M t

_4__
_____7
——
“,7

 

'—+_._.
Igw
_cnooco

THE-Bis

I
200!

This is to certify that the

dissertation entitled

Evidence from Korean—Chinese Intrasentential
Codeswitching

presented by

Chun-Hua Ma

has been accepted towards fulfillment
of the requirements for

 

 

Masters degree in Linguistics

    

 

w Major professor

Date May 9, 2001

MS U is an Affirmative Action/Equal Opportunity Institution 042771

 

LlBRARY
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 wim earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/os"—c:/F'——"mcloateoue.uud-p.1s‘

TESTING THE MLF MODEL:

EVIDENCE FROM KOREAN-CHINESE INTRASENTENTIAL
CODESWITCHING

By

Chun-Hua Ma

A THESIS

Submitted to
Michigan State University
In Partial Fulfillment of the Requirements
For the Degree of

MASTER OF ARTS
Department of Linguistics and Languages

2001

ABSTRACT

TESTING THE MLF MODEL: EVIDENCE FROM KOREAN-CHINESE
INTRASENTENTIAL CODESWTICHING

By
Chun-Hua Ma
This paper presents empirical evidence to test the validity of Myers-
Scotton’s Matrix Language Framework Model by examining Korean-Chinese
intrasentential code switching (CS), a study of languages that are
morphosyntactically more highly disparate from each other than the ones she has
studied, such as English/Swahili, Shona/English, etc.. Naturally occurring CS
data from balanced Korean-Chinese bilinguals are presented to show that,
although the MLF model provides a nice account for the most commonly
occurring data, it fails to provide a satisfactory explanation for some of the
phenomena that have been revealed in language pairs with divergent syntactic
and morphological features. First, the MLF model fails to predict an asymmetry
in terms of the occurrence of switchable content morphemes, a phenomenon
abundantly exemplified in many bilingual CS examples. Second, the central
principles of the MLF model are in conflict with Myers-Scotton’s frequency-
based ML criterion. Since ML assignment functions as a cornerstone of the MLF
model, a more objective criterion of ML designation that is fully compatible with
major claims of the MLF model is required. Finally, the double morphology
data-in its expanded version-challenges the MLF model’s key assumption that
one language is dominant over the other. In conclusion, this study suggests that

the MLF model cannot be postulated as a universal in its current form.

ACKNOWLEDGEMENTS

I would like to express my wholehearted gratitude to Dr. Dennis Preston;
without his guidance and encouragement, I would never be able to complete this
thesis. His enthusiasm in the research as well as his patience with my slow
progress has showed me a perfect example of what is meant to be a good scholar
and professor. I would also like to thank Dr. Yen-hwei Lin and Dr. Barbara
Abbot for their insightful comments on my earlier draft and their willingness to
help and answer questions.

I also want to express special thanks to Mrs. LouAnn Wieland, whose
generous support has made possible my trip back to China to collect data for this
research work. Her unceasing prayers have accompanied every single step of
my journey toward completing my M.A. degree.

Lastly, I would like to thank my family, my former professors and
mentors at Yanbian University of Science and Technology, and my friends,
whose undying love and prayers have greatly strengthened me both spiritually

and emotionally.

iii

TABLE OF CONTENTS

List of Abbreviations

0. Introduction
1.1 Statement of Interest
1.2 Purpose of the Study

1.3 Outline of the Study

1. Theoretical Frame work

1.1 . Bilingual CodeSwitching

1.2. CodeSwitching and Borrowing

1.3. Myers-Scotton’s Matrix Language Frame Model
1.3.1 . Matrix Language vs. embedded language
1.3.2. Content morphemes vs. system morphemes
1.3.3. Principles of the MLF model

1.4. Morphosyntactic Features of Chinese

1.5 Morphosyntactic Features of Chinese

11. Korean—Chinese Bilingual Codeswitching
2.1. Bilingual Data
2.1.1. Description of the Bilingual Community
2.1.2. Subjects
2.1.3. Data Collection & Transcription
2.2. Data Analysis

2.2.1 . EL Nouns in ML+EL Constituents

iv

vi

13

14

17

17

17

20

20

21

21

2.2.2. EL Verbs in ML + EL Constituents
2.2.3. EL Adjectives in ML + EL Constituents

2.2.4. EL island

III. Problems

IV. Conclusions

V. Bibliography

23

24

27

29

46

48

ACC
ACC
ADJ
ADNZ
AGR
CL
CM
COMP
CONJ
CONSEC
COP
CS
DCL
I EL
EXT
GEN
HABIT
HON
INDIC
INFIN
INFL
MANN
ML
MLF
N
NEG
NOM
NOMZ

LIST OF ABBREVIATION

accusative
accusative case
Adjective
adnominalizer (suffix)
agreement

class

code-mixing
complementizer
cnjunctive (particle)
consecutive tense
copula
codeswitching
declarative (ending)
embedded Language
existential

genitive (particle)
habitual aspect
honorific (form)
indicative mood
infinitive

inflection

manner

matrix Language
matrix Language Frame
noun

negative

nominative

nominalizer

vi

OSV
OVS
PL

POST

PP
PRS
PST

RET

SOV
SVO
TOP

VOL
V08
V80

object + subject + verb
object + verb + subject
plural

postposition
prepositional phrase
present Tense (suffix)
past Tense (suffix)
question marker
retrospective Mood (suffix)
subject

subject + object + verb
sbject + verb + object
topic (particle)

verb

volitional Mood (suffix)
verb + object + subject

verb + subject + object

vii

0. Introduction

Over the past decades, increasing interest in syntactic aspects of
codeswitching has triggered a variety of investigations and theoretical
discussions that have shed light on our understanding of bilingual speech
behavior. Among the various analyses that have been proposed to account for
structural constraints of intrasentential codeswitching (hereafter CS) (Timm,
1975; Gumperz & Hernandez-Chavez, 1975; Lipski, 1978; Pfaff, 1979; Poplack,
1980; Sridhar and Sridhar, 1980; Sankoff & Poplack, 1981; Bentahila and
Davies, 1983; Woolford, 1983; Joshi, 1985; Di Sciullo et al., 1986; Pandit, 1990;
Myers-Scotton, 1993a; Belazi et al., 1994; Mahootian, 1996, etc.), Myers-
Scotton’s Matrix Language Frame (hereafter MLF) model has sparked much
research in the field and proven to be efficient on the basis of considerable
evidence from many different language pairs, such as English/Swahili,
Shona/English, Tamil/English, and Hausa/English, etc. However, a cross-
linguistic applicability of this model requires more empirical support from a
wider variety of languages. The motivation for this study primarily comes from
a concern with typological differences, particularly with respect to the
morphosyntactic features of the language pairs studied by Myers-Scotton and the
ones being examined in this thesis. It is well known that Chinese is a SVO
language with very little inflection, while Korean is a typical agglutinated SOV
language with a rather rich morphology. Therefore, with regard to syntactic
features, Korean and Chinese have quite different surface word order; while in

terms of morphological features, undeniably there is vast gulf in morphoogical

complexity between Chinese and Korean, which occupy two extremes
respectively in a continuum with various degrees of inflectional richness.
However, the languages examined by Myers-Scotton to support her MLF model
are either morphosyntatically very rich, such as Swahili, or at least showing
minimum inflectional features, such as English; she examines no language pairs
of which morphosyntactic features drastically differ from each other as Korean
and Chinese. In fact, she examines no language like Chinese. Therefore, testing
the MLF model based on such a language pair will, undoubtedly, provide

convincing evidence to its universal validity.

Another driving force for this study is that despite the fact that both
Korean and Chinese are major language groups and there are almost 2.5 million
Korean-Chinese bilingual population members currently residing in China, no CS
research to date, at least to my knowledge, has ever been done for Korean —
Chinese bilinguals. Therefore, adding Korean/Chinese bilingual codeswitching
to the current literature will broaden our understanding of some phenomena
unique to language pairs that share similar morphosyntactic properties with

Korean and Chinese.

In this paper, I will present data from Korean-Chinese bilingual utterances
to examine the claims pr0posed in Myers-Scotton’s MLF model. The
organization of the paper is as follows. The first part of Section 1 lays out some
terminological conventions, then a sketch of the MLF model proposed by Myers-
Scotton and its central principles will be reviewed, along with a comparison of

the major morphosyntactic features of Korean and Chinese relevant to current

study. In Section 2, I present Korean-Chinese codeswitching data that are
conform to the prediction of the MLF model. Section 3 discusses some
problematic data to show the inadequacy of the MLF model in accounting for
some phenomena that are uniquely revealed in bilingual speech with two
participating languages sharing different morphosyntactic features as that of

Korean and Chinese.

1. Theoretical Framework
Before getting into detail with regard to the theoretical framework for CS,
clarifying some terminology and concepts that are essential for understanding the

theory is in order.

