Title

Language mixing and grammatical development in a Cantonese-English balanced bilingual child in Hong Kong

Author(s)

Yiu, Sze-man, Emily.; .

Citation

Issue Date

2005

URL

http://hdl.handle.net/10722/41327

Rights

The author retains all proprietary rights, (such as patent rights) and the right to use in future works.

Language Mixing and Grammatical Development in a Cantonese-English Balanced Bilingual Child in Hong Kong

Yiu Sze Man, Emily

M. Phil. thesis

THE UNIVERSITY OF HONG KONG AUGUST 2005

Abstract of thesis entitled

Language Mixing and Grammatical Development in a Cantonese-English Balanced Bilingual Child in Hong Kong
Submitted by

Yiu Sze Man, Emily

for the degree of Master of Philosophy at The University of Hong Kong in August 2005 The notion of balanced bilingual development has often been used to refer to two languages which are spoken with equal proficiency during bilingual language acquisition. However, it is still not known how equal the proficiencies of the two languages should be in order to achieve balanced bilingualism. This case study aims to capture this notion of balanced by using the following quantitative measures with respect to grammatical development: mean length of utterance (MLU), upper bound (UB) and lexical diversity (D). Direct comparison between two sets of absolute values does not reliably predict a bilingual childs language dominance due to potential typological differences two languages. The obtained values of these measures are shown to be comparable with those of monolingual children acquiring each language, and thus it is argued that bilingual children who attain such competence in each language are considered as having balanced bilingual development. This study concerns particularly two aspects of bilingual children: a) grammatical development and b) language use in terms of mixing. With regard to a), given that cross-linguistic influence is attested across studies of bilingual language

development, this case study investigates the domains of null objects and wh-interrogatives to see if any cross-linguistic influence takes place. The qualitative similarities found in all Cantonese-English bilingual subjects suggest that the nature of the cross-linguistic influence is similar, whereas the quantitative differences imply that language dominance has determining effect on the phenomenon. Concerning b), since language proficiency is often taken as one of the crucial factors predicting language mixing pattern, this present study examines the relationship of the mixing pattern in the Cantonese-English balanced bilingual child and her language proficiency. The results show that mixing occurs more frequently in Cantonese context than in English context. Since the subject lacks a pattern of language dominance, it is concluded that other factors, rather then language proficiency, play the main role in the asymmetrical mixing observed. It will be proved that mixing by the bilingual subject is influenced by the input properties of bilingual speakers in the speech community in Hong Kong.

Language Mixing and Grammatical Development in a Cantonese-English Balanced Bilingual Child in Hong Kong
by

Yiu Sze Man, Emily

B.A. H.K.U.

A thesis submitted in partial fulfilment of the requirements for the Degree of Master of Philosophy at The University of Hong Kong. August 2005

Declaration

I declare that this thesis represents my own work, except where due acknowledgement is made, and that it has not been previously included in a thesis, dissertation or report submitted to this University or to any other institution for a degree, diploma or other qualifications.

Signed ____________________________________ Yiu Sze Man

Acknowledgments

I am grateful that this thesis was supported by RPG funds offered by University Grant Committee (UGC). Also, I am deeply indebted to Dr. Stephen Matthews for painstakingly reading through my thesis and providing valuable suggestions and comments to my work. The discussions with Dr. Matthews inspired my thesis immensely. I am grateful that Prof. Virginia Yip has been generous with her time to further discuss some issues with me in the final stage of writing the thesis. Their suggestions contribute a lot to the quality of my work. This thesis is thus impossible without their supervision and encouragement which drive me to develop new ideas. Special thanks go to the examiners of my thesis, Prof. Lanza and Dr. Tan who have posed insightful discussions on certain important issues in my work. Furthermore, I have to thank Regine Lai, Vincie Ho, Parco Wong, Charles Marfo, Olivia Lam, Sally Mok, Zhuang Jie in the Department; especially Percy Wong for helping me with his great computer knowledge and John Guest for his generous English suggestions. My life of being a graduate student becomes easier because of them. I am grateful to the audience in the International Bilingualism Symposium 5 (ISB5) for their comments and suggestions on my thesis. I have to owe my thanks to Gary who has pulled me through my difficult time during my graduate study. His help and understanding are unforgettable. Lastly, I would like to express my gratitude to Chung, who has exhausted his rest time to listen to my linguistic ideas and tried to give me comments, although he knows nothing about it.
ii

Table of Contents
Declaration. i Acknowledgements.. ii Table of Contents. iii List of Figures & Tables.. v List of Abbreviations... vi 1 Bilingual Child Language Acquisition. 1 1.1 Definition.. 3 1.2 Dual Input for Bilingual Children.. 4 1.2.1 One-Parent-One-Language Principle.. 4 1.2.2 Lanzas Parental Discourse Strategy... 6 1.2.3 Input in the Community. 11 1.3 Language Mixing in Bilingual Children.. 12 1.3.1 Lexical vs. Grammatical Mixing 13 1.3.2 Syntactic Mixing/Transfer.. 16 1.3.3 Mixing and Language Dominance. 19 1.4 Who are Balanced Bilinguals? . 22 Methodology. 24 2.1 Subject, Data Collection and Transcription. 24 2.2 Measuring Language Proficiency. 26 2.3 Data Selection Criteria for MLU Calculation 27 2.4 Data Selection Criteria for UB Calculation 28 2.5 Data Selection Criteria for D Calculation... 30 2.6 Data Selection Criteria for Mixing Analysis... 31 Defining Balanced Bilingual Children Quantitatively. 35 3.1 Issue of Feasibility of Measuring Cantonese-English Bilingual Childs Proficiency Using MLU, UB and D. 35 3.1.1 Problems of Applying Quantitative Measures to Dual Languages.. 36 3.1.2 Monolingual Contrasts as an Approach to Solve the Problems.. 40 3.2 Findings of Mean Length of Utterance (MLU) & Upperbound (UB)..... 41

iii

3.3 3.4

Findings of Lexical Diversity (D) 47 Discussion: Concepts of Balanced Bilinguals.. 50 3.4.1 Qualitative Evidence Supporting the Result. 52 3.4.2 Possible Cross-linguistic Influence.56

Language Mixing by the Bilingual Child Subject. 62 4.1 Utterances with Lexical Mixing 65 4.1.1 English Nouns in Cantonese Utterances66 4.1.2 English Verbs in Cantonese Utterances. 68 4.1.3 Other English Syntactic Categories in Cantonese Utterances. 71 4.1.4 Cantonese Nouns in English Utterances72 4.1.5 Cantonese Verbs in English Utterances. 73 4.1.6 Other Cantonese Syntactic Categories in Cantonese Utterances. 74 4.2 Discussion: Motivations for Mixing 4.2.1 Input Properties.. 75 4.2.2 Other Motivations for Mixing 83 4.3 Summary of the Analysis.. 85 Conclusions 5.1 Balanced Bilingual. 87 5.2 Mixing in Balanced Bilingual Child. 88 5.3 Further Research 89 References. 92

iv

List of Figures and Tables

Table 2.1 Kathryns transcript files. 25 Table 2.2 MLU values for Kathryns Cantonese and English. 28 Table 2.3 UB values for Kathryns Cantonese and English. 29 Table 2.4 D values for Kathryns Cantonese and English 31 Table 2.5 Breakdown of Language Mixing by Grammatical Category... 34 Table 3.1 Frequency of Null Objects in Kathryns and Timmys Corpora. 60

Figure 3.1 MLU pattern for Kathryns Cantonese and English.. 42 Figure 3.2 MLU pattern for Kathryns English and Adams English... 43 Figure 3.3 MLU pattern for Kathryns English and LLYs Cantonese. 44 Figure 3.4 UB pattern for Kathryns Cantonese and English.. 45 Figure 3.5 UB pattern for Kathryns Cantonese and LLYs Cantonese... 46 Figure 3.6 UB pattern for Kathryns English and Adams English... 47 Figure 3.7 D pattern for Kathryns Cantonese and English..... 48 Figure 3.8 D pattern for Kathryns English and Adams English... 49 Figure 3.9 D pattern for Kathryns Cantonese and LLYs Cantonese... 50 Figure 4.1 The Proportion of Kathryns Mixing in Each Matrix Language 64 Figure 4.2 Percentage of Each Grammatical Category of Kathryns Mixed Utterances.. 66

List of Abbreviations
ASP CL D DEF DP FEM GEN INFL IP MASC MLU POSS PRT SING SFP TTR UB 2SG Aspect Particle Classifier Lexical Diversity Definiteness Determiner Phrase Feminine Genitive Inflection Inflectional Phrase Masculine Mean Length of Utterance Possessive Particle Singular Sentence Final Particle Type-Token Ratio Upper Bound Second Person Singular

vi

Chapter 1: Bilingual Child Language Acquisition

Having more than one language available, bilingual children may show a different acquisition path from the monolingual children. For instance, some studies have found mixing in bilingual childrens language use (e.g. Paradis, Nicoladis & Genesee, 2000; Lanza, 1997); others have reported interaction between two language systems in specific syntactic domains, such as the acquisition of null objects (Mller & Hulk, 2001; Huang, 1999), relative clause construction (Yip & Matthews, 2000), IP construction (Gawlizek-Maiwald, 2003) and etc, whereas some suggest separate development in the respective languages in their subjects (Meisel, 1989; De Houwer, 1990). Moreover, most related investigations only focus on a particular group of bilingual children. Therefore, the findings of one study may not be relevant to the same extent for other groups of bilingual children. For example, Mller (1998) in her study argues that the subordinate clause structure is transferred from languages like French, Italian or English to German because of input ambiguity. However, Yip & Matthews (2000) in their study rule out this possibility in the discussion of transfer of Cantonese to English regarding the wh-in-situ interrogatives and relative clauses since there is little room for ambiguity in English input. In addition, the level of mastery of languages varies from one child to another. Someone may only have minimal control of one language but attain a native level in another; or someone may have an equal of command in both languages. Thus it should be noted that bilingual individuals may perform differently and it is significant to specifically define which type of bilinguals are being investigated. Language input has been claimed to be one of the determining factors in bilingual language acquisition (e.g. Lanza, 1997; De Houwer, 1990). De Houwer
1

suggested that regular input of two languages from birth would ultimately lead to separate development for both languages. Input for early language acquisition is limited to parents and other adults or older bilinguals with close contact. Language use by older bilingual children and adults provides the kind of knowledge necessary for young bilingual children to develop full-fledged competence of both languages. For example, the occurrence of mixing by bilingual children is determined by the development of grammatical and pragmatic competence (Kppe, 1996). On the other hand, the sociolinguistic rules of individual communities underlie the pragmatic knowledge of language use. The effect of parental input will be discussed in 1.2. The existence of mixed utterances produced by the bilingual children has further supported claims of differentiation of the two linguistic systems. Furthermore, the findings of a series of studies have suggested that some cases of mixing can be accounted for by language dominance and/or structural complexity (Genesee, Nicoladis & Paradis, 1995; Bernardini & Schlyter, 2004; Gawlitzek-Maiwald, 2003); and some are due to pragmatic reasons like language preference or language choice, conforming to the rules in the community (Comeau, Genesee & Lapaquette, 2003; Jisa, 2000). In other words, different factors may give rise to different types of mixing during bilingual language acquisition. One of the long-standing controversial issues for bilingual language acquisition is whether a bilingual person can achieve balanced proficiency in two languages. Indeed, the notion of balanced bilingualism needs further research as the criteria between studies vary and hence it cannot be rigidly defined so far. This study aims to investigate a number of issues:

The notion of balanced bilingualism as demonstrated through quantitative measurements, namely Mean Length of Utterance (MLU), Upperbound (UB) and Lexical Diversity (D) (in chapter 3).

The cross-linguistic influence that occurs in the balanced bilingual subject (in 3.4.2)

The asymmetrical mixing pattern that is produced by the subject (in chapter 4)

The plausible reasons for the asymmetrical mixing pattern in the balanced bilingual child (in. 4.2)

1.1

Definition McLaughlin (1978) proposes that children who acquire two languages before

age three are considered as cases of simultaneous acquisition of two languages. By this definition, in a hypothetical case, simultaneous acquisition of two languages could be achieved if a child was exposed to a second language after age two: for instance, where both parents are native Cantonese speakers but their child is taken care of by an English-speaking caregiver has English as the second language, which is a common case in Hong Kong. On the other hand, according to Padilla & Lindholm (1984), simultaneous acquisition of two languages applies only to cases which children are exposed to both languages from birth. This is possible either when each parent speaks a different language or when parents and siblings speak different languages. De Houwer (1990) made a rather stringent definition of bilingual first language acquisition (BFLA). BFLA only refers to situations when (1) a child is first exposed to language B no later than a week after he or she was first exposed to language A and (2) a childs exposure to languages A and B is fairly regular, i.e. the child is addressed in both languages almost every day (De Houwer, 1990 p.3). Recent research on bilingual language acquisition (Hulk & Mller, 2000; Lanza, 1997; Jisa, 2000; Sinka, 2000; Genesee et al, 1995; Nicoladis & Genesee, 1997) often draws data from bilingual children who are exposed to both languages from

parents speaking different languages, following the one-parent-one-language strategy; this form of bilingual language acquisition will be the subject of the present study.

1.2

Dual Input for Bilingual Children Generally, caregivers (e.g. parents or maids) are the main source of input for

young children who stay with them most of the time. It has been reported that the way the parents speak to their children does have an influence on the childrens language competence and this influence is especially salient in bilinguals. For example, if parents speak different native languages, children will receive dual input in the family environment; while other bilingual children only begin to get contact with a second language from a wider community because their parents speak their native language(s) which differ from the majority language spoken outside their families. To various extents, the issue of language use is also taken into consideration. For instance, language mixing in bilingual communities is not unusual and it becomes an essential feature in the input for bilingual children (Lanza, 1997). Depending on the parents attitude or their determination to raise children bilingually, children may end up with bilingual competence to different degrees. Functionally speaking, those who can understand or read two languages are regarded as receptive bilinguals, whereas those who can handle all aspects of language use are termed productive bilinguals (Dpke, 1992). In the following paragraphs, an overview of the features and effects of parental input for bilingual acquisition as discussed in the literature will be provided.

1.2.1 One-Parent-One-Language Principle Ronjat (1913) first gave empirical support to the one-parent-one-language

strategy, successfully raising his child bilingually in that his child separated the two languages from birth. Based on the common scenario that a community usually speaks one majority language, Dpke (1992) investigated the upbringing of bilingual children (age 2;4 2;8) who have to acquire a minority language (with fewer sources in the input and less support) through the one-parent-one-language principle. In her study she especially looked at the extent to which features in the languages spoken by the parents influenced the language acquisition by their bilingual children. Although the language environment in all families is very similar, it is found that two of the subjects, Keith and Fiona were more willing and able to use German (the minority language) than the others in interacting with their German-speaking parents. It uses thus found that both the level of consistency of German-speaking parents in their language choice when addressing their children and their insistence in making their children speak with them in German were correlated with the childrens bilingual language development. Besides the parental attitude towards maintenance of the

one-parent-one-language principle, it is also important to consider the discourse structure in which both parents and children are engaged. It is hypothesized in Dpkes (1992) study that the child would become an active user of the minority language if the interaction between the child and the minority language (i.e. German-speaking in this case) parent is equally or more child-centred than the interaction between the child and the majority language (English-speaking in this case) parent. A child-centred mode of interaction is regarded as being responsive to the childs contribution to the conversation. Keiths German-speaking parent employed a more child-centred approach than his English-speaking parent. Additionally, Keith was shown to be more involved in conversational interactions by his German-speaking father than his English-speaking mother. Regarding the quality

of the parental input, Keiths and Fionas German-speaking parents appeared to provide rich input for German with more teaching techniques than do the parents of other children. Consequently, by the end of the study, Keith and Fiona were considered to have acquired German more successfully than the other children. Keith and Fiona are said to acquire German more successfully than the rest of the subjects in being more active in using German with the German-speaking parents. However, by looking at the measurements with regard to the discourse behavior of parents, the results are not distinctive enough to evaluate to which level the parents should reach in order to help children becoming an active user of the minority language (for details see Dpke; 1992, p.104-5). In addition, Dpke seemed to ignore the mixing phenomenon. She did not explain why Keith and Fiona mixed more frequently towards German-speaking parents than English-speaking parents in both recording sessions, while their mixing rates decreased towards German-speaking parents but increased towards their English-speaking parents. In general, the one-parent-one-language principle favors the upbringing of bilingual children to use the minority language as actively as the majority language if the parents are consistent and insistent in separating the two linguistic inputs. Other studies concerning bilingual acquisition (e.g. Jisa, 2000; Hulk & Mller, 2000; Lanza, 1997) also employed this strategy. The results shows that there is a tendency to nurture active users of both languages through means of teaching techniques by intentions of the parents in a child-centred discourse structure. However, there is no one determining feature in the parental speech addressing the bilingual children that would ensure that the children become active bilinguals.

