Assessing child bilingualism: Direct assessment of bilingual syntax amends caretaker report

Abstract

Aims and objectives:

In this paper we investigate measures for the evaluation of a child’s bilingualism.

Methodology:

We compare a caretaker report on the estimated quantity and quality of a child’s bilingualism to results from direct assessment of the bilingual child’s production of complex sentences in Korean and English. We adopt an integrated methodology in which two case studies, closely matched on caretaker report and general background, are investigated directly through an elicited imitation task, for their knowledge of syntactic factors underlying development of coordinate sentence structure in both languages.

Data and analysis:

The participants are two four-year-old Korean-English bilingual children. We compare the caretaker report on the estimated quantity and quality of the children’s bilingualism to results from the direct assessment of their production of comparable complex sentences in both languages. We adopt an integrated methodology in which the two case studies, closely matched on caretaker report and general background, are investigated directly through an elicited imitation task for their knowledge of syntactic factors underlying development of coordinate sentence structure in both languages.

Findings:

Direct assessment reveals significant differences between the two children in the quantity and quality of their bilingualism, in spite of commonalities in caretaker reports.

Limitations:

This study compared only two children and two languages, and focused on language production. Replication with a larger number of subjects, including variation in child age, and measurement through other tasks, for example, tests of language comprehension, are merited.

Implications:

Results are interpreted as motivating both the refinement of caretaker report questionnaires and the necessity for direct assessment of bilingual participants, and suggest the elicited imitation task as a valuable method for conducting such direct assessment.

Keywords

Child bilingualism coordination caretaker report elicited imitation assessment Korean English

Introduction

Assessing the nature and degree of a child’s bilingualism is extremely challenging. Not only are we lacking a shared definition of the term bilingualism (e.g., Grosjean, 2010; Mackey, 2012), but also the individual variance in the numerous factors involved in any single child’s acquisition of more than one language makes comparisons across children almost intractable. Characterizing balanced versus dominant forms of bilingualism is notoriously vexed (e.g., Flege, MacKay, & Piske, 2002; Genesee, 1989; Genesee, Nicoladis, & Paradis, 1995). At the same time, interpreting the proliferating research on childhood bilingualism (e.g., on its cognitive effects or relation to literacy or on its principled comparison to monolingual language acquisition) rests to a large degree on ascertaining the nature of a child’s bilingualism.

Not surprisingly, the development of relevant bilingual proficiency measures has proved extremely challenging (e.g., Esquinca, Yaden, & Rueda 2005; Gathercole, 2010; Paradis, Emmerzael, & Duncan, 2010). Many assessments of child bilingualism rest on caretaker reports (usually parent or teacher) in answer to questionnaires (several of which are under development now, e.g., Gutierrez-Clellen & Kreiter, 2003; Squires, Bricker, & Potter, 1997; Thordardottir & Weismer, 1996). Standardized tests of early language proficiency are often limited in aspects of language knowledge assessed (e.g., vocabulary), normed on monolinguals, and generally not widely available in cross-linguistic forms.¹ These factors tend to render the comparison of proficiency across a bilingual or multilingual child’s languages intractable.

In this paper, we pursue the challenge of characterizing a child’s bilingualism (or multilingualism) in terms of language competence (i.e., grammar as opposed to overall proficiency). We provide evidence that it is necessary to assess a child’s bilingual language knowledge directly as a complement to caretaker reports, and to do so beyond assessment of vocabulary alone. We exemplify a method for doing so through the bilingual administration of an elicited imitation (EI) task in conjunction with an experimental design to evaluate the development of language syntax in early complex sentence formation. We apply this task directly with bilingual children in addition to a Multilingualism Questionnaire, developed by the Cornell Language Acquisition Lab and the Virtual Center for Language Acquisition (CLAL & VCLA, in prep),² which assesses childhood bilingualism or multilingualism through a caretaker report. Our results suggest that direct assessment of a child’s language knowledge through the EI task matched across bilingual languages may be an effective approach to assess the nature of a child’s bilingualism and that using caretakers’ reports exclusively to assess a child’s bilingual proficiency may be insufficient to characterize the nature of a child’s bilingualism. Caretaker reports can be predictive of child bilingual proficiency to a degree, but must be refined in order to do so. For simplicity, we limit to two languages here, although our proposal extends to multilingualism as well (e.g., Bhatia & Ritchie, 2012; Cabrelli Amaro, Flynn, & Rothman, 2012; Flynn, 2012).

Rationale

In order to assess directly a child’s language knowledge in two languages, it is necessary to adopt a task for language elicitation that is applicable across languages and to choose for study a significant component of language knowledge that can be assessed specifically through comparable cross-linguistic experimental design. By comparing the bilingual’s performance in each language to that of monolinguals one can determine the bilingual child’s relative competence in each language. If so, one can investigate whether the nature and degree of a child’s bilingual competence is quantitatively and qualitatively similar in both languages, thus potentially allowing assessment of balanced or unbalanced (dominant) bilingualism as well as estimation of a child’s competence in each language, relative to monolingual language development. To be able to investigate potential correlations between the caretaker report and child bilingualism as measured directly by child performance, systematic caretaker information must be culled and compared to direct assessment of the child.

