An Optimality-Theoretic Formalization of Stress Pattern in
Seoul Korean
Seongyeon Ko
1. Introduction
2. Stress in Seoul Korean
2.1. Underlying vowel length distinction
2.2. The distribution of stress in Seoul Korean
2.3. A rule-based analysis
3. An Optimality-Theoretic Analysis
3.1. Stress assignment
3.2. Unstressed vowel shortening
3.3. Stressed light syllable lengthening
4. Conclusions
1. Introduction
This paper aims to provide a formalization of stress pattern in Seoul Korean within the
framework of Optimality Theory (Prince and Smolensky 1993). Lee (1989, 1999) asserts
that in Seoul Korean the position of stress in a word is predictable by syllable weight:
specifically, stress falls on the first syllable if it is heavy; otherwise, on the second
syllable. Based on this generalization, I will show what ranking of metrical constraints
accounts for the stress distribution in Seoul Korean. In addition to this static metrical
aspect, I will present an analysis on dynamic metrical processes such as unstressed vowel
shortening and stressed light syllable lengthening.
2. Stress in Seoul Korean
Stress in Seoul Korean has been controversial for a long time. Huh (1985) and Lee
(1989, 1999) argue that there is word-level stress, while Jun (1993) argues that Korean
stress is not a word-level but a phrase-level stress. However, at least they agree that
there exists something that we might call 'stress', that the stress falls on either the first
or the second syllable of the domain (though whether the domain is word or accentual
phrase is not agreed), and that syllable weight plays a crucial role in determining the
placement of stress.
In this paper, I will analyse the stress pattern described in Lee (1989, 1999). The data
of the present study are mainly from Standard Korean Pronouncing Dictionary (Lee 2002)
and other previous literatures (Huh 1985, Lee 1989, 1999, Park 1997).
1
2.1. Underlying vowel length distinction
In order to correctly describe the stress pattern in Seoul Korean we should consider the
vowel length distinction, which has been claimed to affects the distribution of stress.
Vowel length is a contrastive feature in Seoul Korean (Huh 1985, Lee 1989, 1999).
The minimal pairs which show this vowel length contrast are as shown below:
(1)
Short vowel
Long vowel
a. kúl
'oyster'
kúul
'cave'
b. nún
'eye'
núun
'snow'
c.
'horse'
máal
'speech'
d. pám
'night'
páam
'chestnut'
e. pə́l
'punishment'
pə́əl
'bee'
f.
'hand'
sóon
'descendant'
'pine tree'
sóol
'brush'
mál
són
g. sól
(2)
Short vowel
Long vowel
a. sicáŋ
'hunger'
sı́icaŋ
'market'
b. sakwá
'apple'
sáakwa
'apology'
Data given in (1) and (2) suggests that the vowel length distinction should be specified
in the underlying forms. From data given in (1), we cannot find any difference of stress
assignment. However, note that the minimal pairs in (2) show different stress placement:
if the first syllable has a short vowel it does not have stress, while if a long vowel it
does have stress.
2.2. The distribution of stress in Seoul Korean
As shown below, Lee (1989, 1999) claims that stress placement in Korean is highly
predictable from vowel length and syllable structure.
(3)
Stress patterns of Seoul Korean
a. In monosyllabic words, stress falls on the very syllable (Lee 1989:140).
b. In words of more than one syllable stress placement is initial when the first
syllable is a heavy syllable, i.e. one which either contains a long vowel or has a
syllable-final consonant. (Lee 1999:122)
c. All other words of more than one syllable are accented on the second syllable
(Lee 1999:122).
Each of these generalizations can be illustrated by the following example words (Lee
1999:122).
2
(4)
a. Accent on the first syllable:
(i) long vowel in the first syllable:
/káacaŋ/ 'disguise', /sə́əli/ 'acting head'
(ii) closed first syllable:
/sánsu/ 'landscape', /cʰúlku/ 'exit'
(iii) long vowel and closed firt syllable:
/sáansu/ 'arithmetic', /kámsa/ 'thanks'
b. Accent on the second syllable:
/kácaŋ/ 'most', /səlı́/ 'frost', /sadáli/ 'ladder'
Since Lee (1999) shows the stress patterns in disyllabic words, we need consider more
words as follows: monosyllabic words in (5), disyllabic words in (6), trisyllabic words in
(7), and words consisting of more than three syllables in (8). Each set of data is
illustrated with three columns: syllable weight, sample word, and gloss of the word.