1.1 Bilingual Codeswitching

CS is defined by Myers-Scotton as "the selection by bilinguals or
multilinguals of forms from an embedded variety (or varieties) in utterances of a
matrix variety during the same conversation” (1993, 3). The basic assumption
under this definition is that there evidently exists an asymmetry regarding the
participation of the languages involved. The one that is more activated and
functions as the principal language in CS is the matrix language; while the term
embedded languages refer to other participating language(s) inserted into the
matrix frame with less important roles. Although the definition of CS proposed
by researchers varies in its wording (Haugen, 1956; Diebold, 1963; Macnamara,

1967; Hymes, 1974; Scotton and Ury, 1975; Valdes Fallis, 1976; Di Pietro,

1977; Baetens Beardsmore, 1982; Grosjean, 1982; Appel and Muysken, 1987;
Lehiste, 1988; Poplack, 1993), they all roughly agree upon the nature of CS as
the alternation of two or more languages in a conversation. Much debate,
however, has been raised over the issue of what constitutes true bilingual in
terms of their language proficiency. Bloomfield (1933) and others insist only
native-like mastery of a second language other than one’s mother tongue can be a
considered as true bilingual; while others claim that minimal competency in at
least one language skill (speaking, writing, listening, or reading) should be
sufficient to be regarded as bilingual (Macnamara, 1967). These two criteria are
perceptibly of limited merit in that the former one is too restrictive, based on
which only few are eligible as true bilingual; the latter definition, however, is
over inclusive that it applies to everyone who has ever come in contact with
another language. For the purpose of this paper, I will adopt the view that
bilingualism has the characteristics that may exist in a continuum from minimal
competency to complete mastery of more than one language (Hornby, 1977).
However, the subjects to be considered here are individuals at the upper end of
the continuum. In other words, although they are not necessarily native like in
all their linguistic competences in both languages, they should be able to
communicate fluently in both (Loebell, 1989).

Another controversy surrounds the issue of what embedded language
(EL) material actually constitutes true CS, specifically. with regards to single-
lexeme items. Many scholars in the 19705 and far into the 19805 exclude singly

occurring lexemes from CS forms by claiming that only the sentences with EL

phrases or constituents should be considered CS forms (Ryes, 1976). Later some
researchers try to resolve the problem by labeling sentences with single lexeme

items as code-mixing (CM) (Kachru, 1978; Swigart, 1992). Relevant to current
study, we will adopt Myers-Scotton’s view and claim that not only can a phrase,

a clause or a sentence constitute CS, but also a single word.

1.2. Distinctions between CS and Borrowing

Including single lexeme items into CS demands researchers to make a
sound distinction between CS and borrowing, since they appear to be
indistinguishable on the surface. How to objectively draw a fine line between
CS and borrowing, however, has been quite a challenging task and raised much
controversy among the researchers.

In a very general sense, borrowed forms are identical to loanwords,
although some researchers further identify them into three different types:
loanwordsl, loan shifts2 and loan blends3 (Haugen, 1956). The traditional
criterion for distinguishing borrowing and CS is the degree of morphological,
syntactical and phonological integration into the base language (e.g., Weinreich,
1953; Haugen, 1956; Hasselmo, Mackey, 1970; Lance, 1975; Grosjean, 1982;
Lehiste, 1988; Poplack, 1993). Other researchers propose to make the distinction
based on the frequency of occurrence. For example, Myers-Scotton observes the

unpredictability and lower frequency of the CS forms, predictability and higher

' . Loanwords refer to the imported words from the other language with morphological and phonetic
integration with the base language.
2 . Loanshifts refer to the substituted items to represent a new concept or object of the donor language.

frequency of the borrowed forms. She specifies the frequency by adopting a
three-occurrence rule, which is that a borrowed form is the one that occurs in a
relatively large corpus at least three times (1993). However, the definition of
“relatively large corpus” is rather vague and arbitrary, which diminishes the
liability of this mechanism. In this study, because of the familiarity of the
researcher with the community where the data were collected, the distinction of
borrowed forms and CS forms will be mainly, but not exclusively, based on her
own judgment. Roughly speaking, regardless of the frequency of the occurrence,
if a word is a commonly used expression by monolingual speakers in the
community, then it will be considered as borrowed form instead of CS form.
Borrowed forms as well as proper nouns, such as names of persons, places, and

objects, are excluded as code switches.

1.3. The Matrix Language Frame Model

1.3.] Matrix language vs. embedded language

The MLF model differentiates participating languages in CS by
identifying one as the matrix language (hereafter ML) and the other(s) as the
embedded language(s) (hereafter EL). The ML is such labeled because it plays a
predominant role in language production and its grammar sets the
morphosyntactic frame for the sentences, while the EL refers to other languages
that have less important roles in CS. Since the distinction of ML vs. BL is the

basic principle structuring CS within the MLF model, designating the ML in CS

3 . Loanblends refer to the cases where only one part is morphemically imported from another language,
and the rest part still comes from the base language.

utterances becomes obviously crucial. While acknowledging the roles of
psycholinguistic and sociolingusitic criteria in identifying the ML, Myers-
Scotton proposes a frequency based criterion to objectively identify the ML:
“The ML is the language of more morphemes in interaction types including
intrasentential CS” (1993a: 68). Meyers-Scotton further points out that frequency
count should not be based on single sentences, with no consideration of a larger
corpus that contains such CS materials. Assignment of the participating
languages as either ML or EL allows us to categorize the intrasentential CS into
three different kinds of constituent:
1) ML islands--constituents consisting of only ML morphemes; must be
well-formed according to the ML grammar.
2) EL islands—constituents consisting of only EL morphemes; must be
well-formed according to the EL grammar.
3) ML+EL constituents—constituents consisting of morphemes from both
the ML and EL.
The following example of Swahili/English CS illustrates these three types of
constituents:

[1] Setting: several form 4 leavers (who have completed the equivalent of
high school) are talking about job opportunities in the government.

Ah si-vyo, kawaida hu-wa kwa gazeti. Kama last year
ah NEG-MANN usually HABIT-COP in papers[s] as

i-li-ku-w-a gazeti under public service
commission.

CL 9-PAST-INFIN-COP-INDIC paper[s]
Ma-jina i-li-to-lew-a tu hapo na mahali
CL 6-name CL 9-PAST-PLACE-PASS-INDIC just there and place

p-a ku-fanya interview

CL l6-of to-do

“Ah, no, usually it’s published in the papers. For example, last year it
was in the papers under Public Service Commission. The names are just
released [of persons to be interviewed] and the place of doing [the]
interview.”

(Swahili/English; Myers-Scotton 1993a: 81)

There are two EL islands (last year and under Public Service Commission) and
many ML islands. Mahali pa kufanya interview ‘place of to do interview’ is an

ML+EL constituent.

1.3.2. System morphemes vs. content morphemes

The MLF also distinguishes content morphemes and system morphemes
because patterns of occurrence of morphemes in bilingual codeswitchcd speech
are constrained by the status of morphemes. The primary feature for
differentiating these two types of morpheme is the feature [+/- thematic-role
receiver/assigner]. Content morphemes are either thematic-role receiver or
assigner, showing a plus setting for thematic-role. Most nouns are prototypical
thematic-role receivers, while most verbs and prepositions are prototypical

thematic-role assigners. In contrast, system morphemes lack the ability to either

assign or receive thematic-role, and most morphemes belonging to the functional
category, such as inflectional morphemes, are system morphemes.

A second feature that differentiates these two types of morpheme is the
feature [+/-quantification]. The feature [quantification] is defined by its property
of picking out individuals across variables, e.g., determiners specify particular
individuals; tense morphemes choose one specific time-frame. Quantifiers (e. g.
all, any no), determiners (e.g. the, a), possessive adjectives (e.g. my, your, his,
her), tense (tense markers), aspect (aspect markers), and any other category that
can appear at the specifier position of NP or quantificationally have the feature
of [+quantification], therefore, they are all system morphemes. Content

morphemes, on the other hand, show a minus setting for quantification.

1.3.3. Principles of the MLF model

Myers-Scotton proposes a set of interrelated hypotheses in the MLF
model. The first and the most fundamental hypothesis in this model is ML
Hypothesis. The ML Hypothesis states that the morphosyntactic frame, in other
words, the surface structure of the ML+EL constituents, is based on ML
grammar. Several principles further expand this hypothesis. The Morpheme-
order Principle claims that the morpheme order must follow ML morpheme
order; no violation should occur. For instance, an NP involving Swahili/English
CS is shown as in the following example from Myers-Scotton:

[2] ma-mbo m-engi new

CL6-things CL6-many

“many new things”

(Myers-Scotton: Swahili/English No.19)

Since Swahili is the ML, the English lexeme new follows its head mambo mengi

(things-many) according to Swahili syntactic restrictions.

The System Morpheme Principle states that all system morphemes will
come from the ML. Based on Myers-Scotton's criterion for categorizing system
morpheme as mentioned previously, quantifiers, possessives, tense/aspect,
determiners, copula, do verbs, possessive of, complementizers, structurally
assigned agreement and dummy pronominals all fall into the category of system
morpheme. Therefore, according to this principle, all these items should come

from the ML. Consider the following example:

[3] Second group-u] ceyil cwohahay.
second group-Acc. most like-DCL
“[I] like the second group best.”

(Korean/English; Park, Troike, and Mun, 1989: 9)

Note here, an accusative case-marker —ul from the ML Korean appears with the
English content morpheme.