In this project, children’s bilingualism was assessed both through caretaker report via a Multilingualism Questionnaire and through direct assessment. The questionnaire (the Cornell Language Acquisition Lab/Virtual Center for Language Acquisition Virtual Linguistic Lab Child Multilingualism Questionnaire, in prep.) requested information from the caretaker in five parts: (I) information about the child; (II) information about the child’s language, including child language use and exposure to language, as well as caretaker assessments of children’s proficiency across both language production and comprehension; (III) family background; (IV) the child’s and family’s code switching; and (V) the child’s reading and writing ability if applicable.

To test child bilingual knowledge directly in the bilingual child, our study adopted an EI task for use with bilingual children. This EI method has now been well studied and shown to reveal properties of a child’s language competence, when systematized in terms of administration, analyses and scoring, and when combined with experimental design (Lust, Flynn, & Foley, 1996). It has been extensively used in the study of monolingual first-language acquisition with a variety of experimental designs targeting different linguistic areas across a variety of languages, testing children’s competence with a variety of linguistic structures. In addition, it has been previously used to study second- and third-language acquisition in adults as well as in children (e.g., Berkes & Flynn, 2012; Cabrelli Amaro et al., 2012; Flynn, 1987). A foundation thus exists for the interpretation of bilingual children’s data through the use of this EI task addressed to targeted language knowledge.

The syntax of coordination

In previous research with monolingual children acquiring English, experimental designs were created to test the acquisition of a most fundamental form of complex sentence formation, coordination, where varied forms of coordination were tested, as in (1) for example (Lust, 1977). Results in English monolinguals revealed a structured path of development of this form of complex sentence formation. Later, matched cross-linguistic designs were created to test monolingual children acquiring Japanese and Chinese that allowed comparison of development across languages (Lust & Chien, 1984; Lust & Wakayama, 1979). Results of this cross-linguistic research revealed commonalities in the developmental course of the acquisition of coordination, as well as parametric variation in this structured path of development.

(1) a. i. Babies laugh and babies cry. [Sentential Coordination]

ii. Babies laugh and daddies laugh.

b. i. Babies laugh and cry. [Phrasal Coordination]

ii. Babies and daddies laugh.

Monolingual developmental pattern

Previous research on monolingual English coordination acquisition revealed a pattern of development wherein sentential coordination like (1a) is developmentally fundamental; that is, it appears before or at the same time as phrasal coordination like (1b). Thus, sentential coordination has been argued to be a developmental precursor to phrasal coordination in the course of acquisition of coordination (Lust, 1977; Lust, Foley, & Dye, 2009). It appears that developing grammars require the full sentential coordinate form (1a) to be accessible in order to allow development of the structure and interpretation of the phrasal forms of coordination (1b), which involve reduction of redundant elements. In addition, in English, those phrasal coordination structures with forward directions of reduction (e.g., Babies laugh and babies cry > Babies laugh and Ø cry) where the redundancy reduction site follows the antecedent, for example (1bi), have been found to be developmentally primary to those with backward forms of reduction (e.g., Babies laugh and daddies laugh > Babies Ø and daddies laugh) where the reduction site precedes the antecedent, for example (1bii). This has been argued to be in keeping with the right-branching grammar of English. Children in English frequently reduce redundancy in a forward direction, but rarely in a backward direction. We will not pursue details of linguistic representation of coordinate structures in this paper (see Lust, Flynn, Chien, & Krawiec, 2009, for some discussion). The representation of phrasal forms of coordination is debated; in some representations these have been treated as ‘reduced’ from sentential coordination, where there is full specification of redundant elements. If phrasal coordination is represented directly in the grammar without ‘reduction’ then a semantic model must create the full representation of its interpretation. In either case, phrasal coordination can be viewed as more complex than sentential coordination where complete information is overt. We will use the term ‘reduction’ descriptively in this paper to reflect the fact that redundant elements are not overtly represented in the phrasal forms.

Although cross-linguistic studies of coordination acquisition using EI showed a similar developmental pattern for children acquiring left-branching languages in that sentential coordination appeared as developmentally primary to phrasal forms of coordination, differences were found in that backward forms of phrasal coordination were frequently accessible in left-branching languages, such as Japanese and Chinese (see Lust & Chien, 1984; Lust & Wakayama, 1979), in contrast to English. Thus the EI task results used in the study of the acquisition of complex sentence formation (coordination) cross-linguistically allowed the discovery of developmental patterns showing both a structured path of development and parametric variation across languages. Studies of natural speech in English and Japanese (Lust, 1981; Lust & Mervis, 1980; Lust, Wakayama, Snyder, & Bergmann, 1980, respectively) have provided converging evidence to experimental results based on the EI task. Various forms of more complex coordinate structures have been studied as well (e.g., Foley, Núñez del Prado, Barbier, & Lust, 2003; Lust, Flynn, et al., 2009).

We can now use these results from previous applications of the EI task and from previous cross-linguistic study of the acquisition of coordination as a foundation for interpreting data from bilingual children.

In this paper we report first results of the controlled use of this EI task and experimental designs to assess emerging bilingualism in young children. We compare results on direct assessment of children’s language knowledge through this task with those from caretaker reports of these children’s knowledge through the Multilingualism Questionnaire.

Methodology

We use a modified case study approach, in which we combine testing through experimental design with case studies. Experimental design is necessary in order to evaluate systematically the role of hypothesized factors in language acquisition and in order to compare to systematic studies of monolingual language acquisition. Case studies are necessary because of the individual variance that characterizes any two bilingual children. For integration of case study methods with quantitative assessments, see Scholz and Tietje (2002).