(5)
(6)
(7)
monosyllabic words
Syllable Weight
Word
Gloss
a.
L
ı́
'tooth'
b.
H
ı́i
'two'
c.
H
mál
'horse'
d.
H
máal
'speech'
disyllabic words
Syllable Weight
Word
Gloss
a.
LL
sakwá
'apple'
b.
LH
sicáŋ
'hunger'
c.
HL
sáakwa
'apology'
d.
HH
sı́icaŋ
'market'
Syllable Weight
Word
Gloss
a.
LLL
satáli
'ladder'
b.
LLH
panɯ́cil
'needlework'
c.
LHL
catóŋcʰa
'car'
d.
LHH
putóŋsan
'real estate'
e.
HLL
póoŋsuŋa
'touch-me-not'
f.
HLH
cə́ŋkəcaŋ
'station'
g.
HHL
cı́ntallɛ
'azalea'
h.
HHH
cáŋnankʼam
'toy'
trisyllabic words
3
(8)
words consisting of more than three syllables
Syllable Weight
Word
Gloss
a.
LLLL
hɛpálaki
'sunflower'
b.
HLLL
sóksʼakita
'to whisper'
c.
LHHL
alɯ́mtaptʼa
'to be beautiful'
d.
LLLLL
pulə́tʼɯlita
'to break'
e.
HLLHL
sóoŋkusɯləptʼa
'to be sorry'
f.
HHHLL
kə́əcinmalcaŋi
'liar'
From these data, we can see that the generalization described by Lee (1989, 1999) is
correct.
2.3. A Rule-based analysis
Adopting the mechanism proposed in Hayes (1995), Park (1997) provides the following
rule for the stress patterns of Seoul Korean, just presented.
(9)
Set up an non-interative iambic foot at the left edge of a word (Closed
syllables
are heavy).
The stress rule in (9) accounts for most examples from (5) to (8). However, a
monosyllabic word with a light syllable like (5a) ı́ 'tooth' is remained unsolved, because
the foot form (L) does not belong to the universal inventory of iamb (Hayes 1995:71,
Kager 1999:147).
(10)
a. Syllabic trochee (quantity-insensitive): (σ́σ)
b. Moraic trochee (quantity-sensitive):
(LL) (H)
c. Iamb (quantity-sensitive):
(LL) (H) (LH)
Since Park (1997) claims that light monosyllabic words like (5a) form degenerate feet
(Hayes 1995), I rewrite his stress rule in (9) into the following three steps:
(11)
Step 1: Syllabify (open syllables with a short vowel are light, syllables with a
long vowel or closed syllables are heavy).
Setp 2: Assign a non-iterative iamb {(LH), (LL), (H)} at the left edge.
Setp 3: When the entire metrical domain is a single light syllable, assign (L) to it
(Weak prohibitions on degenerate feet: Hayes 1995:87).
This analysis is illustrated by the derivations in (12) below:
4
(12)
'tooth'
'horse'
'apple'
'apology'
'azalea'
'to be beautiful'
/i/
/mal/
/sakwa/
/saakwa/
/cintallɛ/
/alɯmtapta/
Step 1:
i
mal
sa.kwa
saa.kwa
cin.tal.lɛ
a.lɯm.tap.ta
Step 2:
˗˗˗
(mál)
(sa.kwá)
(sáa).kwa
(cı́n).tal.lɛ
(a.lɯ́m).tap.ta
Step 3:
(ı́)
(mál)
(sa.kwá)
(sáa).kwa
(cı́n).tal.lɛ
(a.lɯ́m).tap.ta
The serial derivation like this has some significant problems (see Kager 1999:149-150).
Where the Korean stress pattern be concerned, step 3 is problematic because it only
applies in light monosyllabic words like ı́ ('tooth'). It seems to be the last resort to
satisfy the culminative property - the imperative that every word should have a prosodic
peak.
3. An Optimality-Theoretic Analysis
In this section, instead of the rule-based analysis, I will present an OT analysis,
following Kager (1999, Ch4 pp142-193). First, I will show what constraints and constraint
ranking give us a predictive power about Korean stress assignment. Then, in 3.2 and
3.3, I will deal with some dynamic metrical phenomena in Seoul Korean, which have
drawn relatively little attention until now.