1.2.2 Lanzas Parental Discourse Hypothesis The impact of parental input on language mixing in bilingual children has been

investigated in Lanzas (1997, 2001) study using a language socialization framework as developed by Ochs and Schieffelin (1995). By advocating that all children, monolingual as well as bilingual, are socialized into community norms for language use and language choice through everyday interactions (Lanza, 2001), in Lanzas (1997, 2001) study the parent-child conversational interaction is examined in order to explain the occurrences of lexical mixing in the bilingual child. Lanza stresses the importance in taking the analytical approach of examining the conversation between parents and bilingual children because the response from parents arguably acts as contextualization cue. From a language socialization perspective, contextualization cues serve to signal whether the interactional context is to be more monolingual or bilingual, and this contextualizes bilingual childrens language alternation within the unit of family or community. Hence, the bilingual children are able to use one or another language, or even language mixing, appropriately according to the prevailing sociolinguistic norms. From the data collected from a Norwegian-English bilingual child, Siri, a continuum of parental discourse strategies in response to mixing by bilingual children is proposed. These strategies, whether used consciously or not, constitute either a more monolingual or a more bilingual context. At the monolingual end of the continuum is the Minimal Grasp Strategy, which an adult uses to indicate a lack of comprehension of the mixed utterance by the bilingual child, which in turn, prompts the bilingual child to carry on the conversation in the designated language choice. For example, 1. CHI: klappe hand MOT: Hm? CHI: clap hand

(Lanza, 1997 p. 201)

Siri switches from a mixed utterance to an entirely English utterance upon her

mothers cue to switch to a monolingual context. With the Expressed Guess Strategy, the adult performs a guess as to the mixed item in the bilingual childs utterance through the translation of the mixed item, and thus the child can confirm or disconfirm the guess. Consider example 2: 2. MOT: Yeah, what does the vov-vov want? CHI: ben MOT: A bone? CHI: yeah (Lanza, 1997 p. 263)

The mother recasts the mixed item by using the English equivalent word in question. This strategy ensures a less monolingual context as the parent has implicitly revealed his or her knowledge in that language. Slightly different from Expressed Guess Strategy, the Repetition Strategy responds to the mixed utterance by repeating the meaning using the other language in a non-question form as in example 3, so that no answer is required. 3. CHI: snn og ny diaper Like that and new FAT: Og s en ny bleie And then a new diaper CHI: clothes? (Lanza, 1997 p. 264)

Moving towards a more bilingual mode of interaction, parents may use the Move On Strategy or Code-switching. With the first one, the parents show no sign of problems in understanding the mixed utterance and carry on the conversation whereas in the latter case, the parents would either code-mix or code-switch themselves. Examples are found in the corpus for Kathryn, who is the subject of the present study: 4. CHI:hai6 aa3 em6 # laam4zai2 jiu3 tung4 <leoi5> [//] yes SFP boy have to with leoi5zai2 jat1cai4 marry gaa3 maa3 girl together SFP SFP
8

yes, boys have to marry girls. INV: dim2gaai2 aa3? why SFP Why?

( 4;03.15)

The investigator in example 4 has used the Move On Strategy. The investigator shows comprehension to the childs utterance and carries on the conversation in Cantonese despite mixed item marry in the childs utterance. In example 5, the investigator incorporates the mixed item used previously by the child in her own utterance. 5. CHI: siu2gau2, ei1 lemon+tea bei2 keoi5 sin1 Puppy give it first Puppy, gives lemon tea to it first. INV: zing2 lemon+tea bei2 keoi5 aa4? Make give it SFP Make lemon tea for it?

Lanza found the overall mixing rates of Siri in conversing with her mother and father to be decreasing in general, except for an increase in the mixing rate with her father after a trip to USA between the second and the third recording. Grammatical mixing takes up 48% overall mixing when Siri interacts with her mother, while nearly 100% of mixing by Siri with her father is lexical mixing. Lanza suggested that grammatical mixing in interactions with her mother by Siri is readily accounted for by her dominance in Norwegian and it is the lexical mixing which needs an alternative explanation. By solely looking at the lexical mixing rates in the interactions by Siri with her mother and father, the lexical mixing rate is higher in interactions with her father than with her mother. This finding results from the fact that Siris parents had basically developed different interactional style with her in terms of their discourse strategies towards her lexical mixing. At one side, her mother would attempt to create

a more monolingual context for Siri and on the other side her father allows a more bilingual context, especially near the end of the study. Consequently, the difference of the approach to Siris language socialization between Siris mother and father lead Siri to use mixing in a different fashion to adapt to each parents style. Siri has noticed that her mother refrained from language mixing and provided a negotiation of a monolingual context. The mixing rate in interaction with the mother dropped significantly from the first recording to the subsequent recording and it remains below 6% until it rises to 10% in the last recording. On the other hand, Siris lexical mixing towards her father remained above 10% across the board and reached as high as 50%. This may be attributed to the fact that her fathers use of the Repetition Strategy, Move On Strategy and Code-switching was more frequent in his interaction with Siri than her mothers. As a result, Siri may have perceived mixing as not inappropriate within the conversation. From a language socialization perspective, Lanzas Parental Discourse Hypothesis can indeed provide an insightful analysis of bilingual childrens use of language(s) during the course of language acquisition. However, it seems that deciding which strategy the parent uses at a given turn in response to the childs mixing is somewhat subjective in nature. The lack of independency has posed difficulty for other studies to test the hypothesis and give comparable results (Deuchar & Muntz, 2003; Nicoladis & Genesee, 1998). For example1, 6. CHI: bao. [bath] MOT: what? CHI: baby. MOT: baby yes.

(1;9)

The child was given a picture of a baby but the child first named it as / bao/

The example is taken from Deuchar & Muntz (2003).

10

bath. This prompted the mother to respond with what?. This response could be analyzed as a Move On Strategy if the mother meant to show confusion by the unmatched semantic meaning; or it could be interpreted as Minimal Grasp, the most monolingual strategy, if the mother was querying the language choice of the child. In addition, in the absence of quantitative analysis, it is unknown to what extent differences in interactive style of the adults affect bilingual childrens socialization. Nicoladis & Genesee (1998) and Deuchar & Muntz (2003) attempted to conduct a quantitative study on the relationship of bilingual childrens mixing and parental discourse strategy following Lanzas hypothesis. Both of the studies found overall negative correlation between the two parameters.

1.2.3 Input in the Community In a broader sense of language socialization, Ochs & Schieffelin (1995) assume that childrens cultural competence poses impact on language development. They reported that the use or the non-use of a particular grammatical form is largely influenced by the culturally organized situations. The authors argue that children may not use a form they frequently hear, not necessarily because they lack the relevant grammatical competence, but because it is indexically inappropriate for them to do so. For example, to Samoan children, higher status people would use a lot of imperative forms to them such as come, but they are found to acquire it later in life as the children realize that they are not sanctioned to use this form. Conversely, the authors noticed that the Samoan children may use some grammatical forms which rarely appear within their hearing environment. For example, they use the first person affect-marked pronoun poor me/I in their speech of begging. Though the form is productive, it is infrequently used among adults. It demonstrates that children strategically make use of certain linguistic devices to help fulfill their desires in life.

11

In the same way, the language choice of bilingual children is affected by the underlying language ideology within a community, as when one language has a more recognized role while the other one has a diminished value. Luke (1998) pointed out a similar scene in the Hong Kong speech community. In Hong Kong, while Cantonese is the native language of most of the Chinese people, English has taken on a prestigious status and it is one of the languages taught in schools. Luke claims that in order to be identified as better educated and having a western outlook, people would insert some English items into Cantonese utterances and he called it orientational mixing. These people indeed know the Cantonese counterparts of the mixed English items but they intentionally avoid using the Cantonese ones in order to achieve their goal. Throughout their work, Ochs & Schieffelin (1995) have stressed that grammatical development cannot be adequately accounted for without serious analysis of the social and cultural milieu of the language acquiring child (p.91). Although social knowledge of language use is highlighted, the linguistic competence of grammatical forms remains essential because it has to be there to enable children to understand or produce the language. Therefore, it is believed that grammatical development of bilingual children has to be examined from both linguistic and cultural perspectives.

1.3

Language Mixing in Bilingual Children In recent literature, most of the studies concerning mixing in bilingual children

lend empirical support to the two-system hypothesis of bilingual language acquisition (De Houwer, 1990; Genesee, Nicoladis & Paradis, 1995; Kppe & Meisel, 1995; Mller & Hulk, 2001; Gawlitzek-Maiwald, 2003). Rather than a fusion of two grammatical systems, mixing by bilingual children has been explained by other

12

factors, such as lacking necessary grammatical knowledge or the required social skills for mixing has not yet developed. Therefore, it is important to identify the types of mixing appearing in bilingual children before investigating any underlying reasons.

1.3.1 Lexical vs. Grammatical Mixing From a qualitative perspective, mixing can be classified into lexical and grammatical types. This classification is based on the syntactic category of the mixed item, that is, whether it is a lexical or grammatical morpheme. It is important to note that the distinction between lexical and grammatical morpheme has not reached agreement. In Lanza (1997)s study, lexical morphemes are largely for those items which possess semantic content; the prototype for this includes nouns, verbs, adjectives and adverbs. However, some subclasses of these categories may be taken out from being lexical as they are classified as closed class items, for examples, auxiliaries and modals (subclass of verbs) and adverbs of place (subclass of adverbs). Therefore, they are considered as grammatical morphemes together with other function words and bound morphemes. Lanza found that the conversation between Siri and her mother contains more grammatical mixing than lexical mixing (7i-v). 7. i. Holde Hold + inf. ii. kan do it for Siri Can do it for Siri iii. Mer cookie More cookies iv. kitty eate (all) melk Kitty eats all milk.

In 7i, the English verb hold is inflected by the Norwegian verbal infinitive marker -e when she described what she is going to do with the cat to her mother.

13

Another type of mixing involving verbs is shown in 7ii, where the Norwegian modal auxiliary kan can co-occurs with the English verb do. In the case of nouns, English noun is found to be modified by Norwegian determiners which include demonstratives, quantifiers (such as mer more in example 7iii) and the possessive pronouns. 7iv illustrates both grammatical and lexical mixing in one utterance. The English verb eat is attached with Norwegian verbal infinitive marker and a Norwegian noun melk milk is inserted as its object. Conversely, nearly all the mixing which occurs in Siris conversation with her father is lexical (8i iii). 8. i. Nei bygge house No big.house ii. Masse masse carriage Lots of lots of carriage iii. Namen? Name+DEF:MASC?

8i & ii demonstrate the form of lexical mixing in which English nouns are imported into Norwegian utterances. 8iii is a case of grammatical mixing in which the English noun name is marked by the Norwegian definite marker /-en/ masculine. In another study (Sinka, 2000) of mixing in two Latvian-English bilingual children, Mra and Maija, the term lexical is generally used for categories like nouns, verbs, adjectives, adverbs and prepositions as they contain descriptive content. This type of mixing was mostly found in the subjects. Example 9 i & ii are produced by Mra while 9iii is by Maija. 9. i. MOT: whats this? CHI: its a suns [dog-MASC-NOM] its a dog MOT: its a dog.

(Mra 1;7)

14

ii. CHI: Mummy, dejo! [dance-2SG-IMPERATIVE] Mummy, dance! (Mra 2;1) iii. CHI: Piglet u [sleep-PRES-3] (Maija 1;9) Piglets sleeping In 10i, an English adverbial phrase is inserted in a Latvian utterance. It is to noting that this type of adverb (i.e. adverb of place) is categorized as grammatical morpheme in Lanza (1997)s study. This exposes one of the discrepancies of terms between studies. 10. i. CHI: tur, ne down there there, not down there. (Maija 2;0)

Furthermore, Sinka (2000) identified morphological mixing which involves mixed categories that convey grammatical function and carry information about grammatical properties such as tense, agreement and case. Only a few cases were found . 11. CHI: (m) es uztaisijusi (m) a (m) I make-past participle-FEM-SING a sejias like that! (Maija 2;1) face-FEM-GEN-SING like that (m) I (have) made a face like that! Since the English indefinite article a precedes the Latvian noun in the utterance, this case is classified as morphological mixing. As Latvian does not have indefinite or definite articles and nouns can appear without determiners, the correct use of a is more likely to bring in the grammatical feature (modifying a singular count noun) rather the lexical one. Nevertheless, it is clear that definitions of mixing between studies are incongruent: they may be overlapping with each other to different extents, like in Lanza (1997) and Sinka (2000)s studies, lexical mixing covers essentially the same syntactic categories but the former study has a finer distinction; grammatical mixing

15

and morphological mixing in the respective studies overlap in coverage.

1.3.2 Syntactic Mixing/Transfer Sinka (2000) introduced syntactic mixing as a result of influences from a syntactic structure from one language to another. This is exemplified in the bilingual subjects who applied Latvian word order to otherwise English utterances. 12. i. CHI: she flowers got MOT: whos got flowers? (Mra 1;9) ii. MOT: so what are we doing this morning? CHI: BigEars and Noddy see. (Mra 2;2) 13. CHI: more some FAT: some more?

(Maija 2;0)

In both examples in 12, Mra incorporates the Latvian SOV word order into the English utterances. In addition, the verb form in 12i is unmarked for tense and agreement. In example 13, a switched order of DP is observed. Mller (1998) uses the term transfer to describe the phenomenon of which cross-linguistic influence takes place in a bilingual child. Yip & Matthews (2000) suggest that transfer takes place when a grammatical property of a language is carried over to another language. The authors found that the Cantonese-English bilingual child produces non-target English both qualitatively and quantitatively. For instance, the subjects English contains more wh-in-situ interrogatives than do the monolingual peers. English monolingual children use wh-in-situ primarily in echoing or repeating the prior utterances (14i-ii) 2 , while the pattern used by the subject in Yip & Matthewss study matches the adult Cantonese word order (1517).
2

The examples are taken from Eves data (Brown; 1973).

16

14. i. MOT: Do you know where? EVE: know where? ii MOT: Hes eating what? EVE: eating what?

(1;08) (1;08)

The wh-words in 15i &ii are placed in the same way as in the corresponding Cantonese what- and where- questions (15iii &iv). In 15iv, the subject makes use of to the what to refer to where. 15. i. CHI: This on the what? ii. CHI: You go to the what? iii. Li1 go3 hai6 mat1je5? this CL is what What is this? iv. Lei5 heoi3 bin3dou6? you go where Where (do) you go? (2;04) (2;05)

Example 16 shows that the construction of English why- question matches the Cantonese counterpart (17). 16. CHI: The snail why live in the water? (3;04) 17. zek3 wo1ngau4 dim2gaai2 zyu6 hai2 seoi2 dou6? CL snail why live in water there Why does the snail live in the water? Other syntactic domains like null object and relative clause are found to be transferable from Cantonese to English during bilingual language acquisition. Example 18 shows a target-deviant relative clause in English. This phenomenon is peculiar in the sense that it has not been found in monolingual English data (Yip & Matthews, 2000). This structure resembles the Cantonese counterparts in the way that the relative clause is prenominal in Cantonese grammar, as opposed to postnominal in English. The authors suggested that transfer from Cantonese to English in the domain of relative clause serves as a relief strategy during the

17

language acquisition process3. 18. i. I want to watch videotape. Butterfly. Patrick buy that one. I want Pet-Pet buy that one videotape (2;11.25) ii. [[Patrick buy __]s [that one]NP I want [[Pet-Pet buy__]s [that one videotape]np iii. [[Po4po2 maai5 __]s [di1 tong4-tong2] NP ne1? Grandma buy CL candy-candy SFP What about the candies Grandma bought? (2;07.12) Example of null objects is given below: 19. INV: Wheres your school bag? INV: Any books in it? CHI: Still have.

(2;07.28)

The object of have is omitted in the subjects answer to the previous question. The authors suggested that the Englishs transitive verb structure may be influenced by input ambiguity arise from the optionality of objects in Cantonese grammar. Mller & Hulk (2001) also reported cross-linguistic influence on object omission in bilingual children acquiring Germanic (German and Dutch) and Romance (French and Italian) languages simultaneously. They argue that the Germanic languages exerted influence on the Romance languages since Romance languages allow more than one grammatical analysis and Germanic languages provide positive evidence for one of these possible analyses. From the findings, the bilingual children in their study have illustrated higher object omission rate in their respective Romance language, compared to monolingual children. Monolingual children acquiring the Romance languages with an MLU of about 2.6 produced object drop constructions at about 11% or less, whereas the bilingual data demonstrated above 30% object omission rate. Moreover, the frequency of
The reason(s) for transfer is not within the scope of this discussion. For detailed analysis see Yip & Matthews (2000).
3

18

target-deviant object omission decreases when C-related constructions start to emerge in the Romance languages. Mller & Hulk thus argue that bilingual children find it more difficult to acquire language-specific object omission rules because they are faced with linguistic input of one language which is favored by the positive evidence of the other language. This cross-linguistic influence happens during mapping of universal pragmatic strategies onto language-specific principles, hence the authors call it mapping induced influence. Generally speaking, studies on mixing in bilingual children suggest that mixing is related to the development of syntax. This claim is corroborated in the data of Ivar, one of the German-French bilingual subjects in the study in Kppe and Meisel (1995). They observed that most of Ivars mixing occurred at the switch points of function words before the INFL is in place, while after the functional category is present the mixing rate decreased dramatically as no more mixing of function words is observed. At this later stage, mixing is reduced to only mixing between nouns and determiners in the object NP. Gawlizek-Maiwald (2003) also correlated mixing of VP/IP between German and English with the asynchronic developments of the two languages. Hence, such mixing is only a temporary phenomenon during bilingual child language acquisition.

1.3.3 Mixing and Language Dominance Mixing in bilingual children and language dominance have been seen as interdependent in the literature. Genesee et al. (1995) conducted a study on the predictive value of language dominance with regard to bilingual childrens mixing. They calculated a dominance index which is based on a number of measurements, namely, MLU, Upperbound, MMU and word type, to indicate a bilingual childs language dominance. They found that the rate of inter-utterance mixing is related to

19

language dominance in four out of five bilingual subjects. Children with dominance in one language mix more when they converse with their non-dominant language parent. One balanced bilingual child is found to mix in a comparable amount to both parents. In line with the findings of their study, Bernardini & Schlyter (2004) showed that the higher portion of the syntactic structure of a mixed utterance is lexically realized in the Stronger (dominant) language while the lower portion is realized in the Weaker (non-dominant) language. For example, if French (Fr) is the weaker language and Swedish (Sw) is the stronger language, mixing between these two languages may give the following structures: 20. i. [ett [table]N]DP

(Sw) (Fr) a table ii. [det r [une table] DP ] IP (Sw) (Fr) that is a table. The lower portions of the sentences in the above example, i.e. the head noun in DP and the DP in IP are lexicalized in French while the higher structures are in Swedish. The main distinction of the notions of Stronger and Weaker language lies in the richness of lexicon a language has (one richer and one poorer lexicon). The richness of lexicon is, in turn, measured with MLUw, Upperbound and the appearance and the frequency of use of items lexically instantiating categories that function as the heads. In contrast to Genesee et al. (1995)s findings, Jisa (2000) in her study considers mixing as an indicator of language dominance. She claims that the mixing pattern of the subjects showed a clear dominance for French on the basis of the assumptions laid out in the Dominant Language Hypothesis (Petersen, 1988). The mixing utterances produced by their subjects contain either French or English lexical

20

items with French free grammatical morphemes, but not vice versa in English. Free grammatical morphemes include prepositions (21i), articles (21ii), demonstratives (like the deixis marker in 21iii), pronouns (21iv) and so on. 21. i. the high chair pour Odessa the high chair for Odessa. ii. Is this le sweater mommy? Is this the sweater mommy? iii. a mommy coffee This mommy coffee iv. What this tu got? What this you got? Apart from these findings, language dominance is not found to be a factor to mixing in Deuchar & Muntz (2003)s study. They conducted a correlation test between mixing and dominance and no correlation is found for the bilingual child. Yip & Matthews (to appear) investigate a number of measures of assessing dominance in bilingual children and they do not consider mixing as a potential measure of dominance. Instead, they argue that dominance may be only one of the factors which predict mixing, which is regarded as an aspect of language performance in bilingual children. From a similar point of view, language dominance is found as not reliable in determining the directionality of mixing in the bilingual subjects in Paradis, Nicoladis & Genesee (2000)s study. Mixing utterances of the bilingual subjects are examined based on the Matrix-Language Frame Model proposed by Myers-Scotton (1993). Within the Model, the System Morpheme Principle predicts mixings to be correctly produced under three circumstances: (1) an early or late ML system morpheme in an ML island or ML+EL constituent4; (2) an early or late EL system morpheme in an EL island; or (3) an early EL system

The definition of early vs. late system morphemes are outlined in Myers-Scotton (1993), Myers-Scotton and Jake (2000): ML= Matrix Language; EL= Embedding language

21

morpheme in an ML+EL constituent. Paradis, Nicoladis & Genesee hypothesized that language dominance would lead the bilingual subjects to violate the Principle: the use of a dominant language late system morpheme in an utterance where the ML is the non-dominant language. The results showed that only three out of seven subjects contributed a small amount of violations to the Principle, and the majority of violations were detected in a balanced bilingual child.