Subjects

We report the results of two case studies from matched children becoming bilingual in the preschool period. The fact that the two cases were comparable was established by analyzing their caretakers’ answers to the Multilingualism Questionnaire. Two young Korean-born children (Korean L1, English L2), similar in age (4 years) and gender (male), were evaluated.

One child, MJ2002 (MJ), was evaluated in Ithaca, NY, by members of the Cornell Language Acquisition Lab. A second child matched in general properties of bilingualism, CH2005 (CH), was evaluated in Cambridge, MA, by the MIT team. Both children were tested on production of simple coordination in both L1 (Korean) and L2 (English) in matched experimental designs testing similar calibrated coordinate structures in two languages. CH’s testing was administered in a preschool at Cambridge, MA. MJ’s was administered in the Early Childhood Center (ECC) at Cornell University. For MJ, a Korean native speaker administered the Korean EI task and a multilingual (Polish-Japanese-German-English) speaker administered the English EI; for CH, a bilingual native speaker of Korean and English administered Korean sentences and a native speaker of English administered the English sentences.

Experimental design

Researchers at Cornell University and MIT collaborated to study the two matched bilingual case studies applying matched methods and experimental designs. Complete metadata on experimental designs, standardized methods and data from this study are available through the VCLA infrastructure, including its cybertools, such as a Data Transcription and Analysis Tool (Blume, Flynn, & Lust, 2012; Blume & Lust, 2012). Here an experimental project description is stored in an experiment bank including administration procedures and scoring criteria, as are audio files, transcriptions and scoring results; see also a short project description on the VCLA website (www.clal.cornell.edu/vcla).

Both children were studied through caretaker report and through direct assessment of the children’s production of coordinate sentences through an EI task with an experimental design looking at simple coordination structures in each language, the native language (Korean) and the second language (English). The experimental design of coordinate sentences in both languages replicated earlier studies with monolinguals and thus our use of these designs allowed comparison to previous monolingual results.³ For this study, stimulus sentences were designed in Korean according to the same factors that were varied in the English study (and prior cross-linguistic studies in Japanese and Chinese), with similar controls on structure and lexicon.

The factors varied were Coordinate Structure (Sentential or Phrasal Coordination), Directionality (Forward or Backward redundancy reduction) and Sentence Structure (SV or VO/OV). This was a 2 × 2 × 2 factorial design with eight conditions and two replication sentences in each condition, producing 16 sentences for imitation. Replication items matched in structure but varied in lexicon.⁴ A total of 20 sentences were presented for imitation in each language. Although experimental designs created sentences matched in both English and Korean, English sentences were in the unmarked VO order and Korean sentences were in the unmarked Korean OV order. All sentences were controlled in length (5–6 words, 7 syllables in Korean, 10 syllables in English). They all involved a simple, highly frequent, child familiar lexicon. Two randomized sentence batteries were created for each set of replication items. English and Korean EI tasks were administered in two separate sessions. Tables 4 and 5 in the appendix present the stimuli used in both languages.

For both children, an abbreviated Multilingualism Questionnaire was completed by their mother. Answers were used to begin to assess the nature and degree of each child’s bilingualism according to the caretaker’s perspective, and to establish comparability across the children. Results are now being entered in the master database including the Data Transcription and Analysis Tool at Cornell University.

Subject background

Multilingualism Questionnaire results

As parent answers to the Multilingualism Questionnaire verified, for both children Korean was continuously the main language of the home.⁵ Both were characterized as sequential bilinguals. Both were attending English-speaking nursery schools in the US at the time of testing. MJ entered day care at about 8 months, and nursery school in the US at age 2 years 4 months. CH entered day care in the US at one year of age. Both were continuously enrolled since school entry until the date of testing. Neither had siblings. MJ’s parents both had MA degrees; the mother was a graduate student at Cornell, the father an accountant. CH’s mother had an MA degree; the father had a PhD and was working as a postdoc at MIT. CH’s parents rated their own English proficiency as 3 on a four-point scale and so did MJ’s. The proficiency scale used is presented in Table 1.

Table 1.

English proficiency scale.

1	Cannot speak but understands the general idea of what is being said
2	Limited proficiency and limited vocabulary
3	Good proficiency
4	Native-like proficiency

Both MJ and CH were evaluated by their mothers as using Korean 80% of the time overall, with English only 20%, and were characterized as dominant L1. Both were described as comprehending and producing Korean more proficiently than English and feeling more comfortable in the use of Korean in all contexts. On a general overall proficiency scale of 1–4 (cf. Table 1), both children were rated by their mothers as ranking 4 on their L1 (Korean) and 2 on their L2 (English). Both sets of parents described themselves as encouraging English vocabulary learning in their child by reading English books to them. Both children played with Korean friends, in Korean, about once a week. Outside of school, both children were exposed to and spoke predominantly Korean, according to caretaker reports.

The only reported differences on the Multilingualism Questionnaire between the two children were the following. (1) While both children were reported to use English predominantly at school, MJ spent 40 hours per week at preschool where he was exposed to and spoke English only, but CH spent only 9 hours per week at preschool and used English there about 80% of the time and Korean the rest of the time. (2) On proficiency scores ranging from 1 to 6, both listening and production proficiency were rated for CH as 6 in Korean, but 3 in English. MJ was reported also as 6 in Korean on both listening and production but as 5 in English on both. Tables 2 and 3 present the proficiency scales used.

Table 2.

Proficiency ratings for listening/comprehension.

