology. London:

Arnold.

Gussmann, Edmund. 1992. Resyllabification and delinking: The case of Polish voicing.

__9 A...

Linguistic Inquiry 23.29-56.
Halle, Morris and William J. Idsardi. 1997. r, hypercorrection, and Elsewhere Condition.

gay. “ crr‘rj

Derivations and constraints in phonology, ed. by Iggy Roca, 331-48. Oxford:
Oxford University Press.

Han, J eong-Im. 1992. On the Korean tensed consonants and tensification. Paper
presented at the 28th meeting of Chicago Linguistics Society.

Han, Jeong-Im. 1996. Perception of Korean tense and lax consonants: Evidence for a
geminate analysis of tense consonants. Japanese Korean Linguistics 5, ed. by
Noriko Akatsuka, Shoichi Iwasaki, and Susan Strauss, 407-23. Stanford, CA:
CSLI Publications.

Han, J eong-Im. 1997. Interface between phonological timing and phonetic duration in
Korean. Language Research 33.2.265-92. Seoul: Language Research Institute,
Seoul National University.

Han, Mieko S. and Raymond S. Weitzman. 1970. Acoustic features of Korean /P, T, K/,
/p,t,k/ and /p", t“, khl. Phonetica 22.112-28.

Hardcastle, W. J. 1973. Some observations on the tense-lax distinction in initial stops in
Korean. Journal of Phonetics 1.263-72.

Harris, John. 1996. Phonological output is redundancy-free and fully interpretable.
UCLA Working Papers in Linguistics 8.1-26.

Hayes, Bruce. 1986. Inalterability in CV phonology. Language 62.321-52.

254

Hayes, Bruce. 1989. The prosodic hierarchy in meter. Phonetics and Phonology 1:
Rhythm and Meter, ed. by P. Kiparsky and G. Youmans, 201-60. Orlando, FL:
Academic Press.

Hayes, Bruce. 1996. Phonetically-driven phonology: The role of Optimality Theory and
inductive grounding. Los Angeles: UCLA, ms.

Henderson, J anett B. and Bruno H. Repp. 1982. Is a stop consonant released when

followed by another stop consonant? Phonetica 39.71-82.

Hirose, H., C.Y. Lee, and T, Ushijima. 1974. Laryngeal control in Korean stop

production. Journal of Phonetics 2.145-52.

———*,——..-—

_A‘ .‘ I;

Hooper [Bybee], Joan B. 1976. An introduction to Natural Generative Phonology. New

York: Academic Press.

iFF- ,_...__

Hudson, Grover. 1995. Consonant release and the syllable. Linguistics 33.655-72.

Idsardi, William J. 1998. Tiberian Hebrew spirantization and phonological derivations.
Linguistic Inquiry 29.37-73.

Inkelas, Sharon. 1994. The consequences of optimization for underspecification.
Berkeley, CA: University of California, Berkeley, ms.

Ito, Junko and Armin Mester. 1986. The phonology of voicing in Japanese: Theoretical
consequences for morphological accessibility. Linguistic Inquiry 17 .49-73.

Ito, Junko and Armin Mester. 1993. Licensed segments and safe paths.
Canadian Journal of Linguistics 38.197-213

Ito, Junko and Armin Mester. 1997. Sympathy theory and German truncation.
Proceedings of the Hopkins Optimality Workshop/Maryland Mayfest 1997, ed.
by Viola Miglio and Bruce Moréen. University of Maryland Working Papers in
Linguistics 5.117-39.

Ito, Junko and Armin Mester. 1998. Markedness and Word Structure: OCP effects in
Japanese. Santa Cruz, CA: UCSC, ms. ROA-255, Rutgers Optimality Archive,

http://roa.n.rtgers.edu/.

255

Ito, Junko and Armin Mester. 1999. On the sources of opacity in OT: Coda processes
in German. ROA-347-0999, Rutgers Optimality Archive, http://roa.rutgers.edu/.