1.4

Who are Balanced Bilinguals? Although it is claimed to be hard to achieve balanced bilingualism because

most bilinguals do not use both languages equally frequently in all domains, it does not preclude balanced bilingualism from existing at the syntactic level. Indeed, some studies have identified balanced bilinguals in this regard. Genesee et al. (1995) distinguished bilingual subjects in terms of a dominance index (see section 1.3.3); the closer to zero the index is, the more balanced it indicates. However, problems arise when the value does not represent an obvious tendency towards balanced or dominant, like one of their subjects, Wil, whose score is between balanced and English dominant, and who is classified as balanced but inclining towards English dominance. As it is widely held that there is a continuum of bilingualism, making such a decision is rather impressionistic which may affect the final result. By adopting a similar approach, the Spanish-English bilingual child studied in Deuchar & Muntz (2003)s study shows little difference in the dominance score between her Spanish and English proficiency, hence she is considered as balanced rather than dominant in one language over the other. Apart from quantitative measures, the French-German bilingual child Ivar is identified as balanced because he appeared to acquire some particular grammatical constructions, such as preconjunctional subordinate clauses and adult-like

22

complementizers, in a parallel fashion (Mller & Hulk, 2001; Mller, 1998). Mller abandoned the assumption that he has a preferred language in German as his MLU values for German are little higher than for French until age 2;4; after the age of 2;5, his MLU values show that French has taken up the dominant role. Jisa (2000) claimed that her two bilingual subjects are balanced in comprehending English and French, but were very dominant in French for production. Generally, these studies suggest that balanced bilingual subjects do exist across different language pairs, although they are defined in different ways. The notion of balanced could be determined quantitatively by a number of measurements like MLU, Upperbound and so on. If the two sets of values are alike, which means the childs proficiency of the two languages are at the same level, then they are regarded as balanced. On the other hand, a bilingual child could be rated balanced if he or she shows a parallel grammatical development in both languages in the way that the development concerned is comparable to that of a monolingual. It is important to note that, as many studies comment (Mller, 1998; Hulk & Mller, 2000; Mller & Hulk, 2001), although the development of two languages in bilingual children is largely autonomous, bilingual children are liable to cross-linguistic influence given that the two languages are in contact. Cross-linguistic influence could appear as facilitation/acceleration, delay, or transfer(Hulk & Mller, 2000).

23

Chapter 2: Methodology

In this study, a longitudinal case study of an individual child is adopted. For the purpose of giving a fine-grained analysis of bilingual childrens language acquisition, this study limits the target to a single bilingual child since it can minimize individual differences between subjects which would disturb the result (e.g. De Houwer, 1990; Lanza, 1997; Deuchar & Quay, 2000). Employing longitudinal study aims to observe any developmental changes of the linguistic phenomenon in question in the subject during a substantial data collection period. The present study will use both quantitative and qualitative analyses of a childs speech to address theoretical questions or as evidence to refute certain generalization. For instance, in assessing a bilingual childs language proficiency, quantitative measurements will be used to show a general picture of the language development; for investigating language mixing, quantitative data can again serve to provide general ideas of the linguistic phenomena that are being examined, while qualitative analysis can subsequently distinguish the features of those phenomena.

2.1 Subject, Data Collection and Transcription The subject of the present case study, Kathryn, was brought up in a bilingual family. She was exposed to Cantonese and English from birth and grew up in a one parent-one language environment where her mother was a native speaker of English and her father of Cantonese. Kathryns mother, a housewife at the time of study, was the principal caregiver. The family employed a Filipina domestic helper for a brief period until Kathryn was around age 3, and subsequently a part-time Cantonese cleaner who also spoke fluent English. Kathryn attended the Cantonese section of an

24

international Kindergarten from 2;07. The subjects development in both languages was observed and recorded at weekly or bi-weekly intervals for a period of about one year and nine months, from 2;09.23 - 4;06.07. Kathryns longitudinal data is available in the Hong Kong Bilingual Child Language Corpus (Yip & Matthews, 2000). The recordings were transcribed in CHAT format (MacWhinney, 2000a) which enables investigators to carry out analyses by CLAN programs of the CHILDES project (MacWhinney, 2000a) (see Table 2.1). Transcripts were completed from 3;01.05 to 4;6.07. 17 Cantonese and 16 English transcripts were available for analysis. Age 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 3;01.05 3;02.19 3;03.16 3;04.14 3;06.18 3;07.13 3;08.27 3;09.25 3;10.27 3;11.27 4;00.15 4;01.09 4;02.16 4;03.15 4;04.29 4;05.10 4.06.07 File name (Cantonese context) Kcod950228.cha Kc950411.cha Kcod950509.cha Kcod950606.cha Kcod950810.cha Kcod950905.cha Kcod951020.cha Kcod951117.cha Kcod951220.cha Kcod960119.cha Kc960207.cha Kc960304.cha Kc960409.cha Kc960508.cha Kc960621.cha Kc960703.cha Kc960730.cha Ke950411.cha Ke950509.cha Ke950606.cha Ke950810.cha Ke950905.cha Ke951020.cha Ke951117.cha Ke951220.cha Ke960119.cha Ke960207.cha Ke960304.cha Ke960409.cha Ke960508.cha Ke960621.cha Ke960703.cha Ke960730.cha File name (English context)

Table 2.1 Kathryns transcript files

On the average, each recording session consisted of an hour of audio and in some cases video recordings of the child engaging in her daily activities such as
25

playing, reading and role playing are also available. The subject was encouraged to speak in Cantonese for half an hour and in English for half an hour by separate adult interlocutors. It is noted that the two interlocutors can speak both languages. In addition, despite the efforts made to encourage the subject to speak the specific language during a particular session, the language other than the expected one and mixing of languages are often heard (for discussion see chapter 4).

2.2 Measuring Language Proficiency For the reason that children acquire their language in different pace at a given age, in order to measure Kathryns language proficiency, this study will adopt three measures which are independent of the factor of age. These measures are Mean Length of Utterance (MLU), Upperbound (UB) and Lexical Diversity (D). MLU is the value obtained by averaging the total number of morphemes across all utterances in a sample, while UB calculates the number of morphemes in the longest utterance in the sample of each language As Brown (1973) correlated MLU to a developing childs linguistic complexity, MLU and UB together can serve an index of a childs overall grammatical development. The last measure, D, as its name suggests, is employed to evaluate a childs lexical development. In monolingual child studies, two subjects are considered as having similar degree of language development if they obtain comparable values for these measures. On the other hand, children who show different values for these measures are regarded as developing their languages at different rates (Brown, 1973; Richards & Malvern, 1997). When these measures are applied in studies of bilingual children, a bilingual child would get two sets of values for each measure in each of his/her languages. Presumably, if there was discernable difference between these two sets of values, the child would be considered to show imbalanced development of the two

26

languages. However, studies on bilingual children acquisition (e.g. Dpke, 1998; Yip & Matthews, to appear) indicate problems when applying these measures to studies involving more than one language (see detailed discussion in chapter 3).

2.3 Data Selection Criteria for MLU Calculation In deciding which part of the data is taken to be calculated for MLU values for each language, Browns (1973) criteria are not followed. Brown (1973) suggested that counting MLU on the basis of 100 consecutive and fully transcribed utterances, starting from the 51st utterance in each transcript file regardless of language. As noted by Lanza (1997), counting MLU regardless of the language used by the subject will jeopardize the resulting MLU values. Given the fact that each transcript file of Kathryn contains more then one language code, Browns guideline is not followed because the resulting MLU value cannot represent individual language performance. A solution to this is to gather all the utterances of an individual language from transcripts of both Cantonese and English contexts and the resulting transcript can be processed for MLU values. Furthermore, it is preferable to include all the available utterances in the data because the problem of unfamiliarity between the investigators and the subject has been solved, as transcription did not start from the very beginning of recording sessions when they had been first introduced to each other. Therefore presumably they knew each other quite well after half-year time so the subject did not perform significantly different from her normal speaking performance. Thus, it is unnecessary to remove the first fifty utterances from the calculation of MLU. Table 2.2 shows the MLU results for Kathryns Cantonese and English.

27

Age (year;month.day) 3;02.19 3;03.16 3;04.14 3;06.18 3;07.13 3;08.27 3;09.25 3;10.27 3;11.27 4;00.15 4;01.09 4;02.16 4;03.15 4;04.29 4;05.10 4.06.07

Cantonese MLU 3.034 4.255 4.976 3.14 3.31 4.99 5.242 6.541 5.774 3.831 4.027 5.458 4.588 4.027 4.838 4.78

English MLU 3.126 3.275 3.925 4.521 4.765 3.897 3.227 3.352 3.617 2.645 3.481 4.511 4.076 3.814 3.927 3.517

Table 2.2 MLU values for Kathryns Cantonese and English

2.4 Data Selection Criteria for UB Calculation For the transcript files at hand, UB is calculated by the command maxwd in the CLAN program, by which the longest utterance on the basis of words is generated. It seemingly deviates from the common practice whereby UB is counted in terms of morphemes. However, considering that a unit of word may consist of more than one morpheme, the longest utterance in terms of words should imply the longest utterance in morpheme. The results are shown in Table 2.3.

28

Age (year;month.day) 3;02.19 3;03.16 3;04.14 3;06.18 3;07.13 3;08.27 3;09.25 3;10.27 3;11.27 4;00.15 4;01.09 4;02.16 4;03.15 4;04.29 4;05.10 4.06.07

Cantonese UB 12 17 26 8 12 26 25 29 20 11 12 21 17 14 21 20

English UB 17 17 18 15 19 26 18 20 14 11 16 25 21 22 13 13

Table 2.3 UB values for Kathryns Cantonese and English

In the first attempt to generate UB by the procedure described above, although repetition was excluded from calculation, the result contained two outliers which are suspiciously high in the Cantonese context and one in English context. For Cantonese, one file is reported to contain 62 words in a single utterance at age 3;08.27 while another one reaches 43 words at age 3;10.27. The one in English context reaches 40 words at age 3;09.25. Such result prompts a revision of the data and finds out that the outliners are produced by the subject when she was excitedly giving synopses of the cartoons she usually watched, or reciting the slogans of cartoon figures. This kind of unbroken speech was transcribed as a single utterance. In fact, an utterance can be demarcated by intonational contours in phonology, or it may be treated as a unit conveying one speech act (e.g. requesting information or

29

making an offer). In the absence of agreement on the definition of the term utterance, the analytical unit of an utterance is thus based on the existing transcription. For the purpose of presenting more informative UB values, unrepresentative speech like the story-telling utterances, is isolated from the rest of the corpus in the second time of counting UB. Table 2.3 illustrates the finalized result. 2.5 Data Selection Criteria for D Calculation Richards & Malvern (1997) proposed a new measure, lexical diversity (D), which replaces the original measure, type-token ratio (TTR), and becomes a more robust indicator of the vocabulary development for children. It is shown that the calculation of the traditional TTR has distorted research findings because its validity declines with the increase in sample size (Richards & Malvern, 1997b). In the CLAN program, vocd (the command for computing lexical diversity) uses random sampling of tokens in plotting the curve of TTR against increasing token size for the transcript under investigation. D is then obtained by a model of calculation based on the plotted curve1. An increase in D represents increase in lexical diversity (Malvern & Richards, 2002). As pointed out by Malvern et al (2004), one cannot be sure that the definition of different words is the same among the software packages, transcription systems and researchers. To avoid such anomalies, some editing needs to be done to the transcripts. Firstly, a complete word list for each transcript is obtained by running the command freq in the CLAN program. Since proper names, such as peoples and cartoon characters names, are often spelt or (in the case of Cantonese) encoded inconsistently, they are put in an exclude list (exclude.cut) in which all the entries are

Details are given in the CLAN manual (http://childes,psy.cmu.edu/manuals/CLAN.pdf)

30

excluded for the purpose of calculating D. Moreover, there are a certain number of non-words as well as English alphabetical letters appearing in the word lists which do not have any referential value. Thus, such items are also filtered out by putting them into the exclude file. In addition, retracing induced by either self-repetition or self-correction is excluded from the counting by using +r6 switch in the command. Table 2.4 shows the D result. Age (year;month.day) 3;02.19 3;03.16 3;04.14 3;06.18 3;07.13 3;08.27 3;09.25 3;10.27 3;11.27 4;00.15 4;01.09 4;02.16 4;03.15 4;04.29 4;05.10 4.06.07 Cantonese D 40.42 40.08 48.55 27.85 43.14 39.07 48.36 48.69 44.27 35.02 45.29 50.33 39 31.4 34.81 38.06 English D 68.59 71.09 88.88 54.05 63.57 63.19 55.03 62.58 51.47 60.01 61.46 68.23 75.94 70.64 46.19 80.02

Table 2.4 D values for Kathryns Cantonese and English

2.6 Data Selection Criteria for Mixing Analysis In order to examine any possible contributing factors of mixing utterances in the subject, all the mixed utterances will be considered. Mixing is identified as a case where words from the two languages concerned, namely Cantonese and English, appear within a single utterance. Different from previous studies on language mixing which typically concern questions from either syntactic or pragmatic aspects, the
31

present study will investigate mixing utterances produced by the subject in both syntactic and pragmatic aspects so as to suggest a clearer picture of the language development pattern for the bilingual child in question. Thus, utterances which contain components from two languages will be included for examination. If there is more than one item that is mixed into the other language within an utterance, they will be counted as only one instance of mixing. In Deuchar & Muntzs (2003) study, the term mixed was used to refer to any utterances which were either entirely in the non-context language or utterances which are made up of two languages. The former criterion of mixing is not considered in this study because language context is not strictly maintained due to a main factor: the presence of the interlocutors is not clearly language-differentiated in recording session most of the time. For example, the two interlocutors often participate in the activity during recording. Sometimes they stick to their own responsible languages regardless of the target language of the recording. In an extreme case, the interlocutor who is responsible for English session held the Cantonese session as the Cantonese interlocutor was away from Hong Kong. Besides, as the recordings are held in a natural setting, the interlocutors need to comply with the subjects changes of attention from one interlocutor to the other and so the language may be switched as well. Therefore, the focus will be placed on the intra-sentential kind of mixing, which I will regard as language mixing in the subsequent discussion. Structural mixing (e.g. Gawlizek-Maiwald, 2003; transfer in Mller, 1998 and Yip & Matthews, 2000) is not concerned here because only scant cases of structural mixing are found in the subject2. Since components of both languages are present in mixed utterances, it is
2

Yip & Matthews (in preparation) found that Kathryn produced less wh-in-situ interrogatives (only

two cases) than the Cantonese-dominant bilingual children studied.

32

crucial to determine which one is the mixed item. I adopt the descriptive term matrix language as the language in which mixing occurs and embedded language as the one which provides the mixed items3. The matrix language determines the word order of morphosyntactic elements in the mixed utterance. For example: 1. ho2 m4 ho2ji3 swallow keoi5 gaa3? can-not-can it SFP Can (I) swallow it? Ive got two hak1 car. Ive got two black cars.

2.

Cantonese is regarded as the matrix language and English as the embedded language for example 1 because the English verb is projected within a Cantonese auxiliary phrase and which is constructed in a A-not-A structure (Matthews & Yip, 1994).This structure is unique in Cantonese grammar and not found in English. Example 2 shows a reverse case in which a Cantonese adjective is inserted in the pronominal position to modify the English noun. Should any mixing utterances respect word orders of both languages, the language in which the higher structure of the sentence lexicalized is the matrix language. Given that the language contexts during recording were not

tightly-controlled as the interlocutors are not language-differentiated in the recordings, I will not look at the mixing frequency in each language context. Instead, I will calculate the proportion of Kathryns mixing in each respective matrix language. The mixing utterances are then categorized according to part of speech.

The present study is non-committal with regard to the Matrix Language Frame model proposed by

Myers-Scotton (1993) although I make use of the terms matrix and embedded.

33

According to Table 2.5, the noun-mixing type makes up 89% in Cantonese contexts and 79% in English ones. Noun-type includes proper nouns, bare nouns or full noun phrase projection. Second comes the verbal type of mixing, although it amounts to less than 10% of the mixing corpus (6% in Cantonese context; 9% in English context).

Grammatical Category

Cantonese English utterances with utterances with English items Cantonese items (in %) 89 6 1 4 79 9 9 3

Noun Verb Adjectival Others

Table 2.5 Breakdown of Language Mixing by Grammatical Category

Detailed analyses of the quantitative findings will be presented in the next chapter, while the mixing examples will be discussed in chapter 4.