The Blocking Hypothesis proposes that the ML blocks any EL content
morphemes that do not meet certain congruency conditions with ML content
morphemes. According to Myers-Scotton, two entities (linguistic category in

this case) are congruent if they correspond in respect of relevant qualities. For

example, the ML will block the occurrence of the EL content morpheme if it is

10

realized as a system morpheme in the ML. One specific example concerns
prepositions. Prepositions can be either content morphemes or system
morphemes cross linguistically. In the case of English prepositional phrases of
the form of for + NP, for is a content morpheme since it assigns the thematic role
to its complement NP. However, it does not have congruent status with the
corresponding Swahili morpheme, which is realized as a postpositional suffix;
therefore, for occurs when it heads an EL island, but may not occur in place of a

Swahili counterpart in an ML+EL constituent as supported by the following

example:
[4] Nikam wambia anipe ruhusa niende ni- ka- check for you
ls-CONSEC- check for you
“And I told him he should give me permission so htat I go and check for
you.”
but not:
[5] * Nikam wambia anipe ruhusa niende ni- ka- check for wewe

ls-CONSEC-check for you

“And I told him he should give me permission so that I go and check for

9’

you.
(Swahili/English; Myers-Scotton: 1993: 124)

The EL Island Trigger Hypothesis claims that the accidental accessing of
any EL morpheme that violates either the ML hypothesis or the Blocking
hylaothesis will keep the ML from being activated; therefore, an obligatory EL

iSland will occur. Consider the following example:

11

[6] Wache mimi nielekeee tauni, tukutane this evening at the usual place.
let us meet

“Let me go so that I may reach town, let's meet this evening at the usual
place.”

(Swahili/English; Myers-Scotton, 1993: 140)

Demonstratives have the feature [+ quantification] and therefore are system
morphemes. The unexpected activation of system morpheme this from the EL

forces an EL island to be constructed, here is this evening at the usual place.

The final hypothesis is the EL Implicational Hierarchy Hypothesis, which
proposes that those formulaic, idiomatic, and peripheral constituents that are on
the top of the hierarchy" have more tendencies to appear as EL islands. Most
time adverbials, especially two-word expressions or brief prepositional phrases
(PP) (e. g. next Saturday, every morning, on Saturdays), set expressions such as
old habits die hard, in fact, for personal purposes, and intensifier adverb + an
adjective (very fast, very late, very surprised) are functionally peripheral,
therefore, the EL islandhood will definitely favor these elements as supported

overwhelmingly by Mycrs-Scotton's Nairobi corpus:

[7] Wana some problems.

“They have some problems.”

[8] Ulikuwas ukiongea a lot of nonsense.

“You were talking a lot of nonsense.”

 

4 Myers-Scotton proposes the following Implicational Hierarchy: formulaic expressions and idioms >>
other time and mannar expressions >> quantifier expressions>> non-quantifier, non-time NPs as VP
complements >>Agent NPs >> Thematic role- and case assigners. (1993, p. 144)

12

[9] Hata siyo mwezi jana. Ilikuwa early this morning.

“Not even last month. It was early this morning.”

(Swahili/English; 1993:146-147)

To recapitulate, the MLF model identifies two crucial oppositions in
intrasentential CS: ML vs. EL and content morphemes vs. system morphemes.
Two central principles—Morpheme-order principle and System-morpheme
principle—along with a set of interrelated hypotheses proposed within the MLF
model. Given this account of the theoretical framework, in what follows, I will
examine the morphosyntactic features of the language pair being studied here so
that the readers will have a better understanding of the nature and significance of

the current research.

1.4 Morphosyntactic Features of Chinese

Roughly speaking, Chinese mainly has the features of an SVO language.
However, in terms of word-order typology, particularly in light of Greenberg's
word-order correlation, Mandarin Chinese basically has SVO features as well as
SOV features (Li & Thompson 1981). Prototypical SVO features are such that
head verbs precede the nouns, auxiliaries precede the verbs, and there are
prepositions, etc. However, Chinese also has postpositions, its relative clauses
and genitive phrases always precede the head noun, and the aspect markers
follow the verb, without exception. Morphologically, Chinese is well known for

its impoverished morphosyntactic properties (Chao, 1968; Chen, Tzeng, & Bates,

13

1990; Li, 1989). Some features that are quite common in many Indo-European
languages, such as verb conjugations and noun declension, are completely absent
in Chinese. Nor does Chinese have case markers to signal the grammatical
function the noun has in the sentence, such as subject, direct object, indirect
object, and so on. In the inflectional hierarchy of natural languages, Chinese
probably is one of the least inflected languages. The only markers it has are three
aspect morphemes:

-le “perfective”

-guo “experienced action”

-zhe “durative”

1.5 Morphosyntactic Features of Korean

Korean, however, is canonically an SOV language, with various degree of
other combinations as well, such as, VSO, VOS, OSV, and OVS, depending on
the discourse 0r pragmatic purpose of a speaker (Park 1990). Sentences normally
follow subject-object order with a verb or adjective always showing up at the end
of a sentence or a clause. Modifiers, such as determiners, adverbs, possessive
constructions, and relative clauses precede the modified elements, showing the
characteristics of left-branching languages. In terms of morphology, Korean is
well known as an agglutinative language and it has rather rich system of particles
that are all postpositional (Sohn 1994). As shown in the following example,
Korean noun inflection is realized through such postpositional particles to

represent case relationships (subject marker -- -ka/i; object marker -- ul/lu) and

14

discourse function (topic marker-- un/nun), as well as functions that are carried

out in other language by a preposition.

[10] Nae-ka tosesil -yese Yonghee-[u] mannassta.
I -Nom. library-at name-Acc. Meet

“1 met Yonghee in the library.”

Verb inflection in Korean has further complexity. A Korean verb consists of a
stem, and a sequence of inflectional suffixes. To categorize roughly, there are
two types of verb inflections: pre-wordfinal and word-final (Chang 1996). Pre-
wordfinal suffixes have various social and grammatical functions, including
honorifics (-[u]si), tense markers (-n[n], -[e/a]ss, ~ess), and mood indicators (-
keyss for Volition, -t[e/i/u] for Retrospective ). Word-final suffixes consist of six
different realizations based on their syntactic functions: sentence ending (-ta),
nominalizing-(rki, -m), adnominalizing (-un, n, l), adverbializing (-nikka, -myen,
ciman), conjoining (-ko, -na), auxiliary connecting (-ko, -ci). The following
examples show how verb inflectional endings work in Korean:

1. For the verb po “see”, the pre-wordfinal inflections with —ta as
wordfinal particle, can be as many as follow:

[11] p0- si- ta [12] po-si- n- ta
see-Hon-DCL see-Hon -PRS-DCL
“see (HON)” “seeing (HON)”

[13] p0- si- ess- (ess)-ta [l4] po-si-keyss-ta
see-Hon-PST- PST- DCL see-Hon-Vol-DCL
“saw (HON)” “will see (HON)”

15

2. The following examples show the various wordfinal inflections:

[15] po-ki;
see-NOMZ

“seeing”

[17] o- n- un salam;
come-PRS-ADNZ person

“the man who comes”

pwu-nikka;
blow-ADVZ

[19] palam-i
wind

“since the wind blows”

pwul-meyn
blow -ADVZ

[21] palam-i
wind

“if the wind blows”

[22] 0- ko;
come-CONJ

“come and..”

[24] 0- ko- iss ta
come-aux.con be DCL

“be coming”

[18] o- l-

[16] po-m

see-NOMZ
“to see”

salam
come-ADNZ person

“the man who will come”

[20] alam-i pwul-ciman
wind blow-ADVZ

“Although the wind blows”

[23] 0- na
come-CON]

“come or...”

[25] nwuw- e- iss ta
lie- aux.con be DCL

“be lying”

All the positions are optional except the word-final one. The adjective inflection
has almost the same sequence position, except certain features, such as no
present tense, no imperative or propositive ending and is unable to form the

progressive/perfective aspect with <-ko/e + issta>, etc.

16

In sum, with regard to morphsyntactic features, the dissimilarity between
Korean and Chinese is quite significant. The following comparison chart

provides a summary of two languages:

 

 

Table 1.
Korean Chinese
SOV SVO
N+case marker No case marker
N+ existential V Existential V + N

Complement+copula Copula + complement

N + postposition Preposition + N
Obligatory sentence- Non-obligatory sentence-
ending particle ending particle

Neg.+ V or V + Neg. Neg. + V

Modifier + N/Adj. Same

 

Although the above illustration is not exhaustive, it suffices to suggest
that an investigation of Korean and Chinese bilingual CS may reveal some new

insight on syntactic constraints of intrasentential CS.

II. Korean-Chinese Bilingual Codeswitching

2. 1. Background Information for the Data

21.1.anme

Data come from free conversations audio-recorded in a Korean/Chinese

bilingual community---YanJi, a small city located in the northeast part of China.