3.1. Stress assignment
3.1.1. Weight-By-Position
According to moraic theory (Hyman 1985, Hayes 1989), a short vowel (V) is represented
by one mora. On the other hand, a long vowel (VV), a short vowel plus a coda
consonant (VC), or a long vowel plus a coda consonant (VVC) is represented by two
moras.
(13)
a. Light syllables (one mora)
s
b. Heavy syllables (two moras)
σ
σ
σ
σ
σ
μ
μ
μ μ
μ μ
μ μ
o
i
'bull'
'tooth'
ɛ
n
'stream'
m
a
'horse'
l
m
a
l
'speech'
Whether a language analyses CVC syllables as light or heavy is one of the parametric
factor. In some languages a CVC syllable is regarded as a light syllable, while in others
5
it is regarded as a heavy syllable. The weight of a CVC syllable is determined
depending on whether its coda consonant is moraic or not. Language-specific conditions
on mora assignment is expressed as Weight-By-Position in OT.
(14)
Weight-By-Position (henceforth WBP)
Coda consonants are moraic.
If WBP is undominated, the language counts every CVC syllable as heavy. The stress
patterns discussed above suggest that Seoul Korean belongs to this kind of language.
3.1.2. Rhythm type and foot form
Hayes (1995) provides three basic bounded foot types: syllabic trochee, moraic trochee,
and iamb. In OT, Foot form is explained by the interaction of several constraints such
as RhType=I(or T), Ft-Bin, WSP.
(15)
a. RhType=I
Feet have final prominence.
b. RhType=T
Feet have initial prominence.
(16)
Ft-Bin
Feet are binary under moraic or syllabic analysis.
(17)
Weight-to-Stress-Principle (henceforth WSP)
Heavy syllables are stressed.
Let us consider a quadrisyllabic word such as (8a) hɛpálaki 'sunflower' which consists of
four light syllables (/LLLL/). In this word, stress falls on the second syllable. If we
adopt a trochaic analysis, the metrical shape of (8a) hɛpálaki is represented as L(LL)L.
However, this representation is opposed to the strong cross-linguistic preference,
so-called demarcative property of stress, that stresses fall at edges of the domain.
Since the stress always falls on the first or second syllable, we can assume that
All-Ft-Left, defined below, plays some role regardless of its hierarchical position in the
overall constraint ranking.
(18)
All-Ft-Left
Align (Ft, Left, PrWd, Left)
Every foot stands at the left edge of the PrWd.
In a trochaic analysis, even though RhType=T outranks All-Ft-Left, candidate (19b)
cannot be the optimal output. Instead, candidate (19a) which has stress on the first
6
syllable will be the winner.
(19)
input=/hɛpalaki/
RhType=T
All-Ft-L
*☞ a. (hɛ́.pa).la.ki
☜ b. hɛ.(pá.la).ki
c. hɛ.pa.(lá.ki)
*!
*!*
Even if we consider All-Ft-Right instead of All-Ft-Left as in (20) below, the situation is
still problematic because candidate (21c), instead of candidate (21b), is selected as
optimal.
(20)
All-Ft-Right
Align (Ft, Right, PrWd, Right)
Every foot stands at the right edge of the PrWd.
(21)
input=/hɛpalaki/
RhType=T
All-Ft-R
a. (hɛ́.pa).la.ki
☜ b. hɛ.(pá.la).ki
*☞ c. hɛ.pa.(lá.ki)
*!*
*!
In both tableaus above, candidate b. hɛ.(pá.la).ki cannot be chosen as an optimal output.
So, we can say that it is not easy to explain the Korean stress pattern fully and
plausibly in any trochaic analysis. To defend the trochaic analysis, we need another
constraint which bans the initial stress of (19a) or the final stress of (21c) and it should
be
ranked
higher
than
foot
alignment
constraints.
Constraints
which
encodes
extrametricality on the left or right edge could play such a role.
(22)
a. Extrametricality-Left
No prosodic head is initial in PrWd.
b. Extrametricality-Right (= NonFinality)
No prosodic head is final in PrWd.
(23)
input=/hɛpalaki/
a. (hɛ́.pa).la.ki
Extrametricality-L RhType=T All-Ft-L
*!