34

Chapter 3: Defining Balanced Bilingual Children Quantitatively

In discussing the concept of balanced in language development, it is more convenient to pin it down through the concept of imbalanced because there is a substantial amount of literature on bilingual acquisition and many subjects are categorized as dominant in one of their languages. A bilingual child is identified as dominant in one of the languages if he or she is more proficient at that language. Relatively speaking, those who are estimated to be equally proficient at both languages are considered as balanced. According to Genesee et al. (1995), childrens language dominance is equivalent to their relative proficiency in each language. For young bilingual children and adults, language dominance, in addition to language proficiency, may also reflect bilingual speakers preference for reasons that are independent of proficiency. As preference is too complex to be defined in objective and concrete terms, dominance presumably refers only to language proficiency (Genesee et al., 1995; Meisel, 1994). Besides accounting for language performance, the term dominance could be used to refer to a speakers knowledge, rather than proficiency (Lakshmanan, 1995; Yip & Matthews, to appear). Knowledge of language is considered to be the underlying grammar established in a language learners mind. An example of this view is given by Yip & Matthews (2000). Their Cantonese-English bilingual subject exemplified that transfer is asymmetrical in a way which is predicted by dominance.

3.1 Issue of Feasibility of Measuring Cantonese-English Bilingual Childs Proficiency Using MLU, UB and D Dpke (1998) and Yip & Matthews (to appear) have illustrated that problems

35

arise when comparing MLU values for two languages, especially of different morphological type. As calculation of UB and D is also dependent on morphosyntactic structures, it is important to consider whether these quantitative measures can be appropriately applied to measuring the Cantonese-English bilingual childs proficiency in the present study. In section 3.1.1 I will identify the potential shortcomings of employing MLU, UB and D to cross-linguistic comparison especially involving Cantonese and English. Then I will suggest a possible solution to tackle the problems in section 3.1.2.

3.1.1

Problems of Applying Quantitative Measures to Dual Languages Many studies concerning early bilingual language acquisition measure a

bilingual childs language development in terms of MLU (e.g. Redlinger & Park, 1980; Genesee, Nicoladis & Paradis, 1995; Sinka, 2000; Nicoladis & Genesee, 1997; De Houwer, 1990; Lanza, 1997; Yip & Matthews, 2000). Brown (1973) suggested that MLU is correlated to a childs syntactic development. The higher the MLU value of a language is, the more advanced the language proficiency it indicates. While MLU has been a useful quantitative measure in a substantial amount of first language development studies, it is far from sure whether MLU can serve as effectively as a measure for bilingual children when dual languages have to be handled at the same time. If the language pair involves two different morphological types of languages, for example, inflectional languages like French and English, and isolating languages such as Cantonese, the validity of MLU comparison is in doubt. Lacking inflectional morphemes like person, number or tense as instantiated in languages like English and French, Cantonese seems to deflate the MLU values with respect to an equivalent utterance of an inflectional language. For example, Cantonese would be likely to

36

score lower point in MLU as it allows a wider range of use of subject-drop and object-drop in the adults grammar, compared to English-speakers grammar (Klee et al., 2004). On the other hand, Cantonese would arguably inflate MLU values as it has sets of morphemes (which occur with verbs) such as aspect markers and postverbal particles. The former set is associated with verbs to state the relationship of time of actions, while the latter combines with verbs to indicate either the directional, resultative, quantifying or adversative meanings of actions (Matthews & Yip, 1994). In addition, Cantonese makes use of sentence final particles to express pragmatic functions (details see Matthews & Yip, 1994). Klee et al. (2004 p.1403) discovered that English-speaking and Cantonese-speaking children produced different slopes of age-MLU regression lines and suggested that the above reasons might be at play. Yip & Matthews (2000) have attempted to accommodate the discrepancy of the morphological types between Cantonese and English by treating child English an essentially isolating language, as the inflectional system is yet to take place. Therefore, the authors made use of MLUw instead of MLU in their bilingual child case study. However, as it is hard to ascertain when an English-speaking child starts to manipulate the inflectional system productively, MLUw may underestimate the language proficiency for older children such as Kathryn. A number of studies (e.g. Genesee, Nicoladis & Paradis, 1995) have recognized that MLU is problematic when applied to cross-linguistic comparisons of languages which belong to different morphological types, as has been described above. Yip & Matthews (to appear) have derived MLU differentials that can compare a bilinguals two languages in a relative sense. It is applicable when the language pair is the same across subjects. Since the language pair is kept constant, the problem of comparability also remains constant and so the value of differential is able to reflect

37

to what extent the proficiencies of two languages differ from one another. For instance, two of the subjects, Sophie and Alicia, in their study have nearly one point difference between their Cantonese and English MLU values, where Cantonese MLUs are consistently higher than those for English. It is suggested that these two subjects are relatively more Cantonese dominant while the other subject, Timmy, whose mean MLU differential is 0.39, is considered as less Cantonese dominant. In this sense, the MLU differentials can enhance the comparability of language development between bilingual children. However, the feasibility of its application to all types of bilinguals is in question. The authors argue that it is a viable measure of dominance when there is an imbalance between the childs two languages. It is clear that the MLU differential would go down when the MLU values of the two languages are closer. In the case of balanced bilinguals, who are believed to have minimal (if any) asymmetrical language development, it is uncertain how low the MLU differential would get. It is similar to one of the bilingual subjects in Genesee et al. (1995)s study, who was rated as between balanced and English-dominant. According to the findings of Klee et al. (2004)s study, the MLU values for English and Cantonese begin to show discernable differences after age 36 months. Thus, it is possible that the value of the MLU differential that is yielded by balanced bilingual amounts to the potential difference of the MLU values of the two languages, although this has not yet been demonstrated. Another important point of concern involves the MLU pattern of the bilingual children. Since the bilingual children in the study of Yip & Matthews (to appear) obtain the MLU pattern of which Cantonese is consistently higher than English (except for Timmy), it is then meaningful to interpret the MLU differential as a measure of Cantonese dominance. However, if MLU values for two languages produce an intersecting pattern, it is hard to give conclusive remarks on dominance.

38

We now discuss the problems of other measures of dominance, Upperbound (UB) and Lexical Diversity (D). UB calculates the number of morphemes in the longest utterance in the transcripts of each language. However, the utterance that is elicited by this measure might not always reflect the most complicated grammar at a given time that the child ever used. The longest utterance may, for example, be a string of names of food items or animals which only consists of bare nouns and connectives. Conversely, a shorter utterance may include more advanced grammar like relative clause. Therefore, it has limited significance for measuring language development. In addition, it should be noted that the same cross-linguistic problem for MLU calculation applies for UB too, that is, with respect to English, Cantonese still bears the inflationary/deflationary morphological feature in the calculation of UB. Although D has replaced the traditional TTR as a measure of lexical development for first language acquisition, it is not without its problems. There is an intrinsic problem with this interpretation since the value of D is essentially determined by word types. It is noted that the number of word types is related to the morphosyntactic properties of languages, so direct comparison between two languages would possibly lead to misleading results. For example, come, comes, and came are three inflected forms of the verb come, so they are counted as three types. Malvern et al. (2004) reported that the D values for inflected forms is significantly larger than that of the uninflected forms (root form cited in their work) over time in English. In Cantonese, lei4 come itself can represent the corresponding forms come in English and is taken up as one type only. This is similar to the uninflected form for English and thus Cantonese would produce a D value which is smaller than that of English. In addition, external factors may insert influence on the D values. Klee et al. (2004) and Malvern et al. (2004) assert that a wider range of situational contexts

39

allows more topics to come up within conversations and in turns increase the lexical diversity whereas a restricted context would limit verbal production thus reduce the lexical diversity. Klee et al. (2004 p.1404) have claimed that the regression line of their Cantonese data appears to be continuous with that of the English data from Malvern et al. (2004). These findings altogether seem to imply that lexical diversity of Cantonese and English-speaking children develop in a similar fashion. However, since the contexts for recording differ in the two studies, more evidence is needed to strengthen their claim.

3.1.2

Monolingual Contrasts as an Approach to Solve the Problems It has been demonstrated that the essential problem of comparing Cantonese

and English by the above measures is that they are two unlike languages, in which the absolute values obtained are not directly compatible. In other words, those quantitative measures basically work when comparing similar languages. Assuming that monolinguals and bilinguals go through similar paths in language acquisition (babbling stage, one/two-word stages etc) (Bhatia & Ritchie, 1999; De Houwer, 1990), many studies compare their bilingual subjects to monolingual learners of the respective languages (De Houwer, 1990; Mller, 1998; Yip & Matthews, 2000; Peng, 1998 etc). De Houwer (1990) has shown that her bilingual subject, who received separate input regularly from birth, closely resembles monolinguals of Dutch and English of the same age in different aspects of language development. This in turn suggested that the bilingual childs syntactic development proceeds in a language-specific manner. In the study by Yip & Matthews (2000), the Cantonese-dominant bilingual subjects data has displayed qualitative and quantitative differences from the English monolingual data. Mller (1998), on the other hand, found that the erroneous forms produced by monolingual and bilingual

40

children were qualitatively similar but quantitatively different. Nevertheless, these studies have illustrated that it is reasonable to compare bilingual data against monolingual norms in order to attest if any anomaly is present in bilingual children. Therefore, the quantitative results in the present study will be compared against monolingual data of respective languages. Any similarities and differences will be discussed subsequently.

3.2 Findings of Mean Length of Utterance (MLU) & Upperbound (UB) The bilingual subject in this study, Kathryn, acquired Cantonese and English simultaneously from birth. In chapter 2, the results of the quantitative measurements for Kathryn were presented. This chapter will analyze the data obtained from those measurements and in turn justify whether Kathryn is a balanced or imbalanced bilingual child. Consider Figure 3.1 showing the development of MLU in Kathryns two languages. At the beginning of the study (age between 3;02.19 and 3;08.27), her MLU values for English and Cantonese are intersecting with each other. Therefore, it does not show a clear dominance pattern of one language over the other. From age 3;08.27 to the end of the study, the pattern illustrates that the MLU for her Cantonese is always ahead of the English one. Both Cantonese and English have MLU values exceeding 3 most of the time and it can be as high as 6.54 in Cantonese1. It should be noted that aspect markers, postverbal particles and sentence final particles are counted as separate words in Cantonese. In example 1, the sentence that the bilingual child produced has two sentence final particles (lei4 ge2) and the presence of the classifier go3 indicates

The peak value of MLU for Cantonese is due to the subject being very voluble over the play session,

because she was very enthusiastic about talking about her paintings and playing with toys.

41

the definiteness of the noun phrase. Thus there are five morphemes in the utterance whereas the English counterpart has three only. This can illustrate how Cantonese inflates the quantitative measures with respect to English. 1. CHI:li1 go3 mat1je5 lei4 ge2? Dem CL what SFP SFP What is this? (3;01.05)

7 6 5

MLU

4 3 2 1 0
Cantonese English

.1

.1

.1

.2

.2

0 .1

.1

.2

.1

.0

.1

3; 02

3; 03

3; 04

3; 06

3; 08

3; 09

3; 10

4; 02

.2

Age
Figure 3.1 MLU pattern for Kathryns Cantonese and English

Although it seems that the values of English are lower than those of Cantonese nearly throughout the whole period of study, this does not necessarily mean that Kathryn is more proficient in Cantonese than in English. There is reason to believe that, in general, Cantonese morphology bears features that inflate and deflate MLU at the same time to various extents as discussed in section 3.1. Since neither side of this argument has yet to be demonstrated with supporting evidence, to avoid misleading interpretation, the question of which language is more proficient in terms of MLU has to find another way out. The MLU values for both languages are therefore compared with the monolingual norms in each language respectively. In Figure 3.2, the MLU patterns

42

4; 06

3; 07

3; 11

4; 00

4; 01

4; 03

4; 04

4; 05

.0

.1

.2

.1

for Kathryn and an English monolingual, Adam (Brown, 1973), of compatible age are shown. Kathryns MLU values for English range from 2.645 to 4.765 whereas Adams MLU values range from 3.601 to 4.604. The difference between their MLU values is not statistically significant (p>.05). Therefore, Kathryns English syntactic development (in terms of MLU) has shown to be within the developing norm as the monolingual child.
6 5 4

MLU

3 2 1 0 38 39 40 42 43 44 45 46 47 48 49 50 51 52
Kathryn-Eng Adam

Age (months)

Figure 3.2 MLU pattern for Kathryns English and Adams English

Unfortunately, there is no Cantonese monolingual subject of compatible age range available in other accessible Cantonese-speaking childrens corpora. Supplementary to the lack of comparison across the board for Cantonese, a qualitative description of Kathryns Cantonese utterances will be provided in section 3.4.1. For the purpose of reference, a Cantonese-speaking girl, named LLY (extracted from CANCORP (Lee & Wong, 1998)) whose age range overlaps with Kathryn for a 6-month-period (38 to 44 months), is taken for comparison of MLU values (see Figure 3.3). The range of value is between 3.034 and 5.596 for Kathryn while between 3.624 and 4.942 for LLY. Statistical comparison of Kathryns and LLYs MLU values does not show significant difference (p>.05). Therefore, with respect to syntactic development, Kathryns Cantonese demonstrates monolingual norm.

43

6 5 4

MLU

3 2 1 0 38 39 40 41 42 43 Kathryn lly

Age (months)
Figure 3.3 MLU pattern for Kathryns English and LLYs Cantonese

The measure Upperbound reflects the upper limit of language production capability that a child has. The growth of UB values believes to be more correlated with a childs syntactic development, for instance, a syntactically complex utterance will take up more words. Therefore, UB values are expected to appear in a developmental trend in that the values at the end of the recording period should be higher than those in the beginning of recording. The UB patterns for Cantonese and English show similar kind of fluctuation in Figure 3.4. One may query such fluctuation pattern is unable to reflect a developmental trend. A possible explanation to this pattern is that the measured UB has been contaminated by the subjects language preference according to various situational contexts, for example, the subject may like to speak in one language over the other, therefore, the less preferred language over that session would produce a smaller value of UB. This particularly small value cannot solely be correlated with her syntactic knowledge. In another words, UB values for Kathryn do not reveal merely her syntactic development, but also pragmatic development. Thus, UB offers a secondary reference to the subjects language development apart from MLU.

44

Cantoneses UB values2 range from 8 to 29 whereas Englishs one does from 11 to 22. This suggests that the UB figures are comparable with each other.
40

30

No. of words

20

10

Cantonese English

0
5 8 5 9 4 3 7 9 6 5 6 7 9 7 .1 0 .1 4; 05 .2 .1 .1 .1 .1 .2 3; 10 4; 00 3; 07 4; 01 .1 4; 04 3; 06 3; 09 3; 02 4; 03 4; 06 3; 03 3; 08 3; 04 3; 11 4; 02 .0 .1 .2 .0 .1 .2 .2 7

Age

Figure 3.4 UB pattern for Kathryns Cantonese and English

It should be noted that the same cross-linguistic problem for MLU calculation applies for UB too, that is, with respect to English, Cantonese still bears the inflationary/deflationary morphological feature in calculating UB. LLYs UB ranges from 19 to 23 while Kathryns UB stretches from 8 to 26. Comparing Kathryn with the Cantonese monolingual child (LLY) for the comparable period (Figure 3.5), LLY shows a more steady change pattern than Kathryn. Despite of it, statistically their UB values for Cantonese do not show significant difference (p>.05). It shows that both children are able to produce utterances of much the same length. Indeed, they produce their own longest utterance in a similar fashion (2). 2. i. gam2 zau6 keoi5 keoi5dei6 pou5 zyu6 di1 sai3 zek3 maau1 He they hold ASP CL small CL cat gam2 zau6 go2 go3 maau1 le1 gam2 zau6 heoi3 heoi3 pou5 then that CL cat SFP then it it hold

A few transcript files contain over-long utterances due to transcription errors. Therefore, they are

taken out from the corresponding files and not included in the calculation of UB.

45

zo2 di1 maau1 gaa3 laa3 ASP CL cat SFP SFP They are holding some kittens, they have held some cats. (Kathryn, months)

40

ii. sin1saang1 ne1, gam2joeng6, gam2 ne1, keoi5 ne1, go2 go3 teacher SFP SFP he SFP that CL lat1, e6: ze2ze2 dou1 jau5 faan2 hok6, li4 li2 dou1 jau5 faan2 sister also have go to school also have go to hok6 ma3 school SFP Sister also goes to school. (LLY, 39 months) From the example, both utterances are expanded by gap fillers gam2joeng6, gam2 ne1 and gam2 zau6; and repetition of words conveying the main ideas like /pou5/ hold, /maau1/ cat and /faan2 hok6/ go to school.
30 25

UB (morphemes)

20 15 10 5 0 38 39 40 41 42 43 44 Kathryn lly

Age (months)
Figure 3.5 UB pattern for Kathryns Cantonese and LLYs Cantonese

Considering the Englishs UB values, the English monolingual child, Adam, yields UB from 12 to 22, which is similar to the range (11 -26) produced by Kathryn (Figure 3.6). Again, there is no significant difference (p>.05) between the values obtained by Kathryn and Adam. However, the complexity of their longest utterances seems not comparable to each other. See the following example:

46

3.

i. can I put my head in de mailbox # so de mailman can know where I are and put me in (th)e mailbox ? (Adam, 38 months) ii. you are the girl, you are the girl, I am the girl yellow Power_Ranger, you are the sun Power_Ranger, and and you are the pink Power_Ranger . . (Kathryn 44 months)

The longest utterance produced by Adam is composed of subordinate clauses, but that from Kathryns data is only repetition of simple clause. This finding might suggest that Adams English proficiency is higher than Kathryn in terms of structural complexity. However, Kathryn shows that she is also advanced enough in her English knowledge in order to produce subordinate clause at the age of 38 months (UB = 17) in example 4. 4. i. the tape, because I don't want to stick it on here because it won't work on there ii. I found it in my room <but> [/] but I found it in my [!] room so it's mine

30 25 20 15 10 5 0 38 39 40 42 43 44 45 46 47 48 49 50 51 52

UB (words)

Kathryn Adam

Age (months)

Figure 3.6 UB pattern for Kathryns English and Adams English

3.3 Findings of Lexical Diversity (D) The measure D is employed to assess the lexical development of the bilingual child. Looking at the face value of D shown in Figure 3.7, it appears that English is much more developed than Cantonese throughout the whole period of study.