17

China is well known as a multi-ethnic country with 56 different ethnic groups.
The Korean community in YanJ i is one of the ten biggest minority independent
prefectures along with Tibetan, WeiWuEr, DaLi and many other ethnic groups
within Chinese territory. The population of YanJi is approximately forty percent
Koreans and sixty percent other ethnic groups, mainly Han Chinese population.
The recent statistics, however, show that the Korean population in the area is
rapidly dropping as China gradually opens its door to the world, and the old
notion of sticking to where their ancestors used to live is breaking down. Better
salaried working opportunities and living conditions in major cities within China
as well as in Korea and other developed countries attract talented young
generations to step out of this small community and jump into world
globalization.

The minority nationality autonomous regions allow the use of commonly
employed languages in carrying out local affairs. Accordingly in Yanbian,
where the Korean language and writing system are widely used alongside the
Chinese language and writing system, both languages are used in documents and
announcements of government meetings, in business and commerce, and in
describing titles and writing authorizations to technicians, and workers. The
signs and seals of government, business, and commercial agencies uniformly
appear in Chinese and Korean, and in all matters pertaining to court and police
affairs, both Korean and Chinese are used, allowing the people to communicate
in their own language. Similarly students are given options to be educated and

tested in their own language. China has continued to emphasize the use of

18

national languages in teaching in primary and middle schools. There are two
types of schools in YanJi: one is a Korean school and the other one is a Chinese
school. In Korean schools, Korean is the main language for teaching and
Chinese is taught as a subject. Since students learn Chinese from elementary
school, plus considerable influence from the public media, which is mainly
broadcast in Chinese, most students who go to Korean schools can speak Chinese
very well. And of course, for those students educated in Chinese schools, they
can speak fluent Korean mainly due to the fact that Korean is their mother
tongue, which they have been speaking before entering to the elementary
schools, and it continually functions as the main tool for communication in the
family, particularly with the older generations. Noticeably, minority policy in

. various social dimension including languages provides an answer for the
puzzling question: how come most Korean-Chinese can speak Korean so well
after living in China for four or five generations?—an observation that often
amazes people from outside communities, particularly Korean-Americans, who
are concerned about their second generation will eventually lose their own
cultural identity along with the phenomenon of language loss.

In addition to the broad picture of a successful story of Chinese minority
policy, some other elements at the micro level may attribute to well-maintained
Korean as well. In particular, Koreans, as a nation, are proud of the pureness of
their blood heritage, therefore, everyone senses this invisible solemn mission of
keeping their blood pure, especially Korean men. Living and being educated in

China for generations haven’t really changed people’s way of thinking in this

19

line; attitudes against interracial marriage are still commonly observed among
Korean-Chinese. As a result, Koreans end up marrying their Korean fellows,
with a few extremely rare exceptions. The absence of this mixture with the
mainstream group helped Korean-Chinese keep their language and culture as it is

HOW.

2.1.2. Subjects

Taking into consideration the fundamental purpose of this study, which is
to examine the syntactic constraints on CS instead of the social motivation, data
were collected mainly from a homogeneous group with similar educational
background, language proficiency, age group, social status, etc. All speakers
from these conversations were educated in Chinese schools all the way from
elementary school to high school. Chinese is the most frequently used language
at their workplaces, while they speak Korean at home to their kids, spouses, and
older generations. Apparently, there is little room for questioningtheir status as

balanced bilingual speakers.

2.1.3.2atafiollectinnfilranscrimion

Data collection was undertaken in rather relaxing and natural ways since
the subjects are close acquaintances of the researcher. Prior to recording, the
subjects were told that the purpose of the recording was to analyze the local
color of the language spoken in this area, and the speakers' permission to use the

recorded material in an anonymous fashion was obtained. The conversations

20

vary in length from about thirty minutes to over an hour, and the number of
participants in each conversation varies as well from two to four people. The
data used in this study represent approximately 15 hours of naturally occurring
free conversation.

Yale transcription system was adopted to transcribe the Korean data,
while using romanization pinyin system for Chinese data. Borrowed forms and

intersentential codeswitching data were excluded from the consideration.

2.2. Data Analysis
Since the current research is to test MLF model with Korean Chinese
bilingual data, a good starting point will be to examine those data that can be
nicely accounted for with the principles hypothesized under the MLF model.
The main types of data to be analyzed are ML + EL constituents and EL islands.
Accor'ding‘to Myers-Scotton's feature differentiation of content morpheme
and system morpheme, the most prototypical content morphemes that are highly
likely to show up in ML + EL constituents are nouns, verbs, adjectives,
pronouns, and some prepositions. Numerous examples from my Korean-Chinese
corpus confirm that only these content morphemes are inserted into ML + EL

constituents as predicted by Morpheme Order Principle.

2.2.1. .ELNounainMLiELLonsmnents

As has been observed from various language pairs in the CS literature, the

relative frequency of nouns to appear as EL morphemes is remarkably high.

21

Data from my Korean-Chinese also supports this claim and also confirm the
prediction of the Morpheme-order and System Morpheme Principles. Some
examples are shown below:

[27] Two young women are talking about some apartment certificate related
issues:

...... ku saram-tul -un fayuan ka -se ku dizhang chaca-se
those people-PL-Top court go to-and that backup file found-and

“those people went to the court and found those backup files ..... ”
[28] A young woman is telling her friend the recipe of a Chinese dish:

Jianjiao kkaci neh-myen te mas—iss-ta.
pepper also put-if more tasty-Exit.-DCL

“It tastes even better if you put hot pepper also.”

[29] ...ku saram-un cikum fendian hana kyesok han-ta.
that guy -—Top now branch store one continually running-DCL

“That guy is still running a branch store now.....”

[30] ..... na-nun gongchengbu _-eyse wass-nun-ka haess-ta.
I -Top Dept. of construCtion-from come thought-DCL

“I thought he was from the dept. of construction....”

Notice here that, the NPs ku dizhang “those backup files,” fayuan “the court”
and jianjiao “hot pepper” are VP complements and unmistakenly follow the
structural order of Korean, in which the complement always precedes the verb.
Also, in example [29], the NP fendian hana “one branch office” is a typical ML
+ EL constituent that consists of a head noun plus numerals. Again, the order is

exactly that of Korean instead of Chinese, in which the order is numeral + head

22

noun. Example [30] with a Preposition + NP constituent gongchengbu-ey-se
“from the dept. of construction” again shows the Korean order in which the noun
precedes the postpositional particles. It is obvious that Chinese as the BL is

inserting its content morphemes into the ML, Korean here.

2.2.2. EmenMLLELQQnstimms

Verbs are the second largest category which appear as embedded items in
my Korean- Chinese bilingual corpus, especially in the CS forms with Korean as
ML. Because Korean is an agglutinative language, the embedded Chinese
embedded verbs have to be accompanied by a suffix, -ha support, which is
similar to Do-support in English. Altogether there are more than seventy verb
forms with a Chinese verb stem and the Korean suffix —ha. Below are some of

the examples of BL verbs in CS forms:

[31] A casual conversation between mother and son at the dinner table.

can] yanjitai -eyse hankwuk yenghwa-lul fangying —ha-tu-la.
today YanJi channel-on Korean movie -ACC. show -do support-DCL.

“There was a Korean movie showing on Yanj i Channel.”

[32] Two young female friends are talking about the complicated legal procedure
of interracial marriage.

caki cikcep miandui -ha-ni-kkan pappun-kes-kathae....
self directly face -do support-because hard -seem

“It seems hard since I am currently facing it ...”
[33] A college student is telling her friend how they take notes in her school.

sensayngnim-i jlangke —ha-nun-ke chem -pwuthe kkuth-kac motwu

23

teacher -Nom teaching —do support-thing beginning from end- to all

chao-han-ta.
copy—do-DCL

“. ...we copy down all that the teacher teaches, from the beginning to the end.”

Notice here, the ML, Korean, visibly sets the morpheme order of the entire
sentence, and the Chinese verbs are inserted into the Korean sentences almost
exclusively with ha-support.

One interesting observation in my Korean-Chinese corpus is that Korean,
in most cases, functions as the ML, setting the structural frame of the CS
sentences, and in those few cases of Chinese as the ML, no Korean verbs appear
in a CS sentence. Similar asymmetry has also been noted in other bilingual data
(Stolt, 1964; Park 1990). Since Myers-Scotton's MLF model does not block the
appearance of verbs in any way, we need further explanation for this asymmetry.
We will come back to this point later in Section 3, while dealing with the

problems of the MLF model.

2.2.3. ELAdjseuxeunMLiELmstitnents

Based on the criterion for content morphemes under Myers-Scotton’s MLF
model, most descriptive adjectives are potentially thematic-role assigners;
therefore, they are also content morpheme, which predicts their occurrence as EL
items in CS. My Korean-Chinese corpus supports this prediction, as shown in

the following examples:

24

[34] A free conversation among three friends.

na mankhum pusu -han saram-to tumwul-ta.
I like frugal-ha-support person -also rare-DCL

“.....Someone as frugal as me is also very rare.”

[35] .....ku cip sikkwu-tul-un cinay reqing -han cengto-nun ani-tu-la.
That family —PL-Top very warmhearted-ha Sup. Degree-Top not DCL.

“That family is not that warmhearted.”

[36] umsik-un yeki hwelssin te xinxian -hay-yo.
Food -Top here much more fresh - ha Supp. -DCL

“As for food, here (refer to Yanji) is much fresher.”