Ito, Junko, Armin Mester, and Jaye Padgett. 1995. Licensing and underspecification in
Optimality Theory. Linguistic Inquiry 26.571-614.

Iverson, Gregory K. and Joseph C. Salmons. 1995. Aspiration and laryngeal
representation in Germanic. Phonology 12.369-96.

Iverson, Gregory K. and Hyang-Sook Sohn. 1994. Liquid representation in Korean.
Theoretical issues in Korean linguistics, ed. by Young-Key Kim-Renaud, 77-100.
Stanford, CA: Stanford Linguistic Society, C SLI.

Jakobson, Roman, Gunnar Fant, and Morris Halle. 1951. Preliminaries to speech
analysis. Cambridge, MA: MIT Press.

Johnson, Keith. 1997. Acoustics and auditory phonetics. Cambridge, MA: Blackwell
Publishers Inc.

Jun, Jongho. 1995. A constraint based analysis of place assimilation typology. Los
Angeles: UCLA dissertation.

Jun, Sun-Ah. 1993. The phonetics and phonology of Korean Prosody. Columbus, OH:
The Ohio State University dissertation.

Jun, Sun-Ah. 1994. The status of the lenis stop voicing rule in Korean. Theoretical
issues in Korean linguistics, ed. by Young-Key Kim-Renaud, 101-14. Stanford,
CA: Stanford Linguistic Society, CSLI.

Jun, Sun-Ah. 1995. Asymmetrical prosodic effects on the laryngeal gesture in Korean.
Phonology and phonetic evidence: Papers in laboratory phonology IV, ed. by
Bruce Connell and Amalia Arvaniti, 235-53. New York: Cambridge University
Press.

Kagaya, Ryohei. 1974. A fiberscopic and acoustic study of the Korean stops, affricates

and fricatives. Journal of Phonetics 2.161-80.

256

Kager, René. 1997. Rhythmic vowel deletion in Optimality Theory. Derivations and
constraints in phonology, ed. by Iggy Roca, 461-99. Oxford: Oxford University
Press.

Kager, René. 1999. Optimality Theory. New York: Cambridge University Press.

Kang, Ongmi. 1992. Korean prosodic phonology. Seattle, Washington: University of
Washington dissertation. ’

Keating, Patricia. 1985. CV phonology, experimental phonetics, and coarticulation.
UCLA Working Papers in Phonetics 62.1-13.

Keating, Patricia. 1988. The phonology-phonetic interface. Linguistics: the Cambridge

‘ ’_.' u‘fc’i MARHI'
I

'HIT‘K.

survey. 1. Llinguistic theory foundations, ed. by F. J. Newmeyer, 281-302.

\'.'Y

Cambridge: Cambridge University Press.

11.112.“-
,4

Keating, Patricia. 1990. Phonetic representations in a generative grammar. Journal of
Phonetics 18.321-34.

Keating, Patricia. 1991. The window model of coarticulation: Articulatory evidence.
Papers in Laboratory Phonology 1: Between the grammar and physics of speech,
ed. by John Kingston and Mary E. Beckman, 451-70. Cambridge: Cambridge
University Press.

Keating, Patricia (ed.). 1994. Phonological structure and phonetic form: Papers in
laboratory phonology III. Cambridge: Cambridge University Press.

Kehrein, Wolfgang. 1999. Phonology without affricates. Paper presented at the 4th HIL
Phonology Conference, at Leiden University, January 28-30.
http://www.leidenuniv.nl/hil/confs/hilp4/kehrein.htrn.

Kenstowicz, Michael. 1994. Phonology in generative grammar. Cambridge: Blackwell
Publishers.

Kenstowicz, Michael and Charles Kisseberth. 1977. Topics in phonological theory.
New York: Academic Press.

Kenstowicz, Michael and Charles Kisseberth. 1979. Generative phonology: Description

and theory. New York: Academic Press.

257

Kent, Ray D. and Charles Read. 1992. The acoustic analysis of speech. San Diego, CA:
Singular Publishing Group, INC.

Kim, Chin-Wu. 1965. On the autonomy of the tensity feature in stop classification
(with special reference to Korean stops). Word 21.339-59.