☞ b. hɛ.(pá.la).ki
*
c. hɛ.pa.(lá.ki)
**!
However, since stress falls on the first syllable whenever it is heavy, the prohibition
against stress assignment on left edge seems to be contradictory. Furthermore, if we
7
consider an LL pattern word such as (6a) sakwá 'apple', things would become more
complicated. In order to satisfy extrametricality-Left, the optimal candidate for the input
/sakwa/ would have a degenerate foot (L) as in sa.(kwá). In this case, the optimal
candidate L(L) would violate Ft-Bin and need more entangled constraint interaction.
In contrast with the trochaic analysis mentioned above, a iambic analysis gives us
simpler and easier way to explain the stress pattern of disyllabic and quadrisyllabic
words with only light syllables.
(24)
input=/sakwa/
R=I
Ft-Bin
☞ a. (sa.kwá)
b. (sá.kwa)
(25)
*!
input=/hɛpalaki/
R=I
All-Ft-L
☞ a. (hɛ.pá).la.ki
b. hɛ.(pa.lá).ki
*!
c. hɛ.pa.(la.kı́)
*!*
Note that RhType=I is ranked undominated until the supporting evidence for demoting
it, for I believe that rhythm type of a language seldom alters.
A tableau of a three-syllable word with only light syllables shows the sub-ranking
between Ft-Bin and Parse-Syl. Compare candidate (27a) and (27b) below:
(26)
Parse-Syl
Syllables are parsed by feet.
(27)
Ft-Bin ≫ Parse-Syl
input=/satali/
Ft-Bin
Parse-Syl
☞ a. (sa.tá).li
b. (sa.ta.lı́)
*
*!
Although candidate (27b) with an unbounded foot satisfies all the undominated
constraints such as Rh-Type=I, All-Ft-Left, GrWd=PrWd, it is excluded due to the
violation against Ft-Bin.
3.1.3. Culminativity and GrWd=PrWd
The culminative property is one of the most important cross-linguistic properties of
stress languages. It is directly encoded in OT as following constraint.
8
(28)
GrWd=PrWd
A grammtical word must be a prosodic word.
As we have seen in (5-8), there are no stressless words in Seoul Korean. Moreover, even
monosyllabic words with a light syllable, which has a short vowel and no coda
consonant, are assigned a stress. This is captured by ranking GrWd=PrWd over Ft-Bin.
(29)
GrWd=PrWd ≫ Ft-Bin
input=/i/
GW=PW
Ft-Bin
☞ a. (ı́)
*
b. i
*!
Candidate (29a) with an unpreferred degenerate foot is chosen as optimal output,
because the preference for grammatical words to be prosodic words is stronger than the
prohibition against bad foot forms. Note that, as Kager (1999:152) points out correctly,
this control of culminativity over foot well-formedness captures directly and precisely the
generalization missed by the rule-based analysis. Remember that, in the three-step rule,
step 3 only applies in light monosyllabic words.
(30) below shows the present ranking resulting from the above-mentioned constraints.
(30)
WBP, RhType=I, GrWd=PrWd ≫ Ft-Bin ≫ Parse-Syl
3.1.4. Non-iterative foot formation
Seoul Korean does not show rhythmic alternation of stresses and, consequently, there is
no secondary stress. Instead, It shows left single-sided non-iterative stress pattern.
Non-iterative
stress
pattern
is
captured
by
ranking
All-Ft-Left
(or
All-Ft-Right)
undominated.
First of all, All-Ft-Left dominates Parse-Syl as illustrated by the tableau below.
(31)
All-Ft-Left ≫ Parse-Syl
input=/hɛpalaki/
All-Ft-L
Parse-Syl
☞ a. (hɛ.pá).la.ki
b. (hɛ.pá).(la.kı̀)
**
*!*
Because All-Ft-Left is satisfied only in the case of a single foot standing at the absolute
left edge of the word, the additional foot in candidate (31b) incurs violations against
this constraint despite its contribution to satisfy Parse-Syl.
Also, All-Ft-Left dominates WSP. See the tableau below:
9
(32)
All-Ft-Left ≫ WSP
input=/panɯcil/
All-Ft-L
WSP
☞ a. (pa.nɯ́).cil
*
b. pa.(nɯ.cı́l)
*!
It seems that WSP is ranked relatively low, because we cannot find (H) form feet
anywhere except for in the heavy syllables at the left edge. And the distribution of the
(H) form feet can be fully explained by undominated All-Ft-Left and RhType=I.
However, the constraint demotion is minimized here unless there is a reliable evidence
which supports that WSP is dominated by another constraint.
Finally, All-Ft-Left dominates Uneven-Iamb.