Again, the insertion of BL Chinese adjectives pusu “frugal,” reqing
“warmhearted,” and xinxian “fresh” is realized with ha —support, which is also a
very common inflectional marker for Korean adjectives. In other words, when
Chinese EL adjectives enter into CS sentences, they only substitute the stem of
the words; suffixes will remain as usual. However, there is one exception, which
is labeled by Myers-Scotton as bare form:

[37] ne-nye pwumonim-un cengmal xingfu.
Your parents -Top really happy

“Your parents are really happy ......

Here, the EL Chinese adjective xingfu “happy” appears as a bare form without
the Korean inflectional marker ha-ta. It is rather obvious that Korean is the ML
of this sentence with the Korean system morphemes —un (Topical marker) and
cengmal “really,” and the fact that the Korean adjective suffix fails to show up

could be problematic. However, since this is the only example out of the whole

25

Korean-Chinese bilingual corpus, it can be treated as a production error as
suggested by Myers-Scotton.

Possessive adjectives such as my, your, his and her are system
morphemes; therefore, the occurrence of such adjectives as EL items should be
blocked. My data also confirm this prediction with no EL possessives occurring
in CS sentences.

[38] xiang ni o ppa name aide hua zenmebana!
Like your brother that short —if what to do
“What should we do if (he) is as short as your brother?”
The possible counterexample would be something like this:

[39] * xiang ne-nye gege name ai de hua zenmebana!
Your brother

But my data show no such examples. The only occasion of such possessive
adjectives appearing as EL material is along with noun phrases (NP) to form an
EL island, which rather convincingly supports the hypothesis of the MLF model,
which states that the accidental entering of system morphemes of the EL will
result in an obligatory EL island. Consider the following examples:
[40] ta jiejie -ka sicengpu-ay-se il-han-ta.

Her sister-Nom city hall-at working-DCL

“Her sister is working at the City hall.”

[41] ne enni duo xiwang wo he shui ya!
Your sister how much hope I drink water-DCL

“How much does your sister hope I drink just water!”

26

Notice here, the accidentally appearing EL system morphemes ta “her” and ne
“your” force the whole phrases to be completed in that language; [40] has a
Chinese EL island and in [41] a Korean EL island. This prediction exactly
conforms to the EL Island Hypothesis of MLF model. However, due to the
absence of the topical/nominative case marker for the subject ne enni “your
sister”, which can be dropped in spoken language, sentence [40] can also be
argued to have Korean as ML with the strings of word duo xiwang wo he shui ya
as an EL island. This is one of the criticisms of the MLF model, which we will

come back to it later.

2.2.4. ELJsland

The EL Island Trigger Hypothesis and the EL Implicational Hierarchy
Hypothesis of the MLF model are also well-supported in the Korean-Chinese
corpus. Look at the following examples:

[42] selnal nokum-ha-myen tai za le.
New year’s day recording —do sup. -—if too noisy DCL

“if (you) record on New Year’s Day, then it’s too noisy.”

Adverbs are neither thematic-assigncr nor thematic-receiver; therefore, adverbs
are categorized as system morphemes. Example [42] shows that the EL adverb
tai “too” accidentally enter into the sentence, triggers the island and forces the

rest of the sentence to be completed in Chinese.

27

[43] nay ku-ke mospon-kye weida de shiwu a!
I that-thing not see- fact great pos. mistake DCL

“The fact I didn’t get to see that scene is really a great mistake!”

[44] ku cwung -ye to xiangduilaishuo coh-un saram-to iss-ta.
Those among-Post. also relatively speaking good person-also have-DCL

“Relatively speaking, there are also some good guys among them.”

Here, in [43] the Chinese two-word expression weida de shiwu “great mistake” is
commonly used as a fixed expression with sarcastic rather than literal meaning;
therefore, it again results in an EL island, indicated by the sentence final ending
particle a, an exclamation marker. In [44] the formulaic constituent
xiangduilaishuo “relatively speaking” also forms an island as predicted by the
EL Implicational Hierarchy Hypothesis.
[45] ka-nye mwusun cengsin-ye yizhengtian kongpwu-ha-kyess-ni?

They what mind -Post. all day study- ha —supp-.Q

“How can they study all day long?”
[46] wenming cemsim-ye dou chaocai zuotang.

Wenming at noon always stir frying making soup

“At noon, Wenming always stir frying and making soups.”
The time adverb yizhengtian “all day long” in [45] and brief PP cemsim-ye “at

noon” [46] are all functionally peripheral; therefore, the EL island is favored by

these expressions.

28

To summarize, as proven by the data from many other language pairs,
majority of my Korean-Chinese bilingual data can be accounted for under the
central principles proposed by Myers-Scotton’s MLF model. However, as it is
always the case, some problematic data also exist and have caught our attention
to dig further. In what follows, I will provide a detailed analysis of such data to
point out inadequacy of the MLF model by locating some blind spots neglected

by Myers-Scotton.

111. Problems

First of all, a close observation of my Korean-Chinese corpus following
Myers-Scotton’s criterion for ML reveals a clear asymmetry — the occurrence of
ML + EL constituents with Korean as ML and Chinese EL predominantly
outnumbers those of Chinese as ML and Korean EL. In particular, as briefly
mentioned previously, the Chinese NP + Korean Verb construction is abundant in
my Korean-Chinese corpus, while the Chinese Verb + Korean NP construction
rarely, if ever, occurs. The appearance of Chinese verbs normally triggers EL
islands, forcing the rest of the sentence to be completed in Chinese. Consider the

following examples:

[47] xuexiao-ye ka-ta. Vs. [48] qu hakkyo
School-Post. go-DCL go school
“go to school” “go to school”
[49] xinyifu -lul sa-ta. Vs. [50] mai saeos
new clothes-Ace. Buy-DCL buy new clothes

29

“buy new clothes” “buy new clothes”

The structure in [47] and [49] frequently occurs in Korean-Chinese bilingual
speech compared to that in [48] and [50].
The occurrence of Chinese EL in a sentence is always accompanied by ha-

ta (do-support in Korean) as shown in following example:

[51] ku-ka hong qunzi-lul chuan-haess-ta.
She-Nom. red dress —Acc. wear —do-support-DCL

“She wore a red dress.”

Because the MLF model is based on the nature of morphemes, with no
typological consideration of the two participating languages, it predicts no such ‘
asymmetry at all. For prototypical content morphemes such as verbs and nouns,
nothing should block the appearance of Korean NPs as the complements of
Chinese verbs, nor should the occurrence of a single Chinese verb in a Korean
NP + Chinese V construction be outlawed. Similar asymmetry was observed in
several other studies as well. In one of the earliest studies on the syntax of
bilingual speech, Stolt (1964) shows that Latin nominal and verbal elements can
freely be embedded into German-based sentences, while German insertions into
Latin sentences are rare and restricted in type. More recently, in a study of
Korean/English bilingual codeswitching, Park (1990) observes that only in
Korean-based sentences is code-switching acceptable, while switching from

English to Korean is quite limited. This observation is also supported by the

30

interview data he conducted among bilingual graduate students at the University

of Illinois at Urbana-Champaign. The MLF model appears to be inadequate when
language specific features of two languages with opposite typological values are

examined.

Secondly, because the MLF model argues that one of the participating
languages is dominant over the other and sets the morphosyntactic framework for
the bilingual speech, identifying the ML is apparently crucial. According to
Myers-Scotton, the ML is “the language of more morphemes in interaction types
including intrasentential codeswitching” (1993, 68). In other words, the ML is
the language with higher morpheme frequency in a codeswitching utterance.
Myers-Scotton (1995, 237) further stressed the importance of this ML criterion
by saying that it should be preferred over any structural considerations because
the triggering facts of codeswitching are more sociolingusitic than structural, and
the ML is the sociolinguistically unmarked language across the community as
well as in a given conversational unit. If frequency does play such a pivotal role

in determining the ML, then the following examples will pose a further problem

for the MLF model:
[52] hanam cip -un yijinqu jiushi keting.
Hanam house- Top once get in then is living room.

“That Hanam house, once you get in, then there is a living room.”

[53] ka-nun meiyou wenping.
He- Top no have diploma

“He doesn’t have a (college) diploma.”

31

[54] fuqi -kke-nun ruguo huan ming de hua buyong hua shouxufei.
Husband& wife‘s -—Top if change nam no need pay service charge.

“For husband and wife, if you change the names, then you don’t need to pay
any additional service charge.”

[55] shengiiang-un wei hai keyi, jiaozhe bu haochi.
Ginger -Top smell still ok chew not tasty

“Ginger smells ok, but tastes bad if (you) chew it.”

[56] sejip-i jiu zheme da yipian .....
rented house-Nom only this big piece

“The rented house is only this big.”

[57] sunok-i-do dou guole nage nianling le.
Name-Nom-also already pass-Past that age DCL

“SunOk also has already passed that age.”