1	Understands only a few single words mainly
2	Understands single words but no more than 10 approximately
3	Understands more than 10 words, but single words mainly
4	Understands some sentence combinations, usually with assistance
5	Good understanding of conversations and other language domains, such as TV
6	Native-like proficiency

Table 3.

Proficiency ratings for speaking/production.

1	Can produce a few single words mainly
2	Can produce single words, no more than 10 words
3	Can produce mainly words, more than 10 words
4	Combines words to create simple sentences, pauses, sometimes not understood
5	Combines words to longer sentences, usually understood.
6	Native-like proficiency

Data collection and analysis

Procedures

The EI task was administered to each child under standardized conditions (cf. procedures established in Lust & Blume, under contract).

MJ was tested on L1 (Korean) coordination through EI at age 4;1;20, and on L2 (English) coordination at age 4;1;12. CH was tested on L1 (Korean) at age 4;1;4 and on L2 (English) coordination at age 4;0;1.

EI data were transcribed by bilingual native speakers of Korean. English transcriptions were reliability checked by an English native speaker as well as by a second bilingual speaker. Korean data were reliability checked by a third native speaker of Korean. Data were collected in the VCLA master database and scored at Cornell by two independent bilingual experimenters using standardized scoring criteria and reliability measures to resolve any discrepancies in transcript or scoring. In general, any major structural change, for example change of coordination type or grammatical case, was scored as incorrect, while inflectional, phonological or morphological changes alone were not.

Results

Comparison of English and Korean production in the elicited imitation task

Figure 1 summarizes comparisons of percent correct on the EI task for the two children across their L1 and L2, across all factors of the experimental designs.

Figure 1.

Correct responses to the elicited imitation task on coordination in both languages.

As Figure 1 shows, in spite of commonalities across the children, and in spite of several commonalities in caretaker reports on the Multilingualism Questionnaire, MJ and CH differed significantly in their coordinate sentence production as measured through the EI task. This was true for both languages, Korean and English. While both children show superior performance in Korean, consistent with caretaker reports of Korean dominance in each child, MJ achieved 81% correct on English coordination, while CH achieved only 38%. MJ’s achievement in L2 English is not due to L1 Korean attrition, as he achieves 100% imitation success on Korean sentences, while CH achieves only 50%.

Results from the English elicited imitation task

Figures 2 and 3 compare children’s performance on English coordination according to the structural factors of the design.

Figure 2.

Correct answers to elicited imitation task on English sentence coordination by direction.

Figure 3.

Correct answers to elicited imitation task on English phrasal coordination by direction.

In general, MJ’s control of phrasal coordination shown in Figure 3 is consistent with his attested control of sentential coordination (75%) shown in Figure 2. As Figure 3 suggests, on phrasal coordination in English, MJ shows superior performance on forward forms, over backward, as do English monolingual children.

CH, demonstrating a less advanced developmental level in English, shows performance on phrasal coordination to be generally lagging behind sentential coordination (25% versus 50%), for example see (2) (a backward phrasal coordination). Example (2) shows CH’s imitation of a backward reduction in a phrasal coordination structure. In this and all the following examples, E represents Experimenter.

(2) E: The dolls and the babies jump. [[S&S] V]

CH: Doll is [ə] baby jump. (Battery B, #8)

(The sound [ə] between is and baby was coded as an indefinite determiner.)

Both children made inflectional errors in English, examples of which are presented in (3). Imitations that contained only inflectional errors were not counted as incorrect.

(3) a. E: The teddy bear walks and sleeps.

CH: Teddy bear walk [n] sleeps. (Battery 2, #10)

(The sound [n] was coded as an attempt to produce and.)

b. E: Mommies jump and babies jump.

CH: Mommy jump and baby jump. (Battery 1, #7)

c. E: There are bears and there are dogs.

MJ: There’[s] bears and dogs. (Battery B, #9)

(The sound [s] was coded as an elision of is.)

d. E: The kitties and the dogs hide.

MJ: The kitty and the dogs hide. (Battery A, #1)

At the same time as they made inflection errors, analyses of the structural conditions within the coordination experiment revealed common structural patterns of acquisition across both children, coherent with that of monolingual first-language acquisition of English at different periods of acquisition, MJ more advanced and CH still developing.

In the English experiment, MJ is highly successful on both sentential and phrasal coordination structures. Within sentential coordination, he reduces redundancy by producing well-formed forward reduction patterns, as in (4), demonstrating a preference for forward deletion. This spontaneous reduction of redundant elements was scored as ‘error’, since it significantly changed stimulus sentence structure. This resembles the behavior of English monolingual children during the development of coordination (Lust, 1977):

(4) E: Eat ice cream and eat cookies. (VO & VO)

MJ: Eat ice cream> eat ice cream and cookie. (Battery A, #9)

CH does not reduce redundancy in any sentential coordination structures, and produces only 50% of these structures correctly. His errors were sometimes structural as in (5a), although they often appeared to reflect difficulty with the lexicon rather than with the coordinate structure, as in (5b).

(5) a. E: The kitties and the dogs hide.

CH: Kitty is [da] dog. (Battery A, #1)

([da] was coded as an attempt to produce the.)

b. E: Babies laugh and babies cry.

(The experimenter repeats the sentence.)

CH: Baby laugh, baby [kwet]. (Battery A, #6)

([kwet] was coded as an attempt to produce cry.)

Both children, then, are acquiring coordinate L2 syntax (English) in a normal, structured fashion, relative to English monolingual acquisition of coordination, with a clear difference in level of acquisition at the time of testing.