Kim, Chin-Wu. 1970. A theory of aspiration. Phonetica 21.107-16.

Kim, Hyunsoon. 1997. The phonological representation of affricates: Evidence from
Korean and other languages Ithaca, NY: Cornell University dissertation.

Kim, Hyunsoon. 1998. A phonetic characterization of release and nonrelease: The case
of Korean and English. Language Research 34.2.347-68. Seoul, Korea.

Kim, Hyunsoon and Allard Jongman. 1996. Acoustic and perceptual evidence for
complete neutralization of manner of articulation in Korean. Journal of Phonetics
24.295-312.

Kim, Jong-Mi. 1986. Phonology and syntax of Koran morphology. Los Angeles:
University of Southern California dissertation.

Kim, Sun-Hee. 1989. The behavior of /h/ in Korean. Harvard studies in Korean
Linguistics 3.117-26.

Kim-Renaud, Young-Key. 1974. Korean Consonantal Phonology. Honolulu, HI:
University of Hawaii dissertation.

Kingston, John. 1994. Change and stability in the contrasts conveyed by consonant
releases. Phonological structure and phonetic form: Papers in laboratory
phonology III, ed. by Patricia Keating, 354-61. Cambridge: Cambridge University
Press.

Kingston, John and Mary E. Beckman (ed.). 1991. Papers in laboratory phonology 1:
Between the grammar and physics of speech. Cambridge: Cambridge University
Press.

Kiparsky, Paul. 1982a. Lexical morphology and phonology. Linguistics in the Morning
Calm, ed. by I. S. Yang, 3-91. Seoul: Hanshin Press.

Kiparsky, Paul. 1982b. Explanation in phonology. Dordrecht: Foris.

258

Kirchner, Robert. 1998. An Effort-Based Approach to Consonant Lenition.
Los Angeles, CA: UCLA Dissertation.

Kirchner, Robert. 2000. Geminate inalterability and lenition. Language 76.509-45.

Kirchner, Robert. 2001. Phonological contrast and articulatory effort. Segmental
phonology in Optimality Theory, ed. by Linda Lombardi. Cambridge: Cambridge
University Press.

Klatt, Dennis H. 1976. Linguistic uses of segmental duration in English: Acoustic and
perceptual evidence. Journal of Acoustical Society of America 59.5.1208-21.

Kohler, K. J. and W. A. van Dommelen. 1987. The effects of voice quality on the
perception of lenis/fortis stops. Journal of Phonetics 15.365-81.

Lacy, Paul de. 1998. Sympathetic stress. Amherst, MA: University of Massachusetts,
ms.

Ladefoged, Peter. 1971. Preliminaries to linguistic phonetics. Chicago: The University
of Chicago Press.

Ladefoged, Peter. 1972. Phonetic prerequisites for a distinctive feature theory. Papers in
linguistics and phonetics: To the memory of Pierre Delattre, ed. by Albert
Valdman and Editorial committee, 273-85. Netherlands: Mouton & Co, The
Hague.

Ladefoged, Peter. 1973. The features of larynx. Journal of Phonetics 1.74-83.

Ladefoged, Peter. 1975. A course in phonetics. New York: Harcourt, Brace and
Jovanovich.

Ladefoged, Peter. 1996. Elements of acoustic phonetics. 2nd edition. Chicago: The
University of Chicago Press.

Lass, Roger. 1976. English phonology and phonological theory. New York: Cambridge
University Press.

Lass, Roger. 1984. Phonology: An introduction to basic concepts. New York:

Cambridge University Press.

259

Leben, Williams R. 1973. Suprasegrnental phonology. Cambridge, MA: MIT
dissertation.

Lee, Ho Young. 1996. Korean phonetics. Seoul, Korea: Tae Hak Sa.

Lee, Ik sep and Sangak Lee. 1997. Korean language (in Korean). Seoul: Shin Ku
Publisher.