[58] ziji ban xuesheng po-ta Ire wai de xuesheng gengduo.
Self class student than extracurricular DE student more

“(He) has more extracurricular students than his own in-class students

Based on Myers-Scotton’s ML principle and system morpheme principle, the
topicalization marker —(n)un, nominative case marker -i, postposition —pota,
adverb -to, etc. are all typical system morphemes; therefore, the ML for these
examples should be Korean, and the following Chinese sentences are no more
than long EL islands. However, in terms of morpheme frequency, this is
definitely not the case. Furthermore, it is quite common in spoken Korean to
drop the case markers, especially topic markers and nominative case markers. In
the following two examples, we might assume that Korean is the ML, as

predicted by the MLF model:

32

[59] ka zhuangbuzhu shi.
she keep-not-in things

“She cannot keep anything in her mind.”

[60] wuli yiyuefen kaishi bande.
we January start do

“We started from January.”

However, this is not likely the case. In these two examples, it is more reasonable
to treat the sentence as the Chinese ML with Korean insertion. In this case are
two pronouns -ka “she” and wuli “we.” These two groups of sentences cast
some doubt on the status of Korean as the ML in examples [52] ~ [5 8], as
predicted by the MLF model.

Similar data occurred in Korean/English bilingual data (Park, 1990) and
Japanese/English data (Nishimura 1985; Morimoto 1999), as shown below:

[61] What do you call this statue wa?
Top

[62] I don’t know the bus stop no name.
GEN
“I don’t know the name of the bus stop.”

(Japanese/English; Morimoto 1999)

[63] Camp-seikatsu ga made him rough.
Life Nom.

“(That) camp-life made him rough.”

(J apanese/English; Nishimura 1985)

33

[64] Sometimes wuli-ga can’t know it.
We-Nom.

“sometimes, we can’t know it.”

(Korean/English; Park 1990)

The above examples undoubtedly indicate that the occurrence of such data is not
just a unique phenomenon in the Korean/Chinese bilingual corpus, and it
certainly deserves more attention and a more unified explanation in terms of the
ML assignment since two competing forces, the frequency-based criterion and
system morpheme principle, are at odds with one another.

Myers-Scotton might argue, at this point, that her ML Criterion is based
on a discourse sample rather than a single sentence. However, beyond noting that
at least more than one sentence should be taken into consideration when
determining the ML, the definition as to how large a discourse sample constitutes
a large enough context is rather unclear. In addition, as pointed out by Morimoto
(1999), it’s been observed that bilingual speakers’ intuition with regards to the
language assignment of an individual utterance seems to be independent of
neighboring utterances. In other words, bilingual speakers can rather accurately
pinpoint the base language between two participating language pairs regardless
of whichever language is dominating within the whole conversation. This
Observation is quite deviant from the claims of the MLF model, which predicts

that the speakers’ judgment of language assignment should depend on the whole

34

discourse unit. Therefore, a clause-by-clause basis of determining the ML seems
more plausible.

Finally, the puzzling double morphology data from my Korean-Chinese
corpus raises a crucial challenge for the MLF model. Double morphology is
referred to by Myers-Scotton to as a single head that “has affixes from both the
ML and the EL marking a feature” (1993, 61). The double morphology
phenomenon has been widely observed in the bilingual codeswitching literature
(Lingala/Chiluba/French, Kamwangamalu 1990; Shona/English, Crawhall 1990;
Turkish/Dutch, Backus 1990; Maori/English, Eliasson 1991; Lingala/French,
Bokamba 1988, etc.). Across the language pairs studied so far, the double
morphology phenomenon is mainly exemplified in double-plural affixes (Myers-
Sctton 1993). The following example is from Myers-Scotton’s Shona/English
corpus:

[65] ....dzimwe dzenguva tinenge tichiita ma-game-s panze....

“....sometime we will be doing games outside....”

(Shona/English, Myers-Scotton 1993)

Here, the noun “game” shows the affixes of both the ML and the EL, one is the
Shona Class 6 pluralizing prefix —ma, the other one is the English pluralizing
suffix —s.

Dramatically different from this traditional understanding of the double
morphology phenomenon, my data shows very interesting cases of double

morphology that are quite unique. Consider the following examples:

35

[66]. bu xiang naoxueshuan, 111mm.
Not like cerebral hemorrhage like

“Not like cerebral hemorrhage dcsease.....”
[67]. ta W ca-nun kes kathae,
He seems like sleep-ing fact As if
“It seems like that he was sleeping.”

[68]. ha shi tebie peiyang cheng zhuanye anhflil leymyun....
Not be particularly train become expert not want to

“If you don’t want to train (him/her) to be a specialist (in the field) .....

[69]. you Ireneng nacwung-ye diao -hal kanungseng iss-nun-dey....
Have possible later on fall off do-sup. Possibility be-although

9,

“(pure filling in) will possibly fall off later on, but ......
[70]. saram-i pinwei shi mag hey -r_Q...
men —Nom taste is using tongue — by (using)
“Men are using tongue to taste things....”
[71]. m dabufen dangci di - ha-ci-man
although majority level low —do sup.- although
“Although most of them are pretty shaky....”
[72]. ye slu’ pok-i-ci
also be blessing-be-DCL
“(that) is also a blessing.”
[73]. tamen mi naxie pis-sa-n-ke We yok-hal-kye mwueya?
They buy those expensive-stuff buy-Pst-thing Blame why

“Why would (she) blame them for buying those expensive stuff?”

36

It seems that Chinese morphemes and Korean morphemes with identical
meanings occur simultaneously in one sentence, such as the preposition xiang
“like” and the postposition cherem “like” in [66]; the adverbial phrase haoxiang
“as if” and kath-hae “as if” in [67]; negative adverb bu “no” and anh-i “no” in
[68]; the adverbial phrase you keneng “possibly” and kanungseng “possibility” in
[69]; the preposition yong “with” and the postposition —ro “with” in [70]; the
conjoining adverbial phrase suiran “although” and adverbializer particle ci-man
“although” in [71]; the copula shi “be” and the existential copula i “be” in [72];
and the main verb mai “to buy” and sa-ta “to buy” as shown in [73].

It is worth mentioning here that such phenomena are by no means idiosyncratic
in the Korean-Chinese bilingual data; previous literature in codeswitching has
observed numerous similar examples. Nishimura (1985, 1986), in her study on

J apanese/English intrasentential code-switching, notices that in some CS forms,
which she refers to as mirror image correspondence, the matrix language cannot
be determined. The following are examples of mirror image correspondences
between English and Japanese:

Table 2.

 

English Japanese

 

VO 0V

V complement Complement V
There V NP NP V

P NP NP P

Comp S S particle

 

37

In her data, she observed, “each of the above English constituent orderings is
combined with its Japanese equivalent with a shared element” (Nishimura 1985,
83). Below are some examples given by her:

[74] We bought aboutmm gurai kattekita no.
about bought DCL

[75] There’sghildmn iru yo.
EXT DISC

[76] Let’s becomekgchj ni naroo.
tight Let’s become

Based on Nishimura’s analysis, the underlined parts are shared by the two
languages. In [74], two semantically identical verbs show up in the utterance to
share the common object Was their complement. With the same
pattern, in [75], the NP ghfldmis shared by two copulas from both languages.
In [76], the complement keghi is the shared element of an English X: A
complemenLand a Japanese Weanstruction.

Some comparable data has been observed in Park, Troike, and Park’s
(1989) research on Korean/English CS, further supported in Park’s (1990)
dissertation on Korean/English CS. Consider the following examples:

[77] My parents didnfi helak-haci mam.
Allow-do Neg.

“My parents did not allow (it).”

38

[78] When I was 8 years old (tax, I became a Christian.
When
“When I was 8 years old, I became a Christian.”

[79] enni-nun as an elderly care social worker — lose il-hayyo
sister-Top as

 

“My sister is working as an elderly care social officer.”
[80] wuli-nun 5101:0212 went to ghurch on Sundays.
We- Top church-to
“we went to church on Sundays.”

(Korean/English, Park 1990)

Again, in all these examples, two semantically identical items occur in one
utterance, such as didn ’t and anasseyo “did not” in [77], when and ttay “when” in
[78], as and lose “as” in [79], and kyohoy-ey “to church” and to church in [80].
Auer also mentions the significance of such phenomena, which be labeled as

double marking, and states that such double marking is “a well-known exception
from the generalization that grammatical elements should always be taken from the
matrix language” (1999, 328). The following example is from a case investigated
by Boeschoeten (1983):
[81] ta__ ’ sinp-i sas m.

Until grade six terminative (until)

“until grade six”

(Tajik/Turkic, Boeschoten and Backus 1997)

39

Here, a Turkic suffix gaca “until” and a Tajik preposition ta “until” appear
simultaneously in one utterance. Auer further points out that such phenomena
commonly occur in a pre-lpost-modifying language pair (p. 329), which is
supported by my Korean/Chinese data, together with that of Park (1990),
Nishimura (1985), and Park, Troike and Park (1983).