Results from the Korean elicited imitation task

Figures 4 and 5 compare MJ’s and CH’s performance on Korean sentential and phrasal coordination.

Figure 4.

Correct answers to elicited imitation task on Korean sentential coordination by direction.

Figure 5.

Correct answers to elicited imitation task on Korean phrasal coordination by direction.

Figures 4 and 5 show that in imitation of the matched Korean sentences, MJ achieves perfection on all coordination types, while CH again demonstrates a less developed stage of acquisition, achieving only 50% success on both sentential and phrasal coordination. Sentential coordination forms are equal or superior to phrasal coordination forms for CH across three of the coordination types. As examples (6) and (7) show, a large proportion of CH’s errors in Korean involved omission of the case markers, for example the accusative marker ul in example (6). These types of errors were not produced by MJ.⁶

(6) E: 용돈을 받고 용돈을 쓴다.

Yongton-ul pat-ko yongton-ul ssu-n-ta.

Ø allowance-ACC receive-and Ø allowance-ACC spend-IND-DECL.⁷

‘(Someone) Receive(s) (an) allowance and (someone) spend(s) (an) allowance.’

(The experimenter repeats the sentence.)

CH: 용대 받고 용대 쓴다.

Yongtay pat-ko yongtay ssu-n-ta.

Ø allowance (phonological error) receive-and Ø allowance (phonological error) spend-IND-DECL.

‘(Someone) Receive(s) (an) allowance (omission of ACC) and (someone) spend(s) (an) allowance (omission of ACC).’ (Battery B, #4)

(7) E: 바다를 보고 산을 본다.

Pata-lul po-ko san-ul po-n-ta.

Ø sea-ACC look_at-and Ø mountain-ACC look_at-IND-DECL.

‘(Someone) Look(s) at the sea and (someone) look(s) at (a) mountain.’

CH: 바다 보고 산을 본다.

Pata po-ko san-ul po-n-ta.

Ø sea look_at-and Ø mountain-ACC look_at-IND-DECL.

‘(Someone) Look(s) at the sea (omission of ACC) and (someone) look(s) at (a) mountain.’ (Battery A, #5)

In summary, CH, although revealing a less developed stage of acquisition than MJ in both his first language, Korean, and his second language, English, appears to be following structural patterns of development consistent with monolingual language acquisition in both of his languages, that is, sentential coordination is accessible by the time phrasal coordination in both languages, in accord with monolingual development. In Korean, CH appears to be at a period of development with control of both sentential and phrasal coordination and both forward and backward directionality in these, but he is still acquiring the complex case marking system of Korean, resembling normal monolingual Korean developmental patterns, but at a somewhat slower rate (cf. Kim, 1997).

Comparison to previous studies of monolingual acquisition

On simple coordinate sentences like those tested here in English, it was found for 32 young children (between 2.0 and 3.1 years of age) acquiring English monolingually that children’s language level (measured by mean length of utterance (MLU) ) had an effect. Children by three years of age with a high MLU (⩾ = 4.76) provided 81% correct imitation (Lust, 1977). Children from 2.3 to 2.11 with a lower MLU (3.76–4.75) provided 50% correct imitation. Thus in our current study, MJ is performing overall on a par with children acquiring English complex sentence formation by the age of four, while CH is performing at a level more similar to two year olds at earlier stages of monolingual English language development.

Previous studies in two Asian languages can give us comparison norms for the Korean sentences we tested here. Although we have no previous data for Korean, we believe similar studies conducted in Japanese and Chinese could be used for comparison. On a study of simple coordinate sentences with 38 Japanese children acquiring Japanese monolingually, children with a mean age of 3.10 demonstrated 66% correct imitation; children with a mean age of 5.0 demonstrated 89% correct imitation.⁸ Younger children with a mean age of 2.10 demonstrated 41% correct imitation and children with a mean age of 3.2 year demonstrated 47% correct imitation (Lust & Wakayama, 1979). Children acquiring Chinese on a study of similar coordinate sentences with 68 subjects found that children with a mean age of 4.3 demonstrated 83% correct imitation, while children with a mean age of 2.9 demonstrated 64% correct imitation (Lust & Chien, 1984). Thus again, in this study, MJ appears to be demonstrating performance on his native language, Korean, in a manner compatible with his age counterparts, while CH is demonstrating earlier levels of language acquisition.

Summary of results

Although similar in most aspects of caretaker reports and in many general factors of comparison on these reports, these case studies show on the basis of EI data that descriptively matched children according to caretaker reports may differ significantly in their knowledge of both their first and their second language. Accordingly, they exhibit quite distinct types of bilingualism. Direct testing of the children may reveal what the caretaker report alone does not.

At the same time, both children show coordination acquisition patterns, which are consistent with monolingual first-language acquisition in both languages. Sentential coordination is available early, which is normal for monolingual first-language acquisition in both languages. Where development is still occurring, both show patterns generally consistent with monolinguals in each language with regard to the directionality factors of their language.

While both children are confirmed to be dominant Korean, MJ is perfect in his Korean production of the syntax of coordination tested here, and lacks in English only forms marked in directionality for English as well as some specifics of English inflection. CH, in contrast, is still developing both Korean and English at less than half the rate of MJ. Although both children are still developing their bilingualism, CH reflects a degree of ‘early bilingualism’ relative to MJ.