Lee, Minkyung. 1999. Max feature in Korean consonant reduction. Paper presented at
the annual meeting of the Michigan Linguistic Society, Michigan State University,
East Lansing, MI, October 30.

Lee, Sechang. 1994. Neutralization and Tensification in Korean. Japanese Korean '1
Linguistics 4, ed. by Noriko Akatsuka, 511-20. Stanford, CA: CSLI Publications.

Lee, Sechang. 1998. On the irregular behavior of h in Korean. Japanese Korean
Linguistics 7.585-93.

Liberman, Mark Y. and Alan Prince. 1977. On stress and linguistic rhythm. Linguistic
Inquiry 8.249-336.

Lin, Yen-Hwei. 1993. Degenerate affixes and templatic constraints: Rime change in
Chinese. Language 69.649-82.

Lin, Yen-Hwei. 1997. Syllabic and moraic structures in Piro. Phonology 14.403-436.

Lindau, Mona. 1984. Phonetic differences in glottalic consonants. Journal of Phonetics
12.147-55.

Lindblom, Bjorn, 1990. The status of phonetic gestures. Modularity and the motor
theory of speech perception, ed. by Ignatius G. Mattingly and Michael Studdert-
Kennedy, 7-24. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Lisker, Leigh and Arthur S. Abramson. 1964. A cross-language study of voicing in initial
stops: Acoustic measurements. Word 20.384-422.

Lombardi, Linda. 1990. The nonlinear organization of the affricate. Natural Language
and Linguistic Theory 8.375-426.

Lombardi, Linda. 1995. Laryngeal Neutralization and syllable wellformedness. Natural
Language and Linguistic Theory 13.39-74.

260

Lombardi, Linda. 2001. Segmental phonology in Optimality Theory. Cambridge:
Cambridge University Press.

Maddieson, Ian. 1984. Patterns of sounds. Cambridge: Cambridge University Press.

Makkai, Valeri Becker. 1972. Phonological theory: Evolution and current practice.
New York: Holt, Rinehart and Winston.

Malone, Joseph L. 1993. Tiberian Hebrew Phonology. p. 193. Winona Lake, Indiana:
Eisenbrauns.

Mattingly, Ignatius. 1990. The global character of phonetic gestures. Journal of
Phonetics 18.445-52.

McCarthy, John. 1988. Feature geometry and dependency: A review. Phonetica 45.84-
108.

McCarthy, John. 1998. Sympathy and phonological opacity. Amherst, MA:
University of Massachusetts, ms. ROA-252, Rutgers Optimality Archive,
http://roa.rutgers.edu/.

McCarthy, John and Alan Prince. 1993. Prosodic Morphology I. RUCCS Technical
Report 3. New Brunswick, NJ: Rutgers University Center for Cognitive Science.

McCarthy, John and Alan Prince. 1995. Faithfulness and reduplicative identity.
University of Massachusetts occasional papers in linguistics 18: Papers in
Optimality Theory, 249-384. Amherst, MA: University of Massachusetts,
Graduate Linguistic Student Association.

Mohanan, K. P. 1986. The theory of lexical phonology. Dordrecht: Reidel.

Nespor, Marina and Irene Vogel. 1986. Prosodic phonology. Dordrecht: Foris.

Ni Chiosain, Maire and Jaye Padgett. 1997. Markedness, segment realisation and
locality in spreading. Report no. LRC-97-01, Linguistics Research Center,
University of California, Santa Cruz. RDA-188, Rutgers Optimality Archive,
http://roa.rutgers.edu/.

Nolan, Francis. 1990. Who do phoneticians represent? Journal of Phonetics 18.453-64.

261

Oh, Mira. 1998. The prosodic analysis of intervocalic tense consonant lengthening in
Korean. Japanese Korean Linguistics 8, ed. by David J. Silva, 317-30. Stanford,
CA: CSLI Publications.

Oh, Saenae. 1999. Word medial dissimilation in Modern Korean. Seoul: Chung Ang
University thesis.