The double morphology phenomenon remains one of the unsolved cases
for the MLF model. Recognizing this inadequacy, Myers-Scotton and Jake
(2000) extended the MLF model to the 4-M model, mainly to refine the content
versus system morpheme distinction of the MLF model. The efficacy of the 4-M
model in explaining diverse data has been tested by various researches (Schmitt
2000, Wei 2000, Fuller 2000, Bolonyai 2000, etc.). One of the major
contributions of the 4-M model, as claimed by Myers-Scotton and Jake, is that it
“provides more precise analyses of such phenomena as ‘double morphology’”
(i.e., an early system morpheme from the EL duplicating one from the ML)(5).
Whether this claim holds true cross-linguistically or not, particularly with regard
to the problematic data I presented previously, is still to be answered. In what
follows, I will first briefly introduce the 4-M model and then provide a full
explanation of why and how this extended MLF model doesn’t really solve the
problems we are facing in this study.

Briefly speaking, in addition to the distinction of content and system
morphemes, the 4-M model further classifies morphemes into four types. Rather
than simply using the criterion of thematic role assignment, two more features

are added to distinguish four different types of morpheme: one is [+/-

40

conceptually activated], and the other one is [=referring to grammatical
information outside of its XM‘x] (Myers-Scotton & Jake, 2000). A major premise
of the model is that different morpheme types are activated at three different
levels of production: at the conceptualizer, at the lemma level, and at the

formulator. These classes of morpheme will be discussed separately below:

CanteaLmamhemes

Content morphemes are thematic role assigncrs or receivers and are
present at the conceptual level to convey speakers’ intentions. The lemmas
supporting content morphemes are directly-elected by the bundles of abstract

semantic and pragmatic features.

WW

Indirectly elected by the speaker’s intentions, early system morphemes are
also activated at the conceptual level. They “group with content morphemes as
expressing the bundles of semantic and pragmatic features satisfying the
speaker’s intentions” (Myers-Scotton & Jake 2000, 1962). Early system
morphemes crucially contrast with content morphemes in that they lack the
feature of [+/- thematic role assignment]. Examples of early system morphemes

are the definite article the in [82] and up in [83].

[82] I found thabook that you lost yesterday.

[83] Bora chewed up_the new toy.

41

In [82], book indirectly elects the to complete the semantic/pragmatic feature
bundle called by the speaker’s intentions, adding definiteness to book. In [83],
chew indirectly elects up to convey a different idea than a singly occurring chew

does.

Latemtemmmhemes

Instead of being activated conceptually, late system morphemes are
structurally assigned when a larger constituent is constructed; therefore, their
access in the production process is later than that of early system morphemes.
They function as indicators of the grammatical relations between elements. Two

types of late system morphemes will be discussed separately below:

Bridzamtemmmhemes.

Bridge system morphemes are activated when the structure of their
maximal projection requires them. Bridge system morphemes share a common
property with early system morphemes in their dependence on maximal
projection for their form. However, their relationship with their heads is quite
different: bridge system morphemes depend on the grammatical configurations
instead of on the content morpheme that is the head of that maximal projection.
The preposition of and the possessive —‘s in English are typical bridge system

morphemes, as shown in the following examples:

[84] a friend of Bora

[85] Bora ’s friend

42

Here, both of and -‘s have no semantic-pragmatic relationship with their heads;
the existence of these two elements is a purely grammatical requirement of well-

formed English structural configurations.

WW

Outsider system morphemes, as implied in its wording, depend on an
interaction with many grammatical and pragmatic features OUTSIDE the
immediate maximal projection. For example, English third person singular -s is
under AGR in INFL, but the morpheme’s form depends on coindexing with the
subject NP, showing no relationship with its immediate project (here verb).

Adopting this new classification of system morphemes, Myers-Scotton &
Jake show how the long-existing problems with double morphology data, plural
doubling examples in particular, can be explained easily under the modified
Double —morphology Hypothesis:

In mixed constituents in classic code switching, only embedded-language

early system morphemes double system morphemes from the matrix

language. (2000, 1073)
In other words, the doubled system morphemes of the EL do not invalidate the
ML principle in that they are conceptually activated early system morphemes.

In comparison with Myers-Scotton’s double morphology data, a close
examination of the double morphology data in my Korean/Chinese corpus as well
as others will show that this principle fails.

First of all, Myers-Scotton assumes that the assignment of the ML is

rather obvious in the mixed constituents involving double morphology

43

phenomenon. However, it is not always the case. A few are repeated here for
convenience:
[86]. slim; dabufen dangci di - ha-ci-man
although majority level low —do sup.- although
“Although most of them are pretty shaky....”
[87]. ye all! Pok-i-ci
also be blessing-be-DCL
“(that) is also a blessing.”
[88]. tamen mi naxie pis-sa-n-ke We yok-hal-key mwueya?
They buy those expensive-stuff buy-Pst-thing Blame why

“Why would (she) blame them for buying those expensive stuff?”

Apparently, it is very difficult to accurately distinguish the ML and the EL in
these data, even for native bilingual speakers. Since Myers-Scotton’s Double
morphology hypothesis can only be tested under the assumption mentioned
previously, the position will be even harder to hold.

In addition, according to Myers-Scotton, the double morphology
phenomenon refers only to two affixes from both the ML and the EL attached to
a single head ( 1993, 61). However, numerous examples from various language
pairs provide us with enough evidence to reject this definition of double
morphology; this version is too narrow in that it only applies to some languages
with certain morphosyntactic features. Obviously, the doubled items in
Korean/Chinese, J apanese/English, and Korean/English double morphology data

cover a wide range of categories, from a singly occurring head to a longer

constituent—phrase, or even small clause. The doubled verbs in [89] & [90],
repeated below, are apparently content morphemes.
[89]. tamen mi naxie pis-sa-n-ke Stink; yok-hal-key mwueya?
They buy those expensive-stuff buy-Pst-thing blame why
“Why would (she) blame them for buying those expensive stuff?”
[90] We bought about W gurai kattekita no.
about bought DCL

(J apanese/English, Nishimura 1985)

Furthermore, some doubled constituents are phrases as in [89] (here a PP)
or even longer as in [90]. Consider the following examples:
[91] wuli-nun [mam went techmch on Sundays.
We- Top church-to
“we went to church on Sundays.”

[92] Let’s become mm ni naroo.
tight let’s become

(Japanese/English, Nishimura 1985)

(Korean/English, Park 1990)

Enough evidences have been presented above to show that Myers-Scotton’s MLF
model, along with its extended 4-M model, is inadequate to provide a unified
explanation for codeswitching phenomena; particularly for the ongoing research
on Korean/Chinese bilingual speech, it fails to provide a satisfactory explanation

for the double morphology data.

45

IV. Conclusion

This thesis has attempted to test the validity of Myers-Scotton’s MLF
model by examining Korean-Chinese intrasentential CS, a language pair that are
morphosyntactically more highly disparate from each other than the ones under
her study, such as English/Swahili, Shona/English, etc.. It has been found that,
although the MLF model provides a nice account for most commonly occurring
CS data in Korean-Chinese CS, it still shows insufficiency in providing a
satisfactory explanation for some of the phenomena that have been revealed in
language pairs with divergent syntactic and morphological features.

First, the MLF model fails to predict an asymmetry in terms of occurrence
of switchable content morphemes, a phenomenon abundantly exemplified in
many bilingual CS speeches. Second, the central principles of the MLF model
are in conflict with Myers-Scotton’s frequency-based ML criterion. Since ML
assignment functions as a cornerstone of the MLF model, a more objective
criterion of ML designation that is fully compatible with major claims of the
MLF model is required. Finally, the double morphology data—in its expanded
version—challenges the MLF model’s key assumption that one language is
dominant over the other in terms of activation level.

In conclusion, this study suggests that the MLF model cannot be
postulated as a universal in its present form. Further development and
modification is in need.

Without a doubt, additional studies need to be carried out on other

syntactically and morphological contrasting language pairs to determine whether

46

what have been shown here to be weaknesses in the current formulation of the
MLF model are born out for other language pairs as well. As for the Korean-
Chinese bilingual CS, a language group that has hardly ever received any study,
it is just an initial step for a descriptive study on Korean-Chinese CS; the
presented data are by no means exhaustive. However, it certainly opens a new
avenue for investigation and expands the scope of research on intrasentential CS,
especially studies among language pairs that have significant morphosyntactic

dissimilarity.

47

Bibliography

Appel, R. & Muysken, P. 1987. Some Aspects of Code-Switching in New York
Puerto Rican English. Bilingual Review 11.

Auer, P. 1999. From Codeswitching via language mixing to fused lects: Towards
a dynamic typology of bilingual speech. The International Journal of
Bilingualism. 3, 4, Dec., 309-332.

Backus, A. 1990. ‘Turkish-Dutch Codeswitching and the Frame Process Model’,
paper presented at the International Workshop on Ethnic Minority
Languaes in Europe, Tilburg, Netherlands, Dec.

Baetens Beardsmore, H. 1982. Bilingualism: Basic Principles. Clevedon: Tieton,
1-27.

Belazi, H.M., Rubin, E.J., & Toribio, A.J. 1994. Codeswitching and X-bar
theory: The functional head constraint. Linguistic Inquiry, 25, 221-23 7.

Bentahila, A., & Davies, BE. 1983. The Syntax of Arabic-French code-
switching. Lingua 59, 301-330.

Bloomfield, L. 1933. Language. New York: Holt, Rinehart and Winston.