Conclusions

Our results can contribute in general to the study of relations between first- and second-language acquisition (e.g., Meisel, 2011). In particular, they lead us to several conclusions and directions for future research regarding child bilingualism.

In the estimation of child bilingualism, the caretaker report needs to be supplemented by direct child analyses. Even when child characteristics are very similar according to the caretaker report, we find distinct differences in the degree of children’s bilingualism when we test the child directly. This result generally coheres with other research in the area of child bilingualism (e.g., Pease-Álvarez, Hakuta, & Baylery, 1996, who found non-linear relationships between certain background factors and child proficiency), and with research in other areas of language development, (e.g., studies of infant vocabulary comprehension as in Friend, Schmitt & Simpson, 2011). The results from our caretaker reports suggest that many aspects of child background and estimated proficiency reported there do not distinguish children whose bilingualism does in fact differ when the child is tested directly. For example, on the general overall proficiency scale (ranging from 1–4), both children were rated by their mothers as ranking 4 on their L1 and 2 on their L2, although EI testing results clearly distinguished them.

Matched controlled cross-linguistic experimental designs with a task such as EI testing syntactic properties of language can be effective in this direct child testing, since it did reveal significant differences between the two children in their bilingualism even when many aspects of the caretaker report did not.

At the same time, the Multilingualism Questionnaire completed by parents did differentiate the children in certain ways, which cohered with the observed differences between the children on direct testing. For example, on separate listening/comprehension and production proficiency scores ranging from 1 to 6, these were both rated by the caretaker for CH as 6 in Korean, but 3 in English. However, MJ was reported also as 6 in Korean on both listening and production but as 5 in English on both measures. Although this result did not capture the differences in Korean we observed through the EI data in these children, it did hint at the different English performance we observed. In addition, on the Multilingualism Questionnaire, reported amount of exposure to English (L2) did differ on caretaker’s reports. While MJ was reported to have had 40 hours of estimated exposure a week of English in preschool, CH was reported to have had only 9. We might consider then the hypothesis that ‘amount of exposure’ may be a critical distinguishing factor to explain MJ’s superior performance. Thus, it would be a critical questionnaire diagnostic (cf. Hoff et al., 2012; Pearson, Fernández, Lewedeg, & Oller, 1997 ).

However, in general, amount of exposure cannot be the sole explanation for the observed language differences between these children. While it may have been predictive of MJ’s superior English L2 performance, it is not predictive of his parallel superior performance in Korean, since apparently both children have comparable amounts of exposure to Korean L1 input. Although, based on the mothers’ reports, we could predict a difference in the children’s English acquisition due to the different English exposure patterns, the difference in the children’s Korean acquisition was not predicted. This result coheres with those of Pease-Álvarez et al. (1996), who found that the relation between bilingual children’s proficiency and amount of exposure to a language are not necessarily in a direct relation. Gutierrez-Clellen and Kreiter (2013 ) also found differential effects of exposure across two languages of bilingual children, suggesting that general principles should be investigated here.

The child performing less well on L2 English in our experiment, CH, also showed less developed performance in L1 Korean. This parallelism between L1 and L2 observed in our results raises critical questions about the relation between languages known by bilingual children and the potential mutual reinforcement of language development across languages. It raises the issue of to what degree development in L1 can potentially reinforce development of L2, or vice versa. As has been speculated for example, it is possible that ‘bilinguals succeed more easily than monolinguals in decoding the language-specific coding systems and the underlying grammatical principles’ of their languages (Meisel, 1990, p. 18), presumably because of some form of mutual reinforcement across the learning of more than one language. The parallelism we found in development between Korean and English in the bilingual children, MJ and CH, reinforces the need to directly assess both languages in a bilingual child’s language knowledge. It suggests motivation for reconsidering a ‘mutual reinforcement’ hypothesis in childhood bilingualism.

Finally, our results suggest that a multilingualism questionnaire based on a caretaker report, although it may be predictive of child bilingualism to some degree, deserves refinement and development. Both what questions are asked and how they are asked must be assessed. For example, both parents in this study ranked their children’s proficiency lower when answering questions separately on more refined scales (1–6) for production and comprehension proficiency as opposed to questions on ‘overall proficiency’ (1–4); their ratings on the more refined scales to some degree did predict the general pattern of differences we found between our two children, when the general proficiency measure did not.

In general, our combined results motivate not only development of more refined caretaker estimates of children’s multilingualism but also development of systematic assessments of child bilingualism through direct testing of the child’s knowledge of his/her languages. Further research is now needed to develop a caretaker-based multilingualism questionnaire and to conduct comparisons of various forms of directly elicited child linguistic data, with caretaker reports. Our updated version of the CLAL-VCLA Multilingualism Questionnaire now provides additional refinements. Assessments of caretakers’ reports on various aspects of child behavior and context (e.g., quality or quantity of input, etc.) with both L1 and L2 (or beyond) can be evaluated to assess which reported properties most significantly correlate with child language capacity, when it is directly tested.⁹ Comparisons across various questionnaires should be conducted (e.g., Gutierrez-Clellen & Kreiter, 2013 ) in terms of factors predicting proficiency.

Finally, our results are limited to two comparable cases and one form of bilingualism (English-Korean) and thus require replication. Although we have conducted this study on the basis of direct assessment of child language through of a test of language production (i.e., the EI task), it would be possible in the future to conduct a similar study on the basis of a receptive assessment task as well, especially where such receptive tasks provide comparable results from monolingual language acquisition studies. As argued by Lust, Chien, and Flynn (1987), conducting a comprehension task in addition to a production task can provide ‘converging measures in test of hypotheses regarding specific factors of grammatical competence’ (p. 328).