(33)
Uneven-Iamb
(LH) ≻ (LL), (H)
(34)
(LH) is better iamb than (LL) or (H)
All-Ft-Left ≫ Uneven-Iamb
input=/cəŋkəcaŋ/
All-Ft-L
Uneven-Iamb
☞ a. (cə́ŋ).kə.caŋ
*
b. cəŋ.(kə.cáŋ)
*!
The constraint rankings discussed thus far may be summarized as follows:
(35)
WBP
RhType=I
GrWd=PrWd
≫
Ft-Bin
≫
WSP
Parse-Syl
Uneven-Iamb
All-Ft-Left
The position of Uneven-Iamb in the above ranking will be explained in the next section
3.2, where I will deal with the constraint Max-μ-IO.
With the ranking in (35), we can explain the stress distribution of all the words
given in (5-8). I present here representative three examples: one is a monosyllabic word,
another is a word with heavy first syllable, and a third is a word with light first
syllable. See appendix for the entire analyses on the words in (5-8).
(36)
A tableau for a monosyllabic word
input=/i/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
☞ a. (ı́)
b. i
WSP
Parse-S Uneven
yl
-Iamb
*
*!
*
*
10
(37)
A tableau for a word with a heavy first syllable
input=/cəŋkəcaŋ/ GrWd= RhType All-Ft-L Ft-Bin
PrWd
=I
☞ a. (cə́ŋ).kə.caŋ
Parse-S Uneven
yl
-Iamb
*
b. (cə́ŋ).(kə.cáŋ)
(38)
WSP
**
*!
*
*
A tableau for a word with a light first syllable
input=/hɛpalaki/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
Parse-S Uneven
yl
-Iamb
☞ a. (hɛ.pá).la.ki
**
b. (hɛ.pá).(la.kı́)
*!*
c. (hɛ.pa.la.kı́)
*!
d. hɛ.(pá.la).ki
*!
*
**
*
*
**
3.2. Unstressed vowel shortening
Compound words can have more than one stress in formal speech. However, as claimed
by H. Y. Lee (1996), in casual speech stress falls on only the first element of the
compounds and the vowel shortens. This can be illustrated as follows:
(39)
Simple words
/seekje/ → [(sée).kje]
/tɛɛcən/ → [(tɛ́ɛ).cən]
(40)
Compound words
a. Formal and slow speech
/seekje/+/tɛɛcən/ → [[(sée).kje]PrWd[(tɛ́ɛ).cən]PrWd]ProsodicPhrase
b. Fast and casual speech
/seekje/+/tɛɛcən/ → [(sée).kje.tɛ.cən]PrWd
In stressed syllables, underlying long vowels still have two moras in their output forms.
For this, we set up a sub-ranking between Max-μ-IO and Parse-Syl as follows:
(41)
Max-μ-IO
Input moras must have output correspondents.
(42)
Max-μ-IO ≫ Parse-Syl
a.
input=/seekje/
Max-μ-IO
☞ a. (sée).kje
b. (sé.kje)
Parse-Syl
*
*!
11
b.
input=/tɛɛcən/
Max-μ-IO
Parse-Syl
☞ a. (tɛ́ɛ).cən
*
b. (tɛ́.cən)
*!
On the other hand, in unstressed syllables underlying long vowels change into short
vowels, as shown in (40b). This is captured by an undominated constraint which bans
long vowels from appearing in unstressed syllables. (This must not be confused with
WSP which requires heavy syllables to be stressed.)
(43)
*
[ u n strVessed ]
Long vowels are prohibited in weak syllables.
(44)
*
[ u n strVessed ]
≫ Max-μ-IO
*
input=/seekje/+/tɛɛcən/
All-Ft-L
[ u n strVessed ]
☞ a. (sée).kje.tɛ.cən
Max-μ-IO
*
b. (sée).kje.tɛɛ.cən
*!
c. (sée).kje.(tɛ́ɛ).cən
*!*
In tableau (44), candidate a (sée).kje.tɛ.cən, of which the third syllable loses its weight,
violates Max-μ-IO. However, it is preferred to the other candidates because it satisfies
higher ranking constraint *
[ u n strVessed ] .
The following sub-ranking is a supplementary one, which tells us why Uneven-Iamb
should be ranked relatively low. Because Max-μ-IO is ranked higher than both Parse-Syl
and Uneven-Iamb as shown in (42) and (45) below respectively, we rank Uneven-Iamb
in the lowest position, together with Parse-Syl.