Ohala, John J. 1983. The origin of sound patterns in vocal tract constraints. The
production of speech, ed. by Peter F. MacNeilage, 189-216. New York: Springer-
Verlag.

Ohala, John J. 1990. There is no interface between phonology and phonetics: A personal
view. Journal of Phonetics 18.153-71.

Padgett, Jaye 1996. Partial class behavior and nasal place assimilation. Proceedings of
the Arizona Phonology Conference: Workshop on Feature in Optimality Theory.
Tucson, AZ: University of Arizona. ROA-l 13, Rutgers Optimality Archive,
http://roa.rutgers.edu/.

Paradis, Carole. 1988. On constraints and repair strategies. The Linguistic Review 6.71-
97.

Paradis, Carole. 1997. Non-transparent constraint effects in Gere: From cycles to
derivations. Derivations and constraints in phonology, ed. by Iggy Roca, 529-50.
Oxford: Oxford University Press.

Paradis, Carole and Jean-Francois Prunet, ed. 1991. The special status of coronals.
(Phonology and Phonetics 2.) San Diego, CA: Academic Press.

Park, Duk-Soo. 1990. Lexicon and syntax in Korean phonology. Honolulu, HI:
University of Hawaii dissertation.

Park, Ok-Sook. 1998. Sonorant assimilation within Correspondence Theory. Japanese
Korean Linguistics 8, ed. by David J. Silva, 345-59. Stanford, CA: CSLI
Publications.

Parker, Frank. 1980. The perceptual basis of consonant cluster reduction and final

devoicing. Journal of Phonetics 8.259-68.

262

Pierrehumbert, Janet. 1990. Phonological and phonetic representation. Journal of
Phonetics 18.375-94.

Prince, Alan and Paul Smolensky. 1993. Optimality theory: Constraint interaction in
generative grammar. RUCCS Technical Report 2. New Brunswick, NJ: Rutgers
University Center for Cognitive Science.

Raphael, Lawrence J. and Michael F. Dorman. 1980. Silence as a cue to the perception
of syllable-initial and syllable-final stop consonants. Journal of phonetics 8.269-
75.

Rhee, Seok-Chae. 1997. Constraints on post-obstruent tensification in Korean.
Japanese Korean Linguistics 6, ed. by Ho-min Sohn and John Haig, 253-74.
Stanford, CA: CSLI Publications.

Roca, Iggy (ed.). 1997. Derivations and constraints in phonology. Oxford: Clarendon
Press.

Rubach, Jerzy. 1994. Affricates as Strident Stops in Polish. Linguistic Inquiry 25.119-
43.

Rubach, Jerzy. 1997. Extrasyllabic consonants in Polish: Derivational Optimality
Theory. Derivations and constraints in phonology, ed. by Iggy Roca, 551-81.
Oxford: Oxford University Press.

Rubach, Jerzy and Geert Booij. 1990a. Syllable structure assignment in Polish.
Phonology 7.121-58.

Rubach, Jerzy and Geert E. Booij. 1990b. Edge of constituent phenomenon in Polish.
Natural Language and Linguistic Theory 8.427-63.

Sagey. Elizabeth C. 1986. The representation of features in non-linear phonology.
Cambridge, MA: MIT, dissertation. Published by Garland Press, New York,
1990.

Sawashima, Masayuki. 1971. Devoicing of vowels. Annual Bulletin, Research Institute

of Logopedics and Phoiatrics 5, 25-34. Tokyo: University of Tokyo.

263

Sawashima, Masayuki and Hajime Hirose. 1983. Laryngeal gestures in speech
production. The production of speech, ed. by Peter F. MacNeilage, 11-38.
New York: Springer-Verlag.

Sawashima, Masayuki, Hea Suk Park, Kiyoshi Honda, and Hajime Hirose. 1980.
Fiberscopic study on laryngeal adjustments for syllable-final applosives in
Korean. Annual bulletin, Research Institute of Logopedics and Phoniatrics 14. ‘
125-38. Tokyo: University of Tokyo.

Selkirk, Elisabeth. 1986. On derived domains in sentence phonology. Phonology
Yearbook 3.371-405.