47

120 100 80 D value 60 40 20 0

3; 02 .1 9 3; 03 .1 6 3; 04 .1 4 3; 06 .1 8 3; 07 .1 3 3; 08 .2 7 3; 09 .2 5 3; 10 .2 7 3; 11 .2 7 4; 00 .1 5 4; 01 .0 9 4; 02 .1 6 4; 03 .1 5 4; 04 .2 9 4; 05 .1 0 4; 06 .0 7

Cantonese English

Age
Figure 3.7 D pattern for Kathryns Cantonese and English

According to the results obtained in previous longitudinal studies, for Cantonese-speaking children the median D at age three is 41.06 whereas at age four it is 60.94 (Fletcher et al., 2000); English-speaking children in Bristol obtained 47.14 for age three and 53.8 for nearly age four3 (Malvern et al.; 2004). As Malvern et al. (2004) and Klee et al. (2004) suggested that wider range of situational contexts would likely lead to an increase in the value D, that D values for Kathryns English appear to lead ahead of those of monolinguals would likely be brought about by richer situational contexts created for Kathryn. Although both Malvern et al. (2004) and the present study collect data in the subjects familiar setting (i.e. their homes), the children might arguably perform differently. Children in Malvern et al. (2004) are set to interact with family members without the presence of researchers while Kathryn in this study mainly interacts with unfamiliar interlocutors. Therefore, besides talking about daily routines with the adult speakers, in some instances Kathryn would introduce and explain her family background as extra information to the interlocutors. This may be the reason to account for the higher D value for

The most appropriate age group available in Malvern et al. (2004)s study is 42 months.

48

Kathryn than that have been obtained in the monolingual subjects in Malvern et al. (2004)s study. Looking at Figure 3.8, it shows the D pattern of Kathryn and an English monolingual child, Adam who shares similar recording setting with Kathryn.
120 100 80

D value

60 40 20 0 38 39 40 42 43 44 45 46 47 48 49 50 51 52

Kathryn Adam

Age (months)
Figure 3.8 D pattern for Kathryns English and Adams English

Comparing the trend lines of Kathryns D values with those of the monolingual child, Adam, both lines appear close to each other. Kathryns D values range between 60.38 and 97.75 while Adams is between 71.06 and 101.86. Although both the lower and upper limits have about 10 points in difference, it is understandable as Kathryn can draw on resources from two lexicons in order to express ideas whereas Adam needs to acquire a richer lexicon in only one language. Nonetheless, as opposed to the difference of D value between Kathryns Cantonese and English, the developmental pattern in Kathryn resembles the lexical development of the monolingual child. There is no statistical difference between Adams and Kathryns D value for English (p>.05). These findings suggest that Kathryns English lexicon develops in a similar way to monolingual counterpart. Concerning Cantonese, Kathryns D values do not vary much from the median values in the monolingual study (Fletcher et al., 2000). Furthermore, by closely look

49

at a particular monolingual in Fletchers study, LLY (see Figure 3.9), it shows that Kathryns D values for Cantonese approximate to those of monolingual (LLY). LLYs D value is between 29.27 and 55.17 while Kathryns ranges from 30.42 to 49.31. The upper limit of LLY is slightly higher than that of Kathryns which is also explicable by the fact that Kathryns can resort to the resources from two lexicons. Nevertheless, the difference is too small to claim that Kathryn and LLY have a discernible lexical development with respect to Cantonese and there is no statistical difference between the two (p>.05); instead their lexical development in Cantonese is similar to each other.
60 50 40

D value

30 20 10 0 38 39 40 41 42 43 44 Kathryn lly

Age (months)
Figure 3.9 D pattern for Kathryns Cantonese and LLYs Cantonese

3.4 Discussion: Concepts of Balanced Bilinguals It has been always controversial to identify the notions of balanced bilingualism because it has been shown that bilinguals tend to be more proficient in one language in specific domains. Therefore, balanced bilingualism is considered as ideal but non-existent in this regard. On the other hand, studies about bilingual language development do not deny the notion of balanced as cases illustrated that some bilingual children attain parallel acquisition order of a particular grammatical

50

construction or obtain no clearly diverse values of quantitative measurements between the two languages. While bilinguals may show preference in using one language over the other in a specific domain, such as academic stuff or business matters, this by no means precludes the bilinguals being balanced with regard to language proficiency. Evaluation of the acquisition order of a certain grammatical construction is problematic for this present case. Firstly, by the time Kathryn was recorded, she was old enough to have acquired simple clauses like noun phrase and verb phrase. Since some areas of Cantonese grammar, especially higher level structures like IP/CP, have not yet reached consent among researchers, it may risk erroneous analysis when they are involved. Consequently, I investigate the notion of balanced bilingualism based on the measurements of language proficiency in the bilingual subject. Assessment of a childs bilingualism seems to vary very differently from study to study. At one extreme, Deuchar & Quay (2000) declined to measure language dominance for their subject because there are too many ways in doing so, rather they preferred to give relevant information for readers to interpret on their own. At the other extreme, language dominance is determined by a score which is generated from a number of quantitative indicators such as MLU, UB, MMU etc (Genesee et al., 1995; Deuchar & Muntz, 2003). In between, assessing bilingualism for children could be achieved by a range of methods such as impression of ones language proficiency, parental report, and developmental rate of certain syntactic structures or employing one or two of the above-mentioned quantitative indicators, typically MLU or MLU differential (Yip & Matthews, to appear). In measuring the language proficiency of bilingual, especially involving Cantonese and English, the aforementioned quantitative measures have encountered potential problems arising from contrasts between the two target languages. To make

51

them the most predictive and adaptable for analysis, some adjustments have to be applied to the interpretations of the data derived. In the case of Cantonese and English, the validity of the measures is established by comparisons with monolingual speakers of each language as well as direct comparison of two sets of absolute values. This is due to the fact that the two languages bear different inflection systems and the measures are sensitive to it. With these methodological adaptations, in principle there are two ways to show balanced development. Firstly, it could be shown that both languages resemble monolingual development at the ages concerned. Assuming that interferences are presence in various degrees between two separately developing languages (Mller, 1998), Kathryn seems to reflect a minimal degree of interaction in language usage. Second, the two languages may be shown to develop in parallel (to the extent that the comparisons are valid). Thus, as I have shown above, Kathryn is a balanced bilingual child of Cantonese and English based on the measures of MLU, UB and D. In addition, all three measures maintain at a steady level and do not show observable developmental changes (neither growth nor drop). This suggests that Kathryn may be approaching a ceiling that is near to adult-like linguistic performance, and thus the present case study is different from many previous studies which focus on early language acquisition (usually below age 3).

3.4.1

Qualitative evidence supporting the result Since there is a lack of compatible monolingual Cantonese subject to give a

valid comparison with Kathryns Cantonese proficiency, I will demonstrate that the bilingual subject is native-like in her Cantonese. As mentioned in the above section, Kathryn is near to adult-like linguistic performance; her language proficiency in terms of syntactic constructions is evaluated with reference to adult monolingual

52

Cantonese speaker. Kathryns Cantonese sounds natural and native during the whole period of study. Two areas of Cantonese grammar are chosen to show if they are appropriately manipulated in her utterances: sentence final particles (SFP) and classifiers. Sentence final particles (SFP) is important in Cantonese speech because they serve various pragmatic functions such as indicating speech acts, evidentiality and bring in affective and emotional coloring (Matthews & Yip, 1994). There are a set of basic forms in which each of them can be combined together in a restricted order. In Kathryns corpus, she is found to use a lot of SFP in her Cantonese utterances correctly (5). 5. i. CHI: hei2 san1 laa1 , keoi5 hei2 san1 laa1 . wake up SFP, s/he wake up SFP. Wake up. wake up! (3;08.27)

In example 5i, Kathryn uses laa1 to create an imperative sentence in order to ask the toy dinosaur to wake up. The combination of SFP lei4 ge3 in utterance in 5ii can entail explanatory or discovery function. ii. INV: cai3 dou2 zek3 gau2 ceot1 lei4, zung6 jau5 le1, ni1 assemble PRT CL dog PRT PRT also have SFP this di1 me1 lei4 gaa3 CL what SFP SFP (We) can assemble a dog, what else? CHI: hai6 jip6 lei4 ge3 be leave SFP SFP It is a leaf. (3;08.27) iii. INV: hou2 laa1 dang2 ngo5 gaan2 laa1. good SFP wait me choose SFP. Let me choose. INV: ngo5 soeng2 tai2: +... I want watch I want to watch

53

CHI: li1 go3 laa1 hou2 mou2 aa1? this CL SFP good not SFP. How about this one?

(3;02.19)

One other function of SFP is to make questions. The SFP aa1 is a common particle which serves to provide suggestion in the form of interrogative construction like A-not-A question as in 5iii. Kathryn suggested the interlocutor what to watch on TV by a question marking with aa1 and hou2 mou2, which is a tentative variant form of hou2 m4 hou2 (good-not-good). Example 5iv shows the resultant form zi1 maa3 of another particle combination ze1 and aa1maa3. The former expresses a meaning of just, only while the latter one indicates obvious reasons. This is used correctly within the discourse as Kathryn wanted to say that the size of her dress is not that big. iv. INV: waa1 , zan1hai6 hou2 daai6. PRT really very big. Wow, its really very big! CHI: li1 go3 siu2siu2 daai6 zi1 maa3. This CL little big SFP SFP This is only a little bit big.

(4;04.29)

The above examples have illustrated that Kathryn generally can make use of different SFP to convey different pragmatic functions. Another important feature of Cantonese syntactic is the use of classifiers to modify nouns. Each noun is assigned with a particular classifier. Kathryn was found using them as required in Cantonese grammar, though she might make mistakes sometimes like in the above example (5iv), what was said to be big is her dress, which normally takes the classifier tiu4 (denoting long, thin object), but never go3 (usually referring to non-concrete entities lacking physical features). However, at the very beginning of the recordings, Kathryn has used it correctly (6). 6. INV: hou2 laa1, li1 dou6 jau5 bi4bi1, bi4bi1 li1 dou6 jau5 saam1
54

good SFP here have baby baby here have clothes gaa3 SFP Heres a baby and the baby has clothes. CHI: hai6 aa3, jau5 tiu4 kwan4 gaa3 yes SFP have CL dress SFP Yes, (it) has a piece of dress. (3;01.05) In the case of go3, which was used extensively by Kathryn, it is used to denote people (7); and for items which do not call for a specific classifier (8) (Matthews & Yip, 1994). 7. CHI: mei6 aa3, mei6 hai6 sing1kei4saam1 aa3 not yet SFP not yet be Wednesday SFP Its not yet Wednesday. INV: jiu3 sing1kei4saam1 sin1 jau4 dak1 seoi2 gaa4 need Wednesday first swim PRT water SFP (You) can only swim on Wednesday? CHI: hai6 aa3, ngo5 go3 sin1saang1 waa6 yes SFP I CL teacher say Yes, my teach said so. (3;04.14) CHI: keoi5 go2 go3 uk1kei2 tung4 +/. he that CL home with His home

8.

(3;07.13)

Moreover, she demonstrates a typical use of classifier: an unexpressed noun is acceptable after the classifier when it is understood in the context (Matthews & Yip, 1994). In example 9, the interlocutor and Kathryn are talking about the toy Sailormoon, it is understandable to each other that they are not talking about any other toys when the interlocutor asked Kathryn if she had one. Therefore, Kathryn omitted the head noun Sailormoon in her reply to the interlocutor 9. CHI: <ngo5 jau5> [//] lei5 mou5 Sailormoon, ngo5 jau5 I have you not-have I have

55

Sailormoon I have Sailormoon, you dont have Sailormoon. INV: ngo5 dou1 jau5 gaa3, ngo5 uk1kei2 dou1 jau5 gaa3 I also have SFP I home also have SFP I have (one) too, I have (one) at home. INV: lei5 jau5 mou5 aa3. you have not-have SFP Do you have (one)? CHI: ngo5 jau5 jat1 go3. I have one CL I have one (Sailormoon). (4;05.10) By listening to Kathryns recordings, one could be surprised by her fluency in speech and accuracy tonal articulation, although it is rather impressionistic to justify her pronunciations without carrying out a systemic phonological analysis of her speech.

3.4.2

Possible Cross-Linguistic Influence In response to the argument that when two languages are in contact during

language acquisition, they are liable to cross-linguistic influence (Hulk & Mller, 2000), in this section I will look for possible cross-linguistic influence attested in Kathryn. Yip & Matthews (2000) found that their Cantonese-dominant bilingual subject, Timmy, passed through a stage where his English structures are influenced by his Cantonese counterparts. The authors contributed this finding to dominance as a causal factor for transfer in the domains of wh-interrogatives and null objects. Yip & Matthews (in prep.) have found that Kathryn produced the least amount (13.3%) of wh-in-situ questions in her English, compared with other bilingual children in the corpus: Timmy 92.3%, Sophie 31.3%, Alicia 42.9%, Llywelyn, 66.6% and Charlotte 25%. In the case of where-in-situ, the contrast of the finding is bigger.

56

Kathryn and Charlotte yield 0% whereas Sophie and Alicia get 100% where-in-situ questions; Timmy and Llywelyn get 20% and 33.3% respectively. This quantitative difference between the three Cantonese-dominant bilingual children, Timmy, Sophie and Alicia, and Kathryn supports the argument that transfer is determined by language dominance in that Cantonese structure exerts influence on Englishs. As opposed to this, Charlotte, an English-dominant bilingual child shows that she applied English wh-movement in mixed utterances (10)4 in which Charlotte seems to apply wh-movement to both languages. 10. i. Wheres dang6? chair Wheres the chair? ii. Bin1dou6 chair? where Wheres the chair?

(Charlotte 1;11.05)

(Charlotte 1;11.05)

In Kathryns corpus, cases of wh-movement in Cantonese are not found. Therefore, it conspires to argue that Kathryn shows only minimal cross-linguistic influence in the domain of wh-interrogatives. Qualitatively, the wh-in-situ questions produced by all the bilingual children are similar in that the English wh-in-situ structures match the Cantonese counterparts. Examples are given below: 11. i. Mine is what color? (Kathryn 3;02.19) ii. Ngo5 ge3 hai6 mat1je5 ngaan4sik1? I POSS be what color What color is mine? 12. i. I can hear what noise? (Kathryn 4;04.29) ii. Ngo5 ho2ji5 teng1 dou6 mat1je5 seng1? I can hear ASP what sound What sound can I hear?
4

Example taken from Lai (2005)

57

13. i. This one is what? ii. Li1 go3 hai6 mat1je5 aa3? this CL be what SFP What is this? 14. i. INV: He can bite CHI: Bite the what? ii. Ngaau5 di1 mat1je5 bite CL what What (do you) bite? 15. i. Claire has what? ii. Lei5 jau5 mat1je5? you have what What do you have? 16. i. INV: And this one is CHI: is where? ii. Li1 go3 hai6 bin1 dou6? this CL be where Where is it?

(Alicia 4;10.12)

(Timmy 2;08.18)

(Charlotte, 2;08.06)

(Llywelyn 3;00.27)

This qualitative similarity implies that the underlying cross-linguistic influence is the same for all types of Cantonese-English bilingual children. In the case of null objects, both English and Cantonese licensed optionality of objects of verbs in various degrees (Goldberg, 2001, Yip & Matthews, 2000). Huang (1999) and Yip & Matthews (2000) found Cantonese-influenced null object structure in English in a Cantonese-dominant child, Timmy. The use of existential have in 175 clearly shows a transfer of Cantonese grammar to English. 17. CHI: Zung6 jau5. still have There are still some (there).

(Timmy 2;05.05)

Likewise, the verb put directly followed by a locative phrase is a clear case of

The example is from Timmys parents diary data.

58

transfer from Cantonese to English grammar (18). 18. i. INV: Where shall we stick it? CHI: Put here. ii. Fong3 (hai2) li1dou6. put at here Put (it) here.

(Timmy 2;05.05)

In Kathryns corpus, the existential use of have is not attested. However, there are two cases of put which resemble the Cantonese grammar. In example 19, the first instance of put case appears to lead the second one to happen: since the first case of put is topicalized, the object is considered as overt so it is not a case of null object. Although English grammar has topicalization, it still expects overt element left in the canonical object position. Thus, this topicalized sentence is Cantonese-influenced. The subsequent case of put is regarded as null object because its object is absent, even it is likely to refer to the same item. It can be compared to another case kept in example 21. In example 20, Kathryn produced put here as much like what Timmy did (c.f. example 18). 19. INV: now let's play, let's play play+dough CHI: this, put on here. CHI: put on here and then it will be tea 20. INV: its Snoopy. INV: ah, so, who bought this for you? CHI: put here naked.

(3;02.19)

(3;04.14)

21. INV: now, where's your book? Kathryn, which book is yours? CHI: yes, and that book is mine, I kept on there (3;06.18) Although Kathryn and Timmy exhibit similar construction of null objects in their English utterances, quantitative results draw a difference between them.