(45)
Max-μ-IO ≫ Uneven-Iamb
input=/siicaŋ/
Max-μ-IO
Uneven-Iamb
☞ a. (sı́i).caŋ
b. (si.cáŋ)
*
*!
3.3. Stressed light syllable lengthening
Lee (1989:134-137) states that stressed syllables are pronounced longer than unstressed
12
ones1), based on his experiment with laryngograph. According to the experiment, in the
case that tɛ́ɛməli ('bald head') is wrongly pronounced as [tɛmə́li] by younger generation
of Seoul Korean,2) the duration of the first syllable is 9 ms while that of the second
syllable is 23 ms. On the contrary, in the pronunciation by elderly native speakers of
Seoul Korean, the duration of the first syllable is 25 ms while that of the second is 16
ms. This indicates that if all the first two syllables are light, stress falls on the second
syllable and, as a result, the short vowel of the second stressed syllable is lengthened.
As fas as I know, this stressed light syllable lengthening has not been reflected in their
surface phonetic forms in any pronunciation dictionary. Also, as far as I know, it has
never been clearly formulated as a rule. Therefore, in order to argue that such a
lengthening process exist in the set of rules of Korean phonology, more supporting
evidences should be gathered from both descriptive and experimental phonetics.
However, I assume that the stressed light syllable lengthening works in Seoul Korean
and present here an OT analysis on it. Before than anything else, some representations
of the words given in (5-8) should be modified as follows:
(46)
Syllable Weight
Word
UR
SR
UR
SR
(6a)
/LL/
LH
/sakwa/
[sa.kwáa]
'apple'
(7a)
/LLL/
LHL
/satali/
[sa.táa.li]
'ladder'
(7b)
/LLH/
LHH
/panɯcil/
[pa.nɯ́ɯ.cil]
'needlework'
(8a)
/LLLL/
LHLL
/hɛpalaki/
[hɛ.páa.la.ki]
'sunflower'
LHLLL
/pulətʼɯlita/
[pu.lə́ə.tʼɯ.li.ta]
'to break'
(8d) /LLLLL/
Gloss
All these words have, in their underlying representations, a sequence of two light
syllables at the left edge which changes into a light-heavy sequence in their surface
representations. Note that this vowel lengthening occurs only when the first two
syllables are light.
The lengthening pattern of the stressed syllables is determined by the relative
ranking between Uneven-Iamb and Dep-μ-IO.
(47)
Dep-μ-IO
Output moras must have input correspondents.
(48)
input=/sakwa/
Uneven-Iamb
a. (sa.kwá)
*
b. (sa.kwáa)
Dep-μ-IO
*
1) Lee (1999:122) also states that an accented syllable is longer and louder than an unaccented
one.
2) Vowel length is no more a contrastive feature for them.
13
The tableau in (48) above tells us: if Uneven-Iamb ≫ Dep-μ-IO then lengthening occurs,
whereas if Dep-μ-IO ≫ Uneven-Iamb then lengthening is blocked.
Since I assume that the lengthening occurs in Seoul Korean, the ranking is as
follows:
(49)
Uneven-Iamb ≫ Dep-μ-IO
input=/sakwa/
Uneven-Iamb
a. (sa.kwá)
*!
☞ b. (sa.kwáa)
Dep-μ-IO
*
Though the short vowel in the second light syllable of candidate (49a) (sa.kwá) satisfies
the faithfulness constraint, it incurs a violation against Uneven-Iamb because the
resulting foot is not a canonical (LH). Therefore, it is worse than candidate (49b) (sa.kwá
a) which has a canonical (LH) and satisfies the higher ranking constraint.
In contrast to the lengthening in the underlying LL sequences which aims to make a
preferred iambic foot, the short vowels in light syllables of monosyllabic words are not
lengthened. This is captured by the following tableau:
(50)
input=/L/
Uneven-Iamb
☞ a. (L)
*
b. (H)
*
Dep-μ-IO
*!
In monosyllabic words, even candidate (50b) (H) of which the vowel is lengthened
cannot be chosen as optimal. Since the resulting foot form (H) is not a canonical iamb
(LH), it incurs a violation against Uneven-Iamb just as the faithful candidate does.
Thus, light monosyllabic words can be distinguished from heavy monosyllabic words
with long vowels in their surface forms. This prediction is well consistent with
underlying vowel length distinction given in 2.1.