Shadle, Chritine H. and Celia Scully. 1995. An articulatory-acoustic-aerodynamic
analysis of [s] in VCV sequences. Journal of Phonetics 23.53-66.

Shriberg, Lawrence D. and Raymond D. Kent. 1995. Clinical Phonetics. Needham
Heights, MA: Allyn & Bacon.

Silva, David James. 1993. A phonetically based analysis of [Voice] and [F ortis] in
Korean. Japanese Korean Linguistics 2, ed. by Patricia M. Clancy, 164-171.
Stanford, CA: CSLI Publications.

Silverman, Daniel. 1995. Phasing and recoverability. Los Angeles: UCLA dissertation.

Silverman, D and Jun, Jongho. 1994. Aerodynamic evidence for articulatory overlap in
Korean. Phonetica 5 1 .210-20.

Sohn, Hyang-Sook. 1987. Underspecification in Korean Phonology. Urbana-
Champagne, IL: University of Illinois at Urbana-Champagne dissertation.

Sproat, Richard and Osamu Fujimura. 1993. Allophonic variation in English /1/ and its
implications for phonetic implementation. Journal of Phonetics 21.291-311.

Steriade, Donca. 1987. Redundant values. Papers from the Twenty-third Regional
Meeting, Chicago Linguistic Society, 2. 33 9-62. Chicago: Chicago Linguistic
Society, University of Chicago.

Steriade, Donca. 1989. Affricates and stops. Paper presented at the Feature Conference

MIT, Cambridge, Mass

264

Steriade, Donca. 1991a. Gestures and autosegments: Comments on Browman and
Goldstein’s paper. Papers in laboratory phonology I: Between the grammar and
physics of speech, ed. by John Kingston and Mary E. Beckman, 382-97.
Cambridge: Cambridge University Press.

Steriade, Donca. 1991b. Closure, release and nasal contours. Los Angeles: UCLA, ms.

Steriade, Donca. 1993. Closure, release and nasal contours. Phonetics and phonology 5:
Nasals, nasalization and the velum, ed. by M. K. Huffman & R. A. Krakow, 401-
70. San Diego, CA: Academic Press.

Steriade, Donca. 1994. Complex onsets as single segments: The Mazateco pattern.
Perspectives in phonology, ed. by Jennifer Cole & Charles Kisseberth, 203-93.
Stanford, CA: CSLI Publications.

Steriade, Donca. 1997. Phonetics in phonology: The case of laryngeal neutralization.
Los Angeles: UCLA, ms.

Tatham, Mark. 1999. Articulatory Phonology, task dynamics and computational
adequacy. http://www.essex.ac.uk/speech/pubs/presents/aut-96/aut-96.html

Toebosch, Annenarie. 1999. A morphology sensitive OT analysis of voicing in
Dutch. Paper presented at Linguistics Colloquium, Michigan State University,
East Lansing, MI.

Van der Hulst, Harry and Norval Smith. 1982. An overview of autosegmental and
metrical phonology. The structure of phonological representations, ed. by Harry
Van der Hulst and Norval Smith, 1-45. Dordecht: Foris.

Walker, Rachel. 1999. Reinterpreting Transparency in Nasal Harmony. Los Angeles:
University of Southern California, ms. ROA-306-0399, Rutgers Optimality
Archive, http://roa.rutgers.edu/.

Woo, Nancy. 1969. Prosody and phonology. Cambridge, MA: MIT dissertation.

Yip, Moira. 1989. Feature geometry and cooccurrence restrictions. Phonology 6.349-74.

265

Yoshioka, Hirohide, Anders Lofqvist and Hajime Hirose. 1981. Laryngeal adjustments in
the production of consonant clusters and geminates in American English. Journal
of Acoustic Society of America 70.6.1615-23.

Zsiga, Elizabeth C. 1997. Features, gestures, and Igbo vowels: An approach to the
phonology-phonetics interface. Language 73.227-74.

266

 

     

l11111]lilililljjrrjljjr

930 72