59

Verb

Total No. of occurrence of each verb Kathryn Timmy 169 84 93 35 71 104 26 50 16 12

No. of null object Kathryn 2 0 5 2 0 Timmy 37 6 34 7 1

% of null object Kathryn 1.92% 0% 10% 12.5% 0% 4.88% Timmy 22% 7% 37% 20% 1% 20.4%

get like put take want

Average % of null object

Table 3.1 Frequency of null objects in Kathryns and Timmys corpora

As shown in Table 3.16, the frequency of null objects in Kathryns corpus is much lower than in Timmys case. One of the contributing factors to this discrepancy is that Kathryn is older than Timmy when the recordings were made: Kathryns range is between 3;02.19 and 4;06.07 while Timmys is between 2;04.14 and 3;06.25. Therefore, a developmental reason needs to be taken into account. Indeed, Yip & Matthews (2005) reported that the percentage of null objects in Timmys corpus drops from a higher range 9.1%-28.6% (2;04-2;08) to a consistently lower rate at about 5% (2;09-3;06). On the other hand, Kathryns rate of null object is close to that of an English monolingual child (3.5%) in Huang (1999)s study. It is to note that the high rate of null objects obtained in the case of put in both bilingual children has favored the assumption that Cantonese structure (put followed by locative phrase) exerts influence on its English counterpart7. Yip & Matthews (2000) argue that both input ambiguity and language dominance have played roles in the transfer phenomenon of null objects. In Yip & Matthews (to appear)s study, the authors argue that frequency of null objects appear to correlate with dominance in that less Cantonese-dominant bilingual child, Timmy,
6 7

Timmys data is taken from Yip & Matthews (to appear). Other bilingual subjects in Yip & Matthews (to appear)s study also have their highest null objects

rate for the verb put: Sophie 71% and Alicia 58.7%

60

shows lower rate of null objects than the other two more Cantonese-dominant bilingual children, Sophie and Alicia. As predicted by the authors argument, the present case of a balanced bilingual child receives lesser transfer influence in the domain of null objects. In the study of Paradis & Genesee (1996), they argue that episodic occurrences of some constructions do not necessarily indicate systemic interaction, such as transfer, at the level of competence. Such phenomenon is considered as on-line interaction between the two languages in performance. In the present study, although Kathryn shows the lowest rate of transfer in the domains of wh-interrogatives and null objects, qualitative evidence can demonstrate that the property of transfer is similar to those of Cantonese-dominant bilingual children. The lack of systemic (sustaining over a period of time) evidence for the transfer phenomenon may be due to the fact that Kathryn is linguistically more mature than other Cantonese-English bilingual children in Yip & Matthews (2000, in prep.) studies. To summarize, it is clear that the two languages, Cantonese and English, developed in a parallel fashion in the case of Kathryn, in both qualitative and quantitative aspects. Moreover, the present balanced bilingual case serves to support the argument made in Hulk & Mller (2000) that cross-linguistic influence would be observed in bilingual children as long as two languages are in contact. Regarding the comparisons of null objects and wh-interrogatives between a Cantonese-dominant bilingual child and Kathryn, one could generalize that cross-linguistic influence is more discernable in dominant-children than in balanced bilingual children and the directionality of the influence is likely to be determined by language dominance.

61

Chapter 4: Language Mixing by the Bilingual Child Subject

One of the major differences of language development between monolingual children and bilingual children is that bilingual children are able to draw on linguistic resources from two languages in comprehension and production. Hypotheses or theories on language mixing during language acquisition are usually based on data from bilingual children who show language dominance in one of the two languages. In this regard, the present study hopes to contribute to some of the existing hypotheses or theories concerning language mixing by examining mixing data of a balanced bilingual child. In Kathryns corpus, mixed utterances on the whole make up 5% of the corpus (386 out of 7774), while English and Cantonese share the rest (53% English and 42% Cantonese). Both grammatical and pragmatic aspects of her mixing utterances will be discussed in this chapter. In the literature, mixing has been assumed to be closely related to language dominance of bilingual children. Based on dominance, predictions on language mixing in bilingual children have been formulated: i. According to the Dominant Language Hypothesis (Petersen, 1988), grammatical morphemes of the dominant language may coocur with lexical morphemes of either the dominant or the non-dominant language, but not vice versa. ii. Genesee et al. (1995) hypothesized that children would mix more when speaking in their non-dominant languages than their dominant languages.

62

iii.

Mixing respects the grammar of the stronger (dominant) language more than that of the weaker (non-dominant) language

(Gawlitzek-Maiwald & Tracy, 1996; Bernardini & Schlyter, 2004). As Kathryn is considered as a balanced bilingual child based on the quantitative measures shown in chapter 3, based on i, she is expected to show bi-directional mixing of grammatical morphemes from both languages, such as1: 1. English grammatical items in Cantonese utterance keoi5 ge3 sing4zik1 mei6 can4 above gwo3 average He POSS result not ever ASP His result has not been above average. Cantonese grammatical items in English utterance Im speaking of go3 cost, m4 hai6 functionality CL not be Im speaking of the cost, not functionality.

2.

In the above examples, grammatical morphemes refer to close-class items above, go3 and m4 hai6 and they do not assign or receive semantic roles (c.f. Lanza, 1997). Concerning prediction ii, Kathryn is expected to mix equally (in case she does) when conversing with interlocutors of both languages during recording. Unexpectedly, however she produced an asymmetrical mixing pattern whereby she mixes more frequently in Cantonese than in English as matrix language. Kathryns mixing rate across the recording period is presented in Figure 4.1.

Examples are taken from Leung (2001).

63

Kathryn's Mixing in Each Matrix Language

120.00% 100.00% 80.00% Proportion 60.00% 40.00% 20.00% 0.00%

English as ML Cantonese as ML

Figure 4.1 The Proportion of Kathryns Mixing in Each Matrix Language

As the result shows, mixing is generally more frequent in Cantonese than in English as matrix language. Only one data point (at 3;11.27) shows that mixing is more frequent in English as matrix language (56.25%). There are 197 utterances in Cantonese in which 7 are mixed, whereas 9 out of 202 English utterances are mixed. In fact, one of the mixed utterances in Cantonese as matrix language is a long sentence which contains 7 cases of mixing. Because of methodological reason, this long sentence is counted as mixed once only. Therefore, I would argue that Kathryn is consistent in using more mixing in Cantonese as matrix language. Another data point (at 3;09.25) exhibits frequencies of mixing in the two languages which are very close to each other. Mixing in English as matrix language amounts to 46.67%. There are 14 instances of mixing in which 11 are involved with cartoon names. This means the relatively high rate of mixing is induced by reasons apart from language dominance (more discussion in section 4.2). Nonetheless, this asymmetrical pattern may lead one to suspect that Kathryn might be an English-dominant bilingual

3; 02 .1 3; 9 03 .1 3; 6 04 .1 3; 4 06 .1 3; 8 07 .1 3; 3 08 .2 3; 7 09 .2 3; 5 10 .2 3; 7 11 .2 4; 7 00 .1 4; 5 01 .0 4; 9 02 .1 4; 6 03 .1 4; 5 04 .2 4; 9 05 .1 4; 0 06 .0 7

Age

64

according to prediction ii. According to the quantitative measurements examined in chapter 3, however, there is no evidence that Kathryns English is dominant. Therefore, I argue that the occurrence of Kathryns mixing utterances is influenced by another factor, namely the adult input. Prediction iii distinguishes stronger and weaker languages on the basis of asynchronic syntactic development of the two languages, for instance, one of the language pairs has developed IP earlier than the other one, and therefore it is considered as the stronger language concerning that domain. However, since asychronic syntactic development is not found in Kathryns corpus, the prediction iii cannot be testified.

4.1 Utterances with Lexical Mixing In this section, I will first investigate the structure of the mixed utterances with Cantonese as the matrix language, since this type of mixing is predominant among all the mixing utterances. Then I will examine the reverse pattern, that is, English is the matrix language mixed with Cantonese items. Subsequently, the result can test prediction i. Lexical mixing in this study is used to describe mixings which happen on the lexical level, as opposed to syntactic or phonological mixing. Hence, it does not make a distinction between open and closed-class items (c.f. Lanza, 1997). In the results, it is reported that mixing utterances are mostly noun types (Figure 4.2).

65

90 80 70 Percentage 60 50 40 30 20 10 0 Nouns Verbs

Cantonese as matrix language English as matrix language

Adjectival

Others

Grammatical Categories

Figure 4.2 Percentage of Each Grammatical Category of Kathryns Mixed Utterances

4.1.1

English Nouns in Cantonese Utterances The noun or noun phrase is the most frequent type of mixed item for a wide

range of language pairs (Chan, 2003). This is also exemplified in this case study. Chan (2003) claims that English noun/noun phrase can either exist as in bare form or be marked with number in Cantonese utterances. They may also be embedded within a higher structure like Cantoneses demonstrative phrase, quantifier phrase or classifier phrase. For example: 3. CHI: lei5 geng1 m4 geng1 puppy gaa3?(3;02.19) you scare-not-scare SFP Are you scare of puppies? CHI: ngo5dei6 jau5 [/] jau5 loeng5 go3 turtle gaa3 (3;02.19) we have have two CL SFP We have two turtles! CHI: keoi5 soeng2 daa2 di1 Hyena aa3. (3;08.27) He want hit CL SFP He wants to hit the group of Hyenas.

4.

5.

66

6.

CHI: keoi5 jau6 di1 stripes. He have CL He has some stripes.

(4;00.15)

7.

CHI: Winnie_the_Pooh zou6 mat1je5 aa2? (3;02.19) do what SFP What is Winnie-the-Pooh doing?

Example 3 shows that the English noun appears in its bare form in a Cantonese sentence. However, the noun puppy would be expected to be in plural form according to English grammar as it is used in a generic sense, consider example 8, showing that Kathryn is able to use plural forms to signify generic sense in English. 8. INV: so, ah <what do you> [/] # what do you have <for> [/] for him ? CHI: eggs (3;03.16)

In adult mixing, an English plural noun may appear in its singular form as it has to be congruent with the Cantonese grammar in that Cantonese noun is not inherited with the number feature. See the example2 below: 9. Christmas tree di1 decoration CL the decorations of a Christmas tree

The bare noun turtle of example 4 is likely to be accounted for in the same way, although the quantifier loeng5 implies plurality3. Both the nouns Hyena and stripes are modified by the Cantonese classifier di1 in examples 5 and 6. The meaning of di1 is similar to some in English. Hyena is a proper name of a cartoon character so it is not inflected for number. On the other hand, stripes is marked plural as it does not clash with the plurality feature

Example is taken from Chan (2003). Chan (2003) provides a more detailed analysis on the structure of noun type mixing spoken by Cantonese-English bilinguals in Hong Kong. Chan (2003) has found that English nouns may appear in either singular or plural forms, regardless of whether the numeral/classifier denotes plurality or not.
3

67

denoted by the Cantonese classifier di1 (Chan, 2003). For a proper noun like Winnie-the-pooh in example 7, its structural features are congruent to the Cantonese noun phrase structure in that it does not associate with number by inflection. Thus it does not generate any grammatical problem in Cantonese-English mixing.

4.1.2

English Verbs in Cantonese Utterances Compared to noun type mixing, Kathryn showed much less verb type mixing

between Cantonese and English (see Figure 4.2). One salient contrastive syntactic feature between the two languages is that English verbs are marked with tense/agreement by inflection, while Cantonese verbs are marked with aspect lexically. Chan (2003) pointed out that in most cases of Cantonese-English mixing, the English verbs are the infinitive or root forms although they would be required to be inflected in English grammar. This construction is also illustrated in Kathryns mixing as in examples 10 and 11. In example 10, the English verb swallow is in its root form as it is preceded by a Cantonese modal verb (in A-not-A structure) which is treated as an I-element4. In example 11, it is illustrated that the English verb change is not inflected for person in agreement to the subject keoi5 which is corresponding to the third person singular in English because it is morphologically adapted to Cantonese verb construction. 10. CHI: ho2 m4 ho2ji3 swallow keoi5 gaa3? can-not-can it SFP Can (1) swallow it? 11. CHI: keoi5 hai6 change back to little. it be it changed back to little.
4

(3;03.16)

(4;03.15)

Cantonese sentences are assumed to be headed by I (Aspect) and thus Cantonese verbs are c-selected by aspect markers (Chan, 2003).

68

Another common feature of English verb mixing is that English verbs are root forms and are followed by Cantonese aspect markers like zo2 (perfective). In example 12, the verb plant is c-selected by the aspect marker zo2 so it is uninflected for tense agreement. 12. CHI: hai6 aa3 keoi5dei6 plant zo2 go2 di1 gaa3 laa3. yes SFP they ASP that CL SFP SFP yes, they have planted those. (3;06.18) Example 13 shows that an English verb is not in its root form: the verb catch attached with a past tense /participle marker -ed/ so as to bear passive meaning. The main verb in the utterance /soeng2/ want has been covertly marked with aspectual feature and therefore the English embedded clause be catched is in the infinitive form. Nevertheless, this example also serves as a case of morphological error, the overgeneralization of past tense marker: catch is attached with -ed instead of caught 13. CHI: ngo5 m4 soeng2 be catched aa1 maa3. I not want SFP SFP I dont want to be caught. 14. CHI: Lei5 bump into my fei4 tou5 laam5. you fat belly You bump into my fat belly. (4;02.16)

(4;00.15)

15. a. CHI: dim2gaai2 lei5 throw li1 go3 away ge2? (3;07.13) why you this CL SFP Why did you throw this away? b.. CHI: dont take all the photos out. (4;01.09) In contrast to the entire English VP in example 11, example 14 & 15a show that the VPs are partially in English. These two examples appear to demonstrate a two-step mixing: [IP Can [VP Eng [NP Can]]]

69

Given the above structure, I assume that the matrix language is switched at the boundary of phrases. Since the highest level (IP) of the utterance is lexicalized in Cantonese, the matrix languages of these two mixed utterances are Cantonese. However, the matrix language is switched to English across the boundary of VP. Hence, these examples require a two-fold analysis for their structures. First, consider only the higher part of the sentence [IP can [VP eng]], being consistent with other mixed utterances of English verbs, the verbs bump in 14 and throw in 15 are in their root forms. Therefore they did not forfeit grammaticality. Secondly, the matrix language is switched to English in the lower part of the sentence in examples 14 and 15 (i.e. [VP eng [NP can]]). In example 14, the insertion of /fei4tou5laam5/ fat belly into English DP does not cause ungrammaticality because the structure of DP is congruent in English and Cantonese (c.f. example 18 in section 4.1.4). Example 15 is different from the other examples because it involves mixing of the phrasal verb take away into a Cantonese utterance. According to Ho (2003), monolingual English children prefer to use the structure [V +NP +PRT] over another possible pattern [V +PRT +NP]. In the example (15), Kathryn indeed produces [V +NP +PRT] structure although the object NP is lexicalized in Cantonese as /li1 go3/ this one (c.f. example 15b). Aside from the two-step mixing assumption, Chan (2003 p. 195) has postulated four5 possibilities of mixing between verbs and objects in Cantonese-English in which an English verb complemented with Cantonese NP is one of the frequent types that are attested in his corpora.
The four possibilities are: a. V(can)+ DemP (i.e. [Dem (Num) CL NP (eng)]; b. V(eng)+ DP (i.e. [D(=Art) NP]; c. V(eng) + DemP; d. V(can)+ DP (eng); Chan (2003) found that a.and c. are much frequent than b. and d.
5

70

4.1.3

Other English Syntactic Categories in Cantonese Utterances In Figure 4.2, the results have shown that the noun and verb types of mixing

out-number the mixing of other syntactic categories such as adjectives. Consider example 16, where the verb hai6 takes an adjective as its complement and so dead is appropriately put in place. It is to note that the intransitive verb hurt is also placed correctly in Cantonese grammar. 16. CHI: Keoi5 m4 hai6 dead, keoi5 hurt zi1 maa3. He not be he SFP SFP He is not dead, just hurt. (4;03.15)

Other English lexical items that are mixed into Cantonese utterances included the formulaic expressions yes, byebye, please, no and numbers. In some cases, Kathryn was asked to spell out some words so English letters also appeared in Cantonese utterances (17). 17. INV: lei5 maa1mi4 dim2 cyun3 aa3 ? you mommy how spell SFP How do you spell your mommys name? CHI: ngo5 go2 go3 # ngo3 meng2 jau6 hai6 M U M M Y aa1 I that CL I name have be SFP maa3 SFP Her name is M-U-M-M-Y. INV: gam2 # zung6 jau5 le1, Kathryn dim2 cyun3 aa3 ? still have SFP how spell SFP How about Kathryn? CHI: <ngo5 hai6> [<] K A T H R Y N aa3 I be SFP Im K-A-T-H-R-Y-N. (4;04.29)

71

4.1.4

Cantonese Nouns in English Utterances The overall mixing rate is comparatively very low when English is the matrix

language. Indeed, to my knowledge, there are very few studies focusing on the syntactic structure of Cantonese components being inserted into English utterances (c.f. Chan, 2003). This is likely due to the fact that such kind of mixing is not often found to occur among Cantonese-English bilinguals in Hong Kong, as will be further discussed in section 4.2. In this case, I will illustrate that most of Kathryns mixings in English as the matrix language are grammatical with respect to English grammar. Firstly, let us look at the noun/noun phrase. In reference to the Morpheme Order Principle subsumed under the Matrix Language Frame model of mixing proposed by (Myers-Scotton, 1993), surface morpheme order will be that of the ML in ML + EL constituents and therefore examples 18 to 20 are allowed to occur. 18. CHI: This one is a hou2 jan4. good person This one is a good person (3;10.27)

19. CHI: Im the doctor, and she is the gu1 loeng4 (3;11.27) nurse Im the doctor and she is the nurse. 20. CHI: maa4maa4s bed is with me. Grandmas Grandmas bed is with me. (4;04.29)

Example 18 and 19 show that the placement of the Cantonese NP is congruent to the English structure as Cantonese noun phrase is embedded in the DP projection in English. Moreover, the adjectives of both Cantonese and English are prenominal so /hou2/ good in example 18 does not create any structural clash. In example 20, only the Cantonese head noun is inserted into English DPs and

72

so they are grammatical with respect to English grammar. Besides, example 20 clearly demonstrates that the embedding of the Cantonese noun is adapted to the English grammar since Cantonese does not have inflection and the possession construction with clitic is unique to English. Nonetheless, Chan (2003) has claimed it is rare for English verbs to select an object which is an English DP. Instead, it is common for English verbs to select a DemP6 in Cantonese.

4.1.5

Cantonese Verbs in English Utterances As outlined in section 4.1.2, there may be clashes in terms of syntactic features

between Cantonese and English verbs. Conversely, mixing Cantonese verbs into English utterances seems to be more acceptable in the case of Kathryn. Consider example 21, the verb bin3gyun1 is neutral in tense agreement. At the same time, the subject you selects a verb which is covertly marked with tense agreement. Therefore, we may assume that covert tense marker is still in the English I and so the Cantonese verb can remain as inflection free. 21. CHI: You ought to run and then you bin3gyun1 You ought to run and then you change (3;09.25)

In example 22, the Cantonese verb /zuk1/ catch is uninflected in the English utterance. It does not cause problem to the English structure as it is projected within the English IP headed by the auxiliary verb can. 22.
7

CHI: no, the wild doggie can get the pig . INV: the wild doggie can +/. CHI: can zuk1 the pig catch can catch the pig.