Lengthening does not occur in words with the underlying LH sequence at the left
edge such as sicáŋ 'hunger', either. See the tableau below:
(51)
input=/CVCVC/ Uneven-Iamb
Dep-μ-IO
☞ a. (cv.cv́c)
b. (cv.cv́vc)
*!
The faithful candidate (51a) already has a canonical iamb, so it is selected as winner.
In order to lengthen the second vowel only when the underlying syllable weight
form at the left edge, a rule-based analysis will need the following additional step.
14
(52)
Step 4: Lengthen the vowel of strong light syllables.
This rule, however, provides no explanation of why it should apply only in strong light
syllables. In contrast, the OT analysis accounts for all the cases in (48-51) with a unified
interaction between the preference for a canonical iamb foot form (Uneven-Iamb) and the
preference for input-output faithfulness (Dep-μ-IO).
I finish this section with the following tableau:
(52)
input=/hɛpalaki/
RhType All-Ft-L Ft-Bin
=I
WSP
Parse-S Uneven Dep-μyl
-Iamb
IO
☞ a. (hɛ.páa).la.ki
**
b. (hɛ.pá).la.ki
**
c. (hɛ.pá).(la.kı́)
*!*
*
*!
**
4. Conclusions
In this paper we have formalized the stress pattern in Seoul Korean within the
framework of OT. The analysis presented here includes not only the static aspect of
metrical structure, i.e., stress distribution, but also the dynamic vowel length alternations,
i.e., unstressed vowel shortening and stressed light syllable lengthening. Because vowel length
alternations in Seoul Korean have not been formalized before, the OT analysis in this
paper will make a good contribution to future studies on them.
OT formalization does not need any ad-hoc device or rule for uncanonical cases such
as
(i) degenerate foot construction in light monosyllabic words or (ii) vowel lengthening
which occurs only in words whose first two syllables are light. Instead, it explains all
the patterns in a parallel mode by conflicting interaction between preferences and
dispreferences on the surface metrical forms. Therefore, we can conclude that OT
analysis accounts for the Seoul Korean stress pattern better than previous rule-based
analyses.
Instead of closing remark, I present the overall constraint ranking as follows:
(53)
WBP
RhType=I
Ft-Bin
GrWd=PrWd
≫
All-Ft-Left
*
[
V
u n str essed
]
≫
Max-μ-IO
(WSP)
15
Parse-Syl
Uneven-Iamb
≫
Dep-μ-IO
References
Hyman, Larry. 1985. A Theory of Phonological Weight. Dordrecht: Foris.
Hayes, Bruce. 1989. Compensatory Lengthening in Moraic Phonology. Linguistic Inquiry 20.
253-306.
Hayes, Bruce. 1995. Metrical Stress Theory: principles and case studies. Chicago: University of
Chicago Press.
Huh, Woong. 1983. Korean Phonology - today and tomorrow of our speech sounds. Seoul:
Saymmwunhwasa.
Jun, Sun-Ah. 1993. The Phonetics and Phonology of Korean Prosody. Ph.D. dissertation. The Ohio
State University.
Lee, Hyun Bok. 1989. Korean Standard Pronunciation. Seoul: Kyoyukkwahaksa.
Lee, Hyun Bok. 1999. Korean. Handbook of the International Phonetic Association. 120-123.
Cambridge: University Press.
Lee, Hyun Bok. 2002. Standard Korean Pronouncing Dictionary - Sounds, Accent & Rhythm -.
Seoul: Seoul National University Press.
Lee, Ho-Young. 1996. Korean Phonetics. Seoul: Thayhaksa.
Kager, René. 1999. Optimality Theory. Cambridge: Cambridge University Press.
Park, Hansang. 1997. A Formalization of Stress Pattern in Standard Korean. Malsori 33-34.
137-148. The Phonetic Society of Korea.
Prince, Alan and Paul Smolensky. 1993. Optimality Theory: constraint interaction in generative
grammar. Ms., Rutgers University, New Brunswick and University of Colorado,
Boulder.
16
Appendix
Tableaus for the examples in (5-8)
(5) a.
input=/L/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
☞ a. (L)
b. L
b-d.
input=/H/
WSP
Parse-S Uneven
yl
-Iamb
*
*
*!
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
Parse-S Uneven
yl
-Iamb
☞ a. (H)
b. H
(6) a.
input=/LL/
*
*!