Moreover, although our study has focused on childhood bilingualism, comparable methodology can be developed in the area of adult second-language acquisition, where self-report measures (e.g., Li, Sepansky, & Zhao, 2006) may be evaluated as caretaker reports are for children, with comparison to direct assessments of bilingual language knowledge.

Footnotes

Appendix

Table 4 presents the sentences that were used for elicited imitation in English, while Table 5 presents the sentences that were used for elicited imitation in Korean.

Acknowledgements

We acknowledge and thank Elizabeth Stilwell, director of the Cornell ECC for aiding us in work with children of the Center including MJ. We thank Janet Jeon and Zofia Stankiewicz for sensitive data collection with MJ, Simin Zhang, You Mee Lee, Aeri Lee and Himchan Mona Lee for assistance in data analyses. We thank Nicholas Williamson for work on the CLAL-VCLA Multilingualism Questionnaire.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Science Foundation (grant number 0753415) to María Blume and Barbara Lust for the project ‘Transforming the Primary Research Process Through Cybertool Dissemination: An Implementation of a Virtual Center for the study of Language Acquisition’.

Notes

Author biographies

Barbara Lust is Director of Cornell Language Acquisition Lab and Professor of Developmental Psychology, Linguistics and Cognitive Science at Cornell University. She received her PhD in Developmental Psychology from City University of New York Graduate Center after earlier studies at L’Institut des Sciences de l“Education, at the University of Geneva, Switzerland. She followed this with postdoctoral study in linguistics and philosophy at MIT before coming to Cornell. Her teaching at Cornell involves Developmental Psychology and Linguistics, within an interdisciplinary perspective of Cognitive Science.

Suzanne Flynn is Professor of Foreign Languages and Linguistics Second Language Acquisition at MIT, USA.

María Blume, M.A., PhD, is an Assistant Professor at University of Texas at El Paso, USA. Professor Blume’s research interests are first language acquisition and second language acquisition/bilingualism, with special interest in the acquisition of morphology, syntax and pragmatics.

Carissa Kang is a PhD student at Cornell University. Her research interests includes nilingualism & attentional development (across socioeconomic status and cultures), code-switching and attentional development (executive function) in adults and children.

References

Berkes

Flynn

(2012). Multilingualism: New perspectives on syntactic development. In Bhatia

Ritchie

W. W.

(Eds.), Handbook of Bilingualism and multilingualism (pp. 137-167). Oxford, UK: Blackwell.

Bhatia

Ritchie

(2012). The handbook of bilingualism and multilingualism. Second Edition. Hoboken, NJ: Wiley-Blackwell.

Blume

Flynn

Lust

(2012). Creating linked data for the interdisciplinary international collaborative study of language acquisition and use: Achievements and challenges of a new Virtual Linguistics Lab. In Chiarcos

Nordhoff

Hellmann

(Eds.), Linked data in linguistics. Representing and connecting language data and language metadata. Heidelberg, Germany/Dordrecht, The Netherlands/London, UK/New York, NY: Springer, pp. 85–96.

Blume

Lust

(2012) First steps in transforming the primary research process through a virtual linguistic lab for the study of language acquisition and use: Challenges and accomplishments. Journal of Computational Science Education, 3, 34–46.

Cabrelli Amaro

Flynn

Rothman

. (2012). Third language acquisition in adulthood. Philadelphia, PA: John Benjamins.

Cornell Language Acquisition Lab (CLAL) and Virtual Center for Language Acquisition (VCLA) Virtual Linguistic Lab Child Multilingualism Questionnaire. ( Blume

Lust

with the collaboration of Yang

Sujin

Courtney

Urzúa

Park

Champagne-Williamson

Kang

, & founding members of the VCLA) (in prep.) Cornell University, Ithaca, NY.

Esquinca

Yaden

Rueda

(2005). Current language proficiency tests and their implications for preschool English language learners. In Cohen

McAlister

K. T.

Rolstad

MacSwan

(Eds.), Proceedings of the 4th international symposium on bilingualism (pp. 674–680). Somerville, MA: Cascadilla Press.

Flege

MacKay

Piske

(2002). Assessing bilingual dominance. Applied Psycholinguistics, 23, 567–598.

Flynn

(1987). A parameter-setting model of L2 acquisition: Experimental studies in anaphora. Dordrecht, The Netherlands: Reidel.

10.

Flynn

(2012). Introduction. In Cabrelli Amaro

Flynn

Rothman

(Eds.), Third language acquisition in adults (pp. 1–6). Amsterdam, The Netherlands: John Benjamins.

11.

Foley

Núñez del Prado

Barbier

Lust

(2003). Knowledge of variable binding in VP ellipsis: Language acquisition research and theory converge. Syntax, 6, 52–83.

12.

Friend

Schmitt

Simpson

(2011). Evaluating the predictive validity of the computerized comprehension task: Comprehension predicts production. Developmental Psychology, 40, 1–13.

13.

Gathercole

V. C. M.

(2010). Bilingual children: Language and assessment issues for educators. In Littleton

Wood

Kleine Staarman

(Eds.), International handbook of psychology in education (pp. 713–748). Bingley, UK: Emerald Group.

14.

Genesee

(1989). Early bilingual development: One language or two. Journal of Child Language, 16, 161–179. Reproduced in Wei

(Ed.), The bilingual reader (pp. 327–343). London, UK: Routledge.