(4;02.16)

6 7

DemP= Demonstrative Phrase This mixed word was wrongly transcribed as can chop the pig.

73

4.1.6

Other Cantonese Syntactic Categories in English Utterances Mixing of syntactic categories other than nouns and verbs is much rarer. Some

examples are illustrated below: 23. CHI: Yours is tau3ming4, mine is white and pink Yours is transparent, mine is white and pink (4;05.10)

Example 23 shows that the Cantonese adjective fits well into the position of English adjective. It may be argued that their structures are congruent8 to each other and there is no problem for it to appear. In the following example, the Cantonese adjective ging6 is placed correctly inside an English noun phrase. 24. CHI: go to let you see this ging6 Power-Ranger (4;00.15) Go and let you see this powerful Power Ranger. 25. CHI: tomorrow I have to go to school but seng1kei4ng5 I dont have to go to school. (3;10.27) Tomorrow I have to go to school but on Friday I dont have to go to school A time adverbial /seng1kei4ng5/ Friday is mixed in the sentence in example 25. As stated in the Morpheme Order Principle, the placement of the Cantonese adverbial should adhere to the English grammar. Hence, seng1kei4ng5 should be preceded by an English preposition on. However, it is not the case in Kathryns mixing. Recall the analysis of English prepositions in Cantonese-English mixing produced by adult bilingual speakers (Chan, 2003), the present data is erroneously ruled out by the model proposed in the study. Therefore, I assume that prepositions raise potential problems which preclude the resulting structure of mixing from respecting the grammar of the matrix language; in this case, the insertion of the

Chan (2003) argues that Cantonese and English are not congruent in adjective construction in terms of morphological realization as the author only considers the derivational kind of English adjectives (i.e. V + ed/ ing).

74

Cantonese adverbial does not follow the grammar of English9. In addition, to highlight one more characteristic of the mixing in English context, Cantonese sentence final particles are scarcely employed together with English utterances. From the viewpoint of the claim that sentence final particles are projected to the IP positions in Cantonese, Kathryn demonstrates supporting evidence for separate language development (as she makes use of English syntactic structure for English utterances) during the language acquisition process. In this section, the grammaticality of Kathryns mixing has been illustrated: most of Kathryns mixing has been shown to be grammatical with respect to adult Cantonese-English mixing (Chan, 2003). Concerning prediction i, although Petersen (1988) only made generalizations of mixing pattern for dominant children, this case study has shown that grammatical morphemes of both languages can appear in either language: Cantonese aspect marker is attached to English verb whereas English possessive marker is attached to Cantonese noun (examples 12 &20). This implies that Kathryn does not show language dominance and lends support to my claim that she is a balanced bilingual child.

4.2 Discussion: Motivations for Mixing 4.2.1 Input Properties Based on my finding that Kathryn is a balanced bilingual child, the asymmetric mixing pattern seems not to be triggered by language (in)competence. I will demonstrate how the input in the majority of the bilingual community in Hong Kong affects her language acquisition, given that input is seen to be one of the influential factors to children language development.
According to the opinion from an English native speaker, it is acceptable to say Saturday I dont go to school in which on is omitted before Saturday in colloquial English. If it stands, the utterance in example 23 would be ruled as grammatical in English grammar.
9

75

This case study recorded mainly the conversation between the bilingual subject and the interlocutors, instead of parents. Although her family members, including her two elder brothers, mother and grandmother, were featured in the recordings at times, the data is too sparse for an analysis of parental input to be impractical. Consequently, the input in the interlocutors becomes the prime concern of the present analysis. Both of them are Cantonese-English bilingual speakers, serving as the models of typical bilinguals in Hong Kong. Lanza (1997) proposed parental discourse strategies by which the parents respond to language mixing in bilingual children. In her study, bilingual subjects interacted with their primary caregivers, their parents, through daily routines and thus each parent presumably developed their respective interactional styles with their children. In fact, this is the case in Lanzas study. Concerning the present case study, the interlocutors only visit the subject several times a month; it is questionable whether the interlocutors and the subject could form a clear pattern of interaction as consistency of language choice from both sides of the party is on a knife-edge. Moreover, as mentioned in chapter 2, the two interlocutors basically display a bilingual identity in front of the subject and carry out bilingual communication most of the time. In terms of Lanzas parental discourse strategies, the interlocutors in the present study often react to the subjects mixing by Move On Strategy (example 26) and Code-Switching (examples 27 -28), as indicated by the arrows. 26. CHI: < m4 hai6 > [<] aa3 li1 go3 hai6 ji1saang1 not be SFP this CL be doctor (He) is not, this is a doctor. CHI: li1 go3 hai6 nurse this CL be This is a nurse. INV: o1:, hai6 wo3 Oh, be SFP

76

Oh, (she) is. INV: zan1hai6 jat1 deng2 mou2 gaa3 # go3 wu6si6. really one CL hat SFP CL nurse The nurse really wears a hat. (4;04.29) 27. CHI: and a man want to sit on a horse. INV: mm, and a man who wants to sit on the horse. INV: how about this +... INV: what is this one called. CHI: hoi1 m4 dou2 aa1, hoi1 m4 dou2 aa1 open not PRT SFP open not PRT SFP (I) cant open it, (I) cant open it. INV: hoi1 dou2 aa1 open PRT SFP (It) can be opened. (30;9.25) 28.
10

CHI: hai6 mat6fung1 lei4 gaa3 go3 butterfly be bee SFP SFP CL The butterfly is a bee. CHI: hai6 mat6fung1 lei4 gaa3 be bee SFP SFP (It) is a bee. INV: <mat6fung1> [<] hai6 m4 tung4 butterfly go3 bo3 bee be not with SFP SFP Bees are different from butterflies. (3;02.19)

In some occasions, the interlocutors initiated mixing even though the bilingual subject maintained monolingual speech. For example: 29. CHI: ngo5 sik1 waak6 cin2 gaa3 I know draw money SFP I know how to draw money. INV1: cin2? money Money? INV1: mat1je5 cin2 aa3? what money SFP

10

The child has wrongly correlated the Cantonese word /mat6fung1/ bee with the English word butterfly

77

What money is it? INV2: coin CHI: bat1jyu4 waak6 +... how about draw How about drawing... INV2: coin haa6 maa5? SFP SFP coins?

(4;00.15)

30. CHI: I [!] got to hide. (4;05.10) CHI: lei1 maai4 lok6 cong4 dai2 aa1 hide PRT bed bottom SFP Hide under the bed. INV: oh # I can't INV: because I'm too big INV: look. CHI: you can. INV: yes INV: aai1jaa3, m4 dak1 bo3, bat1jyu4 cing1waa1 Oops not PRT SFP how about frog lei1 maai4 laa1 . hide SFP Oops, (I) cant. How about the frog hides (under the bed)? The examples above illustrated that a) the interlocutors negotiate a more bilingual rather than monolingual context with the subject by using Move On Strategy and Code-Switching especially when both of them are present within the same recording; b) the interlocutors are free to switch from a monolingual to bilingual type of conversation with the subject and c) the subject has no problems handling with these instantaneous switching of interaction styles. Based on the observations a) and b), I argue that the subjects mixing behavior is explicable within a wider framework (the mixing behavior of Hong Kong bilingual adults) rather than a localized one (interaction with interlocutors) as the mixing behavior of the interlocutors indeed represent a typical phenomenon in Hong Kong speech

78

community. Mixing as a Hong Kong linguistic phenomenon has been extremely studied by researchers (Li, 1998, Luke & Richards, 1982; Luke, 1998, Tse, 1992 etc.). Tse (1992) notes that in informal conversation, code-mixing between Cantonese and English is almost ubiquitous in Hong Kong, with Cantonese being the dominant code (p.101). Tse (1992) and Luke (1998) both claimed that one of the reasons of mixing in Hong Kong is that mixing can fill up lexical gaps, especially in fields associated with western culture. Moreover, as English is seen as a prestige language in Hong Kong, people may tend to mix in some instances of English into Cantonese utterances in order to show off or to define their social identity. In this sense, mixing embodies linguistically packaged knowledge which conveys one or any of the

above-mentioned ideologies. A language socialization approach anticipates that children would gradually be socialized into this mixing behavior along the way they acquire language in Hong Kong. Indeed, Kathryn has demonstrated that she is socialized in this way, as evidenced by the mixing rate in English. If mixing is syntactically motivated, as for prediction ii, it would be expected that Kathryn would mix to a similar amount in both Cantonese and English since she is shown to be a balanced bilingual child. In addition, a longitudinal study of mixing in a Cantonese-dominant child, Sophie, found many mixing instances that are structurally similar to those of Kathryns (Kwok, 2004). For example: 31. INV: waa3, hou2 wat6dat6 Woo, very gross Woo, (it is) very gross. CHI: tai2 haa5 go3 witch sin1. Look ASP CL first Look at the witch first

(Sophie 2;10.24)

In 31, an English bare noun is embedded in Sophies Cantonese utterance. The

79

English noun is modified by a Cantonese classifier go3 as same as Kathryns example (4) in section 4.1.1. Example 3211 illustrates a case of verb type mixing. Like Kathryn (12), Sophie attached the mixed English verb wear with the Cantonese aspect marker zo2. 32. MOT: Angel go2 gin6 dim2gaai2 m4 zoek3 aa3? that CL why not wear SFP Why dont you wear that one with Angel? CHI: Jan1wai6.. ngo5 kam4jat6 ji5ging1 wear-zo2 laa3. Wear because I yesterday already ASP SFP gam3 do1 ci3 zou6 mat1je5 aa3? so many CL do what SFP Because I wore it yesterday, why should I wear it so many times? (Sophie 5;01.3) Kwok also claimed that his case study shows the reverse of the results predicted by language dominance, that is, higher mixing rate is observed in Cantonese than in English. Therefore, dominance does not seem to correlate with the mixing pattern in both Sophies and Kathryns corpora. Chan (1993) has pointed out that in most cases English is only learnt by the time the children go to kindergarten after they have acquired nearly full competence in Cantonese. It is less common, though not non-existent, to observe an instance of language mixing in which a Cantonese word/phrase is incorporated in an English utterance. As the majority of the Hong Kong community is made up of this type of speakers, in the perspective of the language socialization approach, children with developing language may be able to observe this practice (seldom mixing Cantonese words in English utterances) and ultimately pick up this practice. In the light of the fact that Kathryn rarely mixes in English context, I argue that Kathryn shows accommodation to this norm.
11

This example is taken from the subjects parents diary.

80

Then, what is the motivation(s) for Kathryn to act against this norm by producing the existing mixing? Although the mixing data in English context are relatively sparse, there are a few salient characteristics that may easily be identified. Firstly, the majority of the mixed items are nouns (48 out of 68) and most of which are proper nouns and kinship terms. It is worth noting that the retention of Hokkien kinship terms has been found in the case of Baba Malay, a creole language based on Malay with a Minnan substrate (Ansaldo & Matthews, 1999). They proposed that kinship term is one of the culture-heavy lexical items which meanings are too specific to be provided by the lexifier language. This reason may be applied in the present case study as Cantonese does make a finer distinction of kinship relationship than English does. For example, Kathryn has used /maa4maa4/ to refer to paternal grandmother in English utterances as in English grandma does not offer such specific meaning. One may suspect that the occurrence of Cantonese kinship terms for grandfather and grandmother in English context is due to lexical gap. However, this possibility is eliminated as the word grandpa has been used by Kathryn in the first transcription file (33) while grandma is in the second transcription (34). 33. CHI: <but don't> [/] don't just disturb <my> [/] my grandpa <because> [/] because he's asleep. (3;02.19) 34. INV: so who's going to take care of you when Mommy is away? CHI: my grandma. (3;03.16) Apart from the cultural motivation, based on the fact that Kathryn addresses her paternal grandparents, who can only speak Cantonese, in only Cantonese, she then extended her addressing habit into English context, as shown in the following example. 35. CHI: maa4maa4's house. INV: yes, Granny's house, right?

81

CHI: no, maa4maa4's house.

(4:06.07)

It is notable that Cantonese kinship terms for grandparents are used in a limited way as shown in 35. Kathryn seems to use the particular word to refer to her own grandma(s) only but not the others. Therefore, she refused the translation equivalent that the investigator gives during the conservation. Interestingly, at a later stage of recording, Kathryn is found to assimilate the pronunciations of Cantonese kinship term for each of the grandparents into English phonology in order to address them within English context as in example 36, and even unable to correlate the terms for both languages as in example 37. 36. CHI: her English name is maa1maa3, and the Chinese name is maa4maa4. (4;01.09) 37. CHI: because it's Alastair's and Alastair always sleeps there. INV: then how about Grandma's bed? CHI: Grandma's, I don't have any grandmas. INV: maa4maa4. INV: what about maa4maa4's bed? CHI: maa4maa4's bed is with me (4;04.29) Furthermore, among Chinese families in Hong Kong, kinship serves as a core social concept which was extended to non-familial relationship. In informal situations, it is not unusual to be addressed by kinship form of address along the scale of intimacy. For example, a child addresses someone older than him/her by following the name of the addressee with ze4ze1 sister or go4go1 brother. It is unnecessary that the speaker and the addressee have sibling relationship, and it may even be the first time they meet. As this practice of addressing appears in the input to Kathryn, in many occasions, mixing is the result of the insertions of these kinship terms. Hence, it is another way to show respect to the social norm in the community. For example: 38. INV: <shall we go on> [/] shall we go on to play it?
82

CHI: I don't www # let's play with Winnie ze4ze1. INV: play with Winnie ze4ze1, where is she?

(3;07.13)

39. Kathryn would address the investigators as Winnie ze4ze1 and Linda ze4ze1 even in English context, as illustrated by 38. To summarize, given a mixing pattern which is unexpected based on prediction ii in section 0 , it is likely, as I argue that the use of mixing in both languages has been integrated into Kathryns language usage as she mixes frequently when speaking Cantonese and rarely mixes when speaking English. Yip & Matthews (to appear) state that mixing is a general feature of most bilingual societies which is taken up as part of their grammar in response to the input.

4.2.2

Other Motivations for Mixing In the study of Kppe & Meisel (1995), they pointed out that bilingual children

use intra-sentential mixing from approximately 3;0 onwards to serve some complex functions. In the present data, Kathryn is old enough (3;02.19 4;06.07) to be observed for the development of pragmatic and metalinguistic knowledge in terms of mixing. She was first observed to make metalinguistic comments at age 3;07.13, which coincides with the finding in Kppe & Meisel (1995)s study. As Kathryn has built an identity of being bilingual, she is likely to make use of mixing as a tool to deal with the problematic spot during conversation. In example 40, the investigator could not sort it out what Kathryn had just said, and Kathryn realized that this problem may arise from the tonal system of Cantonese (/mei5/ tail and /mei4/ eyebrow). Therefore she chose to replace it with the English equivalent term and finally succeeded in clarifying it by mixing. 40. CHI: <bat1gwo3> [<] keoi5 keoi5 zung6 jau5 go3 mei5 gaa3. but it it still have CL tail SFP but, it still has a tail.

83

INV: zung6 jau5 go3 mei4 gaa4. still have CL eyebrow (it) has a eyebrow? INV: li1 tiu4 aa4 ngaan5mei4 aa4. this CL SFP eyebrow SFP This one? Eyebrow? CHI: m4 hai6. not yes No. INV: go3 mei4? CL eyebrow An eyebrow? CHI: zung6 jau5 go3 tail gaa3. still have CL SFP It still has a tail! INV: go3 mei5, o1, jau5 tiu4 mei5 ge3 CL tail have CL tail SFP A tail, oh, (it) has a tail.

(3;10.27)

Some mixing examples like 41 & 42 demonstrate that she has language preference given that she recognized her bilingual ability. She clearly demonstrated that she prefers to include the Cantonese names of the cartoon characters /ng5seng1zin3deoi6/ Power Rangers and /mei5siu3neoi5zin3si6/ Sailor Moon although she might not be lacking the English equivalents. Having this language preference on cartoon names is the main reason behind the comparatively high frequency of mixing in English as matrix language at 3;09.25. 41. CHI: Who wants Power_Ranger. to be the ng5seng1zin3deoi6, (3:11.27) the

42. INV: oh, it's a Barbie. CHI: some kind of Barbie, it's mei5siu3neoi5zin3si6 Barbie. INV: is that <Sailor> [/] Sailor_Moon_Barbie? INV: no. CHI: it's exactly same. (4;04.07)

84

Moreover, Kppe & Meisel (1995) reported that the bilingual subject, Ivar, from 4;4 onwards, began to use marked language choice in order to make fun out of playing with the two languages. In this data, Kathryn performed this language use (example 43) a bit earlier, at the age of 3;09.25. She coins new noun phrases with both English and Cantonese modifiers. 43. INV: the sun will rise every day, every morning, except rainy days and cloudy CHI: and windy day INV: and windy day and cloudy day CHI: and the cloudy day CHI: and a foot day, and a head day, and a eye day <and> [/] and a nose day <and a> [/] and a bei6si2 day, and a tau4faat3 day, and # and a ngaa4ci2 day, and saam1 day. and a foot day, and a head day, and an eye day and a nose day and a boogie day and (a) hair day and a tooth day and clothes day. (3;09.25) Among other factors, the existence of lexical gaps is still an important one to explain mixing behavior in bilingual children. In examples 44 and 45, the meanings of the Cantonese items are relatively special in the language and hence Kathryn might tend to preserve the meaning by incorporating them into English utterances. 44. INV: I like to read some books here. CHI: ok, you can look at it, lei5 haam4sap1. ok, you can look at it, youre so lecherous 45. CHI: no, Winnie is Miss ba ba. INV: Miss bla bla. CHI: that means Winnie is fong3 cau3pei3. that means Winnie farts 4.3 Summary of the Analysis I have shown that Kathryn, like dominant bilingual children, does exhibit

(4;06.07)

(4;04.29)

85

mixing during her language development. In most cases, the mixed utterances are grammatical with respect to mixing constraints. This is to say that Kathryn has acquired adequate grammatical knowledge to perform mixing. Moreover, Kathryn, like children with language dominance, yields an asymmetric mixing pattern in that mixing is predominantly in one dimension. The most likely reason accounting for this is that Kathryn is socialized into the common usage of mixing within the speech community in Hong Kong: mixing English items into Cantonese but seldom in the other way round and this forms part of her pragmatic knowledge. Apart from this, since Kathryn has established a sense of bilingual identity from the age of 3;07.13; she intentionally uses mixing to carry out a number of functions, like clarifying problematic spots in bilingual communication or overcoming lexical gaps. All these functions are sustained by her metalinguistic knowledge To summarize, the asymmetric mixing pattern in the balanced bilingual child is the end product of combining all these knowledge described above. In another words, mixing is unlikely to be caused solely by general language proficiency. Based on the predictions outlined in the beginning of this chapter, the present findings can contribute to a clearer picture of mixing behavior in bilingual children. As stated in section 4.1, our balanced bilingual shows evidence of bi-directional mixing of grammatical morphemes. Secondly, mixing predominantly in one language does not necessarily result from dominance in that language. This case study has suggested plausible reasons for mixing other than language dominance.