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
*
Parse-S Uneven
yl
-Iamb
☞ a. (L.L)
*
b. (L).L
*!
c. (L.L)
d. LL
b.
input=/LH/
*
*
*!
*!
**
GrWd= RhType All-Ft-L Ft-Bin
PrWd
=I
WSP
Parse-S Uneven
yl
-Iamb
☞ a. (L.H)
b. L.(H)
*!
c. (L).H
d. LH
c.
input=/HL/
*!
*!
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
*
*
*
*
*
*
**
WSP
☞ a. (H).L
b. H.(L)
*!
c. (H.L)
d. HL
d.
input=/HH/
*
*
*
*
*
*
**
*!
*!
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
☞ a. (H).H
b. H.(H)
*!
c. (H.H)
d. HH
*
Parse-S Uneven
yl
-Iamb
WSP
Parse-S Uneven
yl
-Iamb
*
*
*
*
*
*
**
**
*!
*!
17
(7) a.
input=/LLL/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
☞ a. (L.L).L
b.
b. L.(L.L)
*!
c. L.(L).L
*!
d. (L.L.L)
*!
e. (L.L).L
*!
input=/LLH/
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
*
b. L.(L.H)
*!
c. (L.L).(H)
*!*
input=/LHL/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
e.
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
WSP
*!*
c. L.(H).H
*!
*!
*
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
Parse-S Uneven
yl
-Iamb
*
*
Parse-S Uneven
yl
-Iamb
Parse-S Uneven
yl
-Iamb
*
**
WSP
Parse-S Uneven
yl
-Iamb
☞ a. (H).H.L
WSP
Parse-S Uneven
yl
-Iamb
**
*!
c. H.(H).L
*!
18
*
*
WSP
*
b. (H).(H).L
*
**
*!
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
*
*
*
b. (H).(L.H)
input=/HHL/
*
*!
☞ a. (H).L.H
g.
*
**
b. (H).(L.L)
input=/HLH/
*
*
☞ a. (H).L.L
f.
**
**
*
b. (L.H).(H)
input=/HLL/
*
**
*!
☞ a. (L.H).H
d. L.(H.H)
*
*
b. L.(H).L
input=/LHH/
*
*
☞ a. (L.H).L
d.
*
*
☞ a. (L.L).H
c.
Parse-S Uneven
yl
-Iamb
*
Parse-S Uneven
yl
-Iamb
**
*
*
**
**
*
h.
input=/HHH/
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
☞ a. (H).H.H
**
**
*
*!
*
*
*
c. H.(H).H
*!
**
**
*
*
*
input=/LLLL/
*!
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
WSP
☞ a. (L.L).L.L
*!*
c. (L.L.L.L)
*!
d. L.(L.L).L
*!
input=/HLLL/
*
*
**
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
b. (H).(L.L).L
*!
c. H.(L.L).L
*!
input=/LHHL/
WSP
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
☞ a. (L.H).H.L
input=/LLLLL/
WSP
*
b. (L.H).(H).L
d.
*!*
WSP
☞ a. (L.L).L.L.L
b. (L).L.L.L.L
*!
c. (L.L).(L.L).L
e.
input=/HLLHL/
*!*
*!
GrWd= RhType All-Ft-L Ft-Bin
PrWd
=I
☞ a. (H).L.L.H.L
f.
input=/HHHLL/
WSP
*
b. (H).L.(L.H).L
Parse-S Uneven
yl
-Iamb
***
*
*
**
**
*
Parse-S Uneven
yl
-Iamb
**
*
GrWd= RhType
All-Ft-L Ft-Bin
PrWd
=I
d. (L.L).L.L.L
*
**
☞ a. (H).L.L.L
c.
Parse-S Uneven
yl
-Iamb
**
b. (L.L).(L.L)
b.
Parse-S Uneven
yl
-Iamb
b. (H).(H).H
d. (H.H).H
(8) a.
WSP
*!*
*
Parse-S Uneven
yl
-Iamb
***
*
****
*
*
*
***
*
Parse-S Uneven
yl
-Iamb
****
*
**
*
GrWd= RhType
Parse-S Uneven
All-Ft-L Ft-Bin WSP
PrWd
=I
yl
-Iamb
☞ a. (H).H.H.L.L
**
b. (H).(H).(H).(L.L)
*!*****
19
****
*
****