Boeschoten, H. & Backus, A. 1997. Code-switching and ongoing linguistic
change. Turkic Languages 1, 41-62.

Bokamba, E. 1988. ‘Are There Syntactic Constraints on Code-Mixing?’, in K.
Denning (ed.) Variation in Language, 35-51. Stanford, Calif.: Stanford
Univ. Press.

Bolonyai, A. 2000. “Elective Affinities”: Language contact in the abstract
lexicon and its structural consequences. International Journal of
Bilingualism, 4, 81-106.

48

Chang, Suk-Jin. 1996. Korean. Philadelphia: John Benjamins.

Chao, Y.R. 1968. Language and symbolic systems. Cambridge: Cambridge
University press.

Chen, S., Tzeng, O., & Bates, E. 1990. Sentence interpretation in Chinese
aphasia. Unpublished manuscript, University of Southern California;
University of Calilfornia, riverside; University of California, San Diego.

Crawhall, N. 1990. Shona/English Data. Unpublished.

Diebold, AR. 1963. Code-switching in Greek-English Bilingual Speech. Reprot
of the Thirteenth Annual Round Table Meeting on Linguistics and
Language Studies. Washington: Georgetown.

Di Sciullo, A. M., Muysken, P., & Singh, R. 1986. Government and code-
switching. Journal of Linguistics, 22, 1-24.

Di Pietro, R. 1977. Code-Switching as a Verbal Strategy among Bilinguals. In F.
Eckman (ed.), Current Themes in Linguistics: Bilingualism, Experimental
Linguistics and Language T ypologies. Washington, DC: Hemisphere
Publishing.

Eliasson, S. 1989. ‘English-Maori Language Contact: Code-Switching and the
Free-Moprheme Constraint’, Reports from Uppsala University department
of linguistics, 18: 1-28.

Fuller, J. 2000. Morpheme types in a Matrix Language Turnover: The
introduction of system morphemes from English into German.
International Journal of Bilingualism, 4, 45-58.

Grosj can, F. 1982. Life with Two Languages: An Introduction to Bilingualism.
Massachusetts: Harvard University Press, 55-99.

Gumperz, J .J . & E. Hernandez-Chavez. 1975. Cognitive Aspects of Bilingual

49

Communication. In E.Hernandez-Chavez et al. (eds.), El Lenguaje De Los
Chicanos, Arlington, 54-64.

Hasselmo, N. 1970. Code-switching and Modes of Speech. In Glenn G. Gilbert
(ed.), Taxes Studies in Bilingualism. Berline: DeGruyter, 179-210.

Haugen, E. 1956. Bilingualism in the Americas: A Bibliography and Research
Guide. Alabama: University of Alabama Press.

Hornby, P.(Ed.) 1977. Bilingualism: Psychological, Social, and Educational
Implications. Academic Press, New York.

Hymes, D. 1974. Foundation in Sociolinguistics: An Ethnographic Approach,
Philadelphia.

J oshi, AK. 1985. Processing of sentences with intrasentential code-switching.
In D. Dowty, L.Kartunnen, & A. Zwicky (Eds.), Minority Language today
(pp.40-51). Edinburgh: State Mutual Book.

Kachru, BB. 1978. ‘Code-Mixing as a Communicative Strategy in India’, in J .E.
Alatis (ed.), International Dimensions of Bilingual Education, 107-24.
Washington, DC: Georgetwon Univ. Press.

Kamwangamalu, NM. 1990. ‘Bilingual Code-Mixing: Syntax and Constraints’,
rev. version of paper given at African Linguistics Conference, Univ. of
Georgia, Athens, Apr.

Lance, D.M. 1975. Spanish-English Code-switching. In E. Hemandez-Chavez
(eds.) El Lenguaje de Los Chicanos: Regional and Social Characteristics
Used by Mexican Americans. Arlington, Virginia: The Center for
Applied Linguistics, 138-153.

Lehiste, I. 1988. Lectures on Language Contact. Cambridge, Massachusetts: The
MIT Press, 1-27, 99, 123.

Li, P. 1989. What Cues can Chinese speakers use in sentence processing? Paper

50

presented at the workshop on cross-linguistic studies of sentence
processing, Department of Psychology, Carnegie-Mellon University.

Li, C. & Thompson, S. 1981. Mandarin Chinese: A Functional Reference
Grammar. Berkeley & Los Angeles, CA: University of California Press.

Lipski, J .M. 1978. Code-switching and the problem of bilingual competence. In
M. Paradis (Ed.), Aspects of bilingualism (pp. 250-264). Columbia, SC:
Hombeam.

Loebell, H. 1989. Syntactic Aspects of Bilingual Language Production.
Unpublished MA Thesis. Michigan State Unviersity.

Macnamara, J. 1967. The Bilingual’s Linguistic Performance: A Psychological
Overview. Journal of Social Issues 23, 59-77.

Mahootian, S.1996. A competence model of codeswitching. In J .Arnold, R.
Blake, B. Davidson, S. schwenter, & J. Solomon (Eds.), Sociolinguistic
Variation: data, theory and analysis (pp. 387-399). Selected Papers from
NWAV 23 at Stanford, Stanford, CA: CSLI Publications.

Morimoto, Y. 1999. Making words in two languages: 3 prosodic account of
Japanese- -English language mixing. The International Journal of
Bilingualism. 3,1, Jan., 23 -.44

Myers-Scotton, C. 1993a Duelling Languages: Grammatical Structure in
Codeswitching. Oxford: Clarendon.

. 1995. A lexically based model of codeswitching. In L. Milroy
& P. Muysken (Eds.), One Speaker two languages: Cross-disciplinary
perspectives on codeswitching. Oxford, Clarendon Press.

 

& Jake, J. 2000. Four types of morpheme: evidence from
aphasia, code switching, and second-language acquisition. Linguistics,
38-6, 1053-1100.

 

51

Nishimura, M. 1985. Intrasentential Code-switching in Japanese and English.
Unpublished Ph.D. dissertation, University of Pennsylvania.

Pandit, I. 1990. ‘Grammaticality in Code-Switching’, in R. Jacobson (ed.),
Codeswitching as Worldwide Phenomenon, 33-69. New York: Peter Lang.

Park, J ., Troike, R.C., & Park, M. 1989. ‘Constraints in Korean/English Code-
switching: A preliminary study. Paper presented at the 9“1 Second
Language Research Forum, University of California, Los Angeles.

Park, J .E. 1990. Korean/English Intrasentential Code-Switching: Matrix
Language Assignment and Linguistic Constraints. Ph.D. dissertation.
University of Illinois at Urbana-Champaign.

Pfaff, C. 1979. Constraints on language mixing: Intrasentential code-switching
and borrowing in Spanish/English. Language, 55, 291-318.

Poplack, S.1980. Sometimes I will Start a Sentence in Spanish Y TERMINO EN
ESPANOL: Toward a Typology of Code-Switching. Linguistics 18, 581-
618.

Poplack, 1993. Variation Theory and Language Contact: Concepts, Methods and
Data. In D.R. Preston (ed.), American Dialect Research.
Amsterdam/Philadelphia: Benjamins, 1-32, 251-86.

Ryes, R. 1976. ‘Language Mixing in Chicano Bilingual Speech’, in J .D. Bowen
and J. Ornstein (eds.), Studies in Southwest Spanish, 182-8. Rowley,
Mass: Newbury House.

Sankoff, D., & Poplack, S. 1981. Formal grammar of code-switching (Working
Papar no. 8). New York: City University of New York, Centro de Estudios
Puertorriquenos.

Schmitt, E. 2000. Overt and covert codeswitching in immigrant Russian children.
International Journal of Bilingualism 4, 9-28.

52

Scotton, C.M. & Ury, W. 1975. Bilingual Strategies: The Social Functions of
Code-Switching. International Jouranl of the Sociology of Language.

Sohn, Ho-Min. 1994. Korean. New York: Routledge.

Sridhar , S., & Sridhar, K. 1980. The syntax and psycholinguistics of bilingual
code-mixing. Canadian Journal of Psychology/Revue Canadienne de
Psychologies, 34, 407-416.

Stolt, B. 1964. Die Sprachmischung in Luthers Tischreden; Studien zum Problem
der chisprachigkeit. Stockholm, Almquist & Wiksell

Swigart, L. 1992. ‘Practice and Perception: Language Use and attitudes in
Dakar’, Ph.D. Seattle: University of Washington.

Timm, LA. 1975. Spanish-English Codeswitching: E1 Porque Y How-Not-To.
Romance Philology 28, 473-82.

Valdesat F allis, G. 1976. Social Interaction and Code-Switching Patterns: A Case
Study of Spanish/English Alternation. In G. Keller (ed.), Bilingualism in
the Bicentennial and Beyond. New York: Bilingual Press/Editorial
Bilingue.

Wei, L. 2000. Types of morphemes and their implications for second language
acquisition. International Journal of Bilingualism 4, 29-43.

Weinreich, U. 1953. Languages in Contact. The Hague: Mouto; repr. 1967 .

Woolford, E. 1983. Bilingual code-switching and syntactic theory. Linguistic
Inquiry 14, 520-536.

53

            
   

iiiiiiiijiiilflijiijiiuijiiiifliil

9