86

Chapter 5: Conclusions

5.1 Balanced Bilingualism This study has investigated the language development of a Cantonese-English bilingual child, Kathryn (age 3;02.19 4;06.07). In order to measure the language proficiency of the child in each language, the present study has made use of a batch of quantitative measurements: MLU, UB and D which reveal her syntactic and lexical development respectively. In section 3.1.1, I pointed out that these measures are sensitive to morphological features, for example, Cantonese bears

inflationary/deflationary morphological features in calculating MLU and UB when it is compared with inflectional languages like English. Therefore, direct comparison of absolute values of the measurements between two morphologically different languages (in this case, Cantonese and English) would likely result in an inconclusive picture. In addition, one should be cautious of applying the measure D to bilingual children acquisition involving two morphological types of languages. Since English is inflectional and Cantonese is isolating, Cantonese is liable to generate a smaller value of D than English. Also, the contexts in which studies are conducted are crucial because a wider range of situational contexts allows more topics to come up within conversations which in turns increase the lexical diversity, whereas a restricted context would limit verbal production thus reduce the lexical diversity. In order to make these measurements applicable to the present case study, the values obtained from the bilingual subject are compared against those of monolingual children in each language, based on the assumption that the developmental pattern of monolingual children represents normality in language

87

acquisition. It is discovered that the developmental trends of the given quantitative indicators resemble those of monolingual children in both languages (in section 3.2). Therefore I argue that this is one essential checkpoint for identifying balanced bilinguals. Apart from quantitative measurements, in section 3.4.2 it has been shown that Kathryn exhibited transfer in the domains of null objects and wh-interrogatives to a lesser extent compared with English-dominant and Cantonese-dominant bilingual children. Although there are only scant cases of the non-target structures produced by Kathryn, it nevertheless follows the argument that cross-linguistic influence in the form of transfer exists when the two languages are in contact in bilinguals (Mller, 1998). In addition, the present study enriches the argument made in Yip & Matthews (to appear) that frequency of null objects produced by bilingual children is correlated with language dominance because the frequency of null objects is shown to increase along the scale of Cantonese dominance: Balanced bilingual less Cantonese-dominant more Cantonese-dominant.

Increase in the frequency of null objects

5.2 Mixing in Balanced Bilinguals As with other bilingual children, mixing of languages is one of the salient features of Kathryns language performance. The asymmetric mixing pattern produced by Kathryn disproved the predictions that occurrence of mixing is determined by language dominance. With mixing predominant existing within Cantonese utterances, it seems to favor the assumption that Kathryn is English-dominant. However, it would contradict the finding that Kathryn is a balanced bilingual child, on the basis of the quantitative indicators of language proficiency. Moreover, mixing within Cantonese utterances appears to be

88

qualitatively similar to those produced by Cantonese-dominant bilinguals. Thus, I believe that mixing by bilingual children is not essentially explicable by language dominance (section 4.1). An explanation of Kathryns mixing behavior can be derived from the language socialization approach. For the case of frequently mixing English elements into Cantonese utterances, it is well-documented that mixing is ubiquitous within the bilingual community in Hong Kong, and children are inextricably socialized into adults use of mixing. This explanation is supported by the similarities in the types of mixing produced by Kathryn and Hong Kong adult bilinguals. On the other side of the coin, mixing of Cantonese elements into English utterances is rare within the speech community in Hong Kong and consequently children mix less in such way. In section 4.2.1, I have demonstrated that mixing within English utterances can be accounted for by pragmatic reasons like using conventionalized Chinese kinship terms, solving problematic spots arise in bilingual dialogue or overcoming lexical gaps.

5.3 Further Research Due to the limitations of the corpus, analyses involving developmental features cannot be demonstrated in Kathryn as she appears to be too old (linguistically mature) in the earliest transcript for discernable developmental changes to be observed. Therefore the case study leaves open to future research the study of balanced cases concerning a younger age, preferably from the onset of two-word stage where grammar is believed to start to emerge. Moreover, it has been seen that cross-linguistic influence arises in the balanced bilingual subject. In the hope of revealing a clearer picture of how the two languages interact with each other during bilingual language development, the present study

89

prompts further research on a systemic examination of cross-linguistic influences such as transfer in null objects and wh-interrogatives, that may be taking a place in the bilingual child. One could start with including more types of verbs in analyzing target-deviant null objects in English, as in the present study only five transitive verbs were considered. Moreover, in order to test the hypothesis that dominance determines the directionality of transfer, research could draw on data from English-dominant bilingual bilinguals and balanced bilinguals to see if the opposite direction of transfer is attested in the first case and bi-directional transfer in the latter case. Besides these two domains of grammar, Yip & Matthews (2000) also found that relative clause is another vulnerable area of grammar where transfer from Cantonese to English is observed in their Cantonese-dominant bilingual subjects. The English relative clauses were found in prenominal rather than postnominal position, which matches the Cantonese relative clause pattern. Hence it is interesting to investigate whether or to what extent such cross-linguistic influence affects her language development. Although Kathryn had acquired adult-like level of grammar, relative clauses are rare in her corpus. For example: 1. CHI: babe is a pig INV: is a pig ? CHI: <the> [//] a pig that can talk. INV: no, something like a hand which can catch people CHI: it's um, it's a thing that can catch bad guy. INV: ooh, it looks like a weapon CHI: it is a present that I eat, lai5mat6. [present]

(4;02.16)

2.

(4;02.16) (3;07.13)

3.

According to the preliminary result, no case of transfer-based relative clause structure is found in Kathryns corpus. Nor is there any sign of resumptive pronouns

90

as observed in Matthews & Yip (2003)s data from Cantonese-dominant children. For example, in 3, she produces a well-formed object relative clause and she gives further description to the object with the mixed Cantonese word, /lai5mat6/, present for good measure.

91

References

Ansaldo, U. & Matthews, S. (1999). The minnan Substrate and Creolization in Baba Malay. Journal of Chinese Linguistics vol 27, no.1 pp. 39-68 Bernardini, P. & Schlyter, S. (2004). Growing Syntactic Structure and Code-mixing in the Weaker Language. Bilingualism: Language and Cognition. 7.1:49-69. Bhatia, T. K. & Ritchie, W. C. (1999). Language Mixing and Second Language Acquisition: Some Issues and Perspectives. In Klein, Elaine C. & Gita Martohardjono (eds.), The Development of Second Language Grammars. Amsterdam: John Benjamins Pub. Co. Brown, R. (1973). A First Language: The Early Stages. Cambridge, Mass.: Harvard University Press. Chan, H-S. Brian. (1993). In Search Of The Constraints And Processes Of Code-Mixing In Hong Kong Cantonese-English Bilingualism. City Polytechnic of Hong Kong. Chan, H-S, Brian (2003). Aspects Of The Syntax, The Pragmatics, And The Production Of Code-switching: Cantonese And English. Peter Lang, New York Comeau, L., Genesee, F. & Lapaquette, L. (2003). The Modeling Hypothesis and Child Bilingual Codemixing. International Journal of Bilingualism. Vol 7, 2, 113-126 De Houwer, A. (1990). The Acquisition of Two Languages From Birth: a Case Study. Cambridge: Cambridge University Press. De Houwer, A (1994). The Separate Development Hypothesis Method And Implications. In Extra & Verhoeven (eds). The Cross-linguistic Study Of Bilingual Development. Amsterdam: North-Holland. Deuchar, M. & Muntz, R. (2003). Factors Accounting for Code-mixing in an Early Developing Bilingual. In (In)vulnerable Domains in Multilingualism by N. Mller (ed). Amsterdam, Philadelphia: J. Benjamins Pub. Co. Deuchar, M. & Quay, S. (2000). Bilingual Acquisition: Theoretical Implications of a Case Study. Oxford University Press Dpke, S. (1992). One Parent One Language: An Interactional Approach. Amsterdam, Philadelphia: J. Benjamins Pub. Co. Fletcher, P., Leung, C.-S., Stokes, S. F. & Weizman, Z. (2000). Cantonese Pre-School Language Development: a Guide. Hong Kong: Department of Speech and Hearing Sciences, University of Hong Kong Gawlitzek-Maiwald, I. (2003). Approaches to Bilingual Acquisition Data. In (In)vulnerable Domains in Multilingualism by N. Mller (ed.). Amsterdam,
92

Philadelphia: J. Benjamins Pub. Co. Gawlitzek-Maiwald, I. & Tracy, R. (1996). Bilingual Bootstrapping. Linguistics 34: 901-926 Genesee, F., Nicoladis, E. & Paradis, J. (1995). Language Differentiation In Early Bilingual Development. Journal of Child Language 22, 611-631. Goldberg, A. E. (2001). Patient Arguments of Causative Verbs Can Be Omitted: The Role of Information Structure in Argument distribution. Language Sciences 23: 503-524 Ho, S-K. Samantha. (2001). Syntactic Transfer In Cantonese-English Bilingual Development: A Corpus-based Study Of Phrasal Verbs. Undergraduate. Thesis. The University of Hong Kong Huang, P-Y. (1999). The Development of Null Arguments in a Cantonese-English Bilingual Child. Unpublished M. Phil. Thesis, Chinese University of Hong Kong Hulk, A. & Mller, N. (2000). Bilingual First Language Acquisition At The Interface Between Syntax And Pragmatics. Bilingualism: Language and Cognition 3 (3) 227-244. Jisa, H. (2000). Language Mixing In the Weak Language: Evidence From Two Children. Journal of Pragmatics 32, 1363-1386. Klee, T., Stokes, S., Wong, A. M. Y., Fletcher, P. & Gavin, W. J. (2004). Utterance Length and Lexical Diversity in Cantonese-Speaking Children with and without SLI. Journal of Speech, Language, and Hearing Research. Kppe, R. (1996). Language Differentiation In Bilingual Children: The Development Of Grammatical And Pragmatic Competence. Linguistics 34, 927-954. Kppe, R. & Meisel, J. M. (1995). Code-switching In Bilingual First Language Acquisition. In Milroy, L. & Muysken, P. (eds). One Speaker, Two Langugaes: Cross-disciplinary Perspectives On Code-switching. Cambridge. Kwok, W. H. R. (2004). Cantonese-English Code Mixing in Early Bilingual Development. Undergraduate Thesis. The University of Hong Kong Lai, R. (in prep.) Language Mixing in an English-Cantonese Bilingual Child with Uneven Development. M.Phil. thesis. The University of Hong Kong Lakshmanan, U. (1995). Child Second Language Acquisition of Syntax. Studies in Second Language Acquisition 17.301-329. Lanza, E. (1997). Language Mixing In Infant Bilingualism: a Sociolinguistic Perspective. Oxford: Clarendon Press Lanza, E. (2001). Bilingual First Language Acquisition: A discourse Perspective on Language Contact in Parent-child Interaction. In J. Cenoz & F. Genesee (eds). Trends in Language Acquisition Research. Amsterdam: John Benjamins Pub. Co.
93

Lee, Thomas H.T. & Colleen Wong (1998) "CANCORP: the Hong Kong Cantonese Child Language Corpus", Cahiers de Linguistique Asie Orientale vo. 27, no. 2, pp. 211-228. Leung, T. C. (2001) An Optimality-Theoretic Approach To Cantonese/English Code Switching. M. Phil. Thesis. The University of Hong Kong Li, D. C-S (1998). The Plight of the Purist. In M.C. Pennington (ed.), Language in Hong Kong at Centurys End. Hong Kong: Hong Kong University Press. 161-190 Lindholm, K. J. & Padilla, A. M. (1978). Language Mixing in Bilingual Children. Journal of Child Language, 5, 327-35 Luke, K. K. (1998). Why Two Languages Might Be Better Than One: Motivations of Language Mixing in Hong Kong. In M. C. Pennington (ed), Language in Hong Kong at Centurys End. Hong Kong University Press Luke, K. K. & Richards, J. C. (1982). English in Hong Kong: Functions and Status. English Worldwide, 3, 47-64. MacWhinney, B. (2000a). The CHILDES project: Tools for analyzing talk. Third Edition. Mahwah, NJ: Lawrence Erlbaum Associates Malvern, D. & Richards, B. (2002). Investigating Accomodation in Language Proficiency Interviews Using a New Measure of Lexical Diversity. Language Testing, 19, 85-104. Malvern, D, D., Richards, B. J., Chipere, N. & Durn, P. (2004). Lexical Diversity and Language Development: Quantification and Assessment. Basingstoke: Palgrave. Matthews, S. & Yip, V. (1994). Cantonese: a Comprehensive Grammar. London: Routledge. Matthews, S & Yip, V. (2003). Relative clauses in early bilingual development: transfer and universals. In A. giacalone (ed.), Typology and Second Language Acquisition, pp. 39-81. Berlin: Mouton de Gruyter. McLaughlin, B. (1978). Second-language Acquisition in Childhood. Hillsdale, NJ: L. Erlbaum Associates. Miller, J. F. (1981). Assessing Language Production in Children: Experimental Procedures. Baltimore: University Park Press. Meisel, J. (1989). Early Differentiation of Languages in Bilingual Children. In K. Hyltenstam & L. Obler (eds.), Bilingualism Across the Lifespan: Aspects of Acquisition, Maturity, and Loss, pp. 13-40. Cambridge: Cambridge University Press. Meisel, J. M. (1994). (ed.).Bilingual First Language Acquisition: French and German Grammatical Development. Amsterdam/Philadelphia: John Benjamins.
94

Mller, N. (1998). Transfer in Bilingual First Language Acquisition. Bilingualism: Language and Cognition 1 (3) 151-171 Mller, N. & Hulk, A. (2001). Crosslinguistic Influence In Bilingual Language Acquisition: Italian And French As Recipient Languages. Bilingualism: Language and Cognition 4 (1) 1-21 Myers-Scotton, C. (1993). Duelling Languages: Grammatical Structure in Codeswitching. Oxford: Clarendon. Nicoladis, E. & Genesee, F. (1997). The Role of Parental Input and Language Dominance in Bilingual Childrens Code-mixing. In Proceedings of the 21st Annual Boston University Conference on Language Development, Vol. 2, 439-464. (eds) by E. Hughes, M. Hughes & A. Greenhill Nicoladis, E,.& Genesee, F. (1998). Parental Discourse and Codemixing. The International Journal of Bilingualism 2 91): 85-99 Ochs, E. & Schieffelin, B. (1995). The Impact of Language Socialization on Grammatical Development. In Fletcher, P. & MacWhinney, B. (eds). Handbook of Child Language. Oxford, UK: Blackwell. Owen, A. J. & Leonard, L. B. (2002). Lexical Diversity in the Spontaneous Speech of Children with Specific Language Development: Application of D. Journal of Speech, Language, and Hearing Research, 45, 927-937. Padilla, A. & Lindholm, K. (1984). Child Bilingualism: The Same Old Issues Revisited. In J. L. Martinez & R. Mendoza (eds.), Chicano Psychology, 2nd. New York: Academic Press Paradis, J. & Genesee, F. (1996). Syntactic Acquisition in Bilingual Children: Autonomous or Interdependent?. Studies in Second Language Acquisition, 18, 1-25. Cambridge University Press Paradis, J., Nicoladis, E. & Genesee, F. (2000). Early Emergence of Structural Constraints on Code-mixing: Evidence from French-English Bilingual Children. Bilingualism: Language and Cognition, 3 (3), 2000, 245-261 Peng, L.L. (1998). The Development of Wh-questions in a Cantonese/English Bilingual Child. Unpublished M. Phil. Thesis, Chinese University of Hong Kong Petersen, J. (1988). Word-internal Codeswitching Constraints in a Bilingual Childs Grammar. Linguistics 26: 479-493 Redlinger, W. & Park, T. (1980). Language Mixing in Young Bilinguals. Journal of Child Language 7: 337-352. Richards, B. J., & Malvern, D. D, (1997). Quantifying lexical diversity in the study of language development. Reading: University of Reading, The New Bulmershe Papers.
95

Richards, B. J., & Malvern, D. D. (1997b). Typetoken and Type-Type measures of vocabulary diversity and lexical style: an annotated bibliography. Reading: Faculty of Education and Community Studies, The University of Reading. Ronjat, J. (1913). Le Dveloppement du langage observe chez un enfant bilingue. Paris: Champion. Sinka, I. (2000). The Search for Cross-Linguistic Influences in the Language of Young Latvian-English Bilinguals. In Cross-Linguistic Structures in Simultaneous Bilingualism. (ed) by S. Dpke, Amsterdam, Philadelphia: J. Benjamins Pub. Co. Tse, A. (1992) Some Observations On Code-switching Between Cantonese And English In Hong Kong. Working Papers in Languages and Linguistics 4: 101-108. Department of Chinese, Translation and Linguistics, City Polytechnic of Hong Kong. Yip, V. & Matthews, S. J. (2000). Syntactic Transfer In a Cantonese-English Bilingual Child. Bilingualism: Language and Cognition 3 (3): 193-208. Yip, V. & Matthews, S. J. (2005). Dual Input and Learnability: Null Objects in Cantonese-English Bilingual Children. Proceedings of the 4th International Symposium on Bilingualism, 2421-31 Yip, V. & Matthews, S. J. (to appear). Assessing Language Dominance in Bilingual Acquisition. Language Assessment Quarterly. Yip, V. & Matthews, S. J. (in prep.) Bilingual Child: Early Development and Language Contact

96