15.

Genesee

Nicoladis

Paradis

(1995). Language differentiation in early bilingual development. Journal of Child Language, 22, 611–631.

16.

Grosjean

(2010). Bilingual: Life and reality. Cambridge, MA: Harvard University Press.

17.

Gutierrez-Clellen

Kreiter

(2013). Understanding child bilingual acquisition using parent and teacher reports. Applied Psycholinguistics, 14, 267–288.

18.

Hoff

Core

Place

Rumiche

Señor

Parra

(2012). Dual language exposure and bilingual development, Journal of Child Language, 39, 1–27.

19.

Kang

Kim

Lust

(in prep.). Examining variables that affect bilingual children’s language development: Analysis of caregiver report for English-Malay bilinguals in Singapore. Ms., Cornell University.

20.

Kim

Y.-J.

(1997). The acquisition of Korean. In Slobin

(Ed.), The cross-linguistic study of language acquisition, vol. 4. (pp. 335–443). Hillsdale, NJ: Lawrence Erlbaum.

21.

Sepanski

Zhao

(2006). Language history questionnaires: A web-based interface for bilingual research. Behavior Research Methods, 38, 202–210.

22.

Lust

(1977). Conjunction reduction in child language. Journal of Child Language, 4, 257–297.

23.

Lust

(1981). On coordinating studies of coordination: Problems of method and theory in first language acquisition - A reply to Ardery. Journal of Child Language, 8, 457–470.

24.

Lust

Blume

(under contract). With the collaboration of Chien, Y., Dye, C., Kedar, Y., Woetzel, T., Somashekar, S., Sanchez, L., Foley, C., Lakshmanan, U., Kalashnikoa, M., Rayas, M., Yang, S., & Buitrago, N. and members of the VCLA. Research methods manual: Scientific methods for the study of language acquisition. Washington, DC: American Psychological Association/Walter de Gruyter.

25.

Lust

Chien

Y.-C.

(1984). The structure of coordination in first language acquisition of Chinese. Cognition, 17, 49–83.

26.

Lust

Chien

Y.-C.

Flynn

(1987). What children know: Methods for the study of first language acquisition. In Lust

(Ed.), Studies in the acquisition of anaphora (Volume II). Dordrecht, Holland: D. Reidel Publishing Company.

27.

Lust

Flynn

Chien

Y.-C.

Krawiec

(2009). First language acquisition of coordination: The Mud-Puddle study and beyond. In Lewis

Karimi

Harley

S. Farrar

(Eds.), Time and again. Theoretical perspectives on formal linguistics (pp. 151–176). Oxford, UK: John Benjamins.

28.

Lust

Flynn

Foley

(1996). What children know about what they say: Elicited imitation as a research method. In McDaniel

McKee

(Eds.), Methods for assessing children’s syntax (pp. 55–76). Cambridge, MA: MIT Press.

29.

Lust

Foley

Dye

(2009). The first language acquisition of complex sentences. In Bavin

(Ed.), The Cambridge handbook of child language (pp. 237–257). Cambridge, UK: Cambridge University Press.

30.

Lust

Mervis

C. A.

(1980). Coordination in the natural speech of young children. Journal of Child Language, 7, 279–304.

31.

Lust

Wakayama

T. K.

(1979). The structure of coordination in young children’s acquisition of Japanese. In Eckman

Hastings

(Eds.), Studies in first and second language acquisition (pp. 134–152). Rawley, MA: Newbury House.

32.

Lust

Wakayama

Snyder

Bergmann

(1980). The acquisition of complex sentences in first language acquisition of Japanese: Coordination. Unpublished manuscript, Cornell University. Paper presented at Fifth Annual Boston University Conference on Child Language. Boston, MA.

33.

Mackey

(2012). Bilingualism in North America. In Bhatia

Ritchie

(Eds.), Handbook of bilingualism and multilingualism (pp. 707–724). Oxford, UK: Blackwell.

34.

Meisel

J. M.

(1990). Grammatical development in the simultaneous acquisition of two first languages. In Meisel

J. M.

(Ed.), Two first languages: Early grammatical development in bilingual children (pp. 5–22). Amsterdam, The Netherlands: John Benjamins.

35.

Meisel

J. M.

(2011). First and second language acquisition. Cambridge, UK: Cambridge University Press.

36.

Paradis

Emmerzael

Duncan

(2010). Assessment of English language learners: Using parent report on first language development. Journal of Communication Disorders, 43, 474–497.

37.

Pearson

B. Z.

Fernández

S. C.

Lewedeg

Oller

D. K.

(1997). The relation of input factors to lexical learning by bilingual infants (ages 10 to 30 months). Applied Psycholinguistics, 18, 41–58.

38.

Pease-Álvarez

Hakuta

Baylery

(1996). Spanish proficiency and language use in California’s Mexicano community. Southwest Journal of Linguistics, 15, 137–151.

39.

Scholz

Tietje

O. I.

(2002). Embedded case study methods: Integrating quantitative and qualitative knowledge. Thousand Oaks, CA: Sage Publications.

40.

Squires

Bricker

Potter

(1997). Revision of parent-completed developmental screening tool: Ages and stages questionnaire. Journal of Pediatric Psychology, 22, 313–328.

41.

Thordaardottir

Weismer

(1996). Language assessment via parent report: Development of screening instrument for Icelandic children. First Language, 16, 265–285.