predicting minority children's bilingual proficiency: child, family, and institutional factors

29
Language Learning 41:2, June 1991, pp. 205-233 Predicting Minority C hi1 dre n’s Bilingual Proficiency: Child, Family, and Institutional Factors* Ludo T. Verhoeven Ti1 b u rg University The purpose of the present study was to identify factors that predict the first- and second-language proficiency of ethnic minority children at the age of 6 years. A sample of 72 six-year-old Turkish children, living in The Netherlands since their infant years, was identified prior to their entrance into the first grade of primary school. Predictor measures originatingfiom the child, his or her family, and the institu- tional care the child had gone through were collected, along with direct and indirect first- and second-language profi- ciency measures. The results ofthe study make clear that two dimensions underlie the children’s proficiency in either lan- guage: communicative skills versus cognitivdacademic skills. Measures of the cultural orientation ofthe children and their parents turned out to be related to all of the proficiency levels under consideration. The extent of caretaker interaction in the first language was also positively related to the children’s bilingual proficiency level. Moreover, there was evidence for the notion of interdependency in bilingual development in that cognitive/academic abilities in the second language could be predicted from similar abilities in the first language. *Requests for reprints may be sent to the author at Linguistics Department, Tilburg University, Post Office Box 90153,5000 LE Tilburg, Holland. 205

Upload: ludo-t-verhoeven

Post on 03-Oct-2016

219 views

Category:

Documents


2 download

TRANSCRIPT

Language Learning 41:2, June 1991, pp. 205-233

Predicting Minority C hi1 dre n’s Bilingual Proficiency:

Child, Family, and Institutional Factors*

Ludo T. Verhoeven Ti1 b u rg University

The purpose of the present study was to identify factors that predict the first- and second-language proficiency of ethnic minority children at the age of 6 years. A sample of 72 six-year-old Turkish children, living in The Netherlands since their infant years, was identified prior to their entrance into the first grade of primary school. Predictor measures originatingfiom the child, his or her family, and the institu- tional care the child had gone through were collected, along with direct and indirect first- and second-language profi- ciency measures. The results ofthe study make clear that two dimensions underlie the children’s proficiency in either lan- guage: communicative skills versus cognitivdacademic skills. Measures of the cultural orientation ofthe children and their parents turned out to be related to all of the proficiency levels under consideration. The extent of caretaker interaction in the first language was also positively related to the children’s bilingual proficiency level. Moreover, there was evidence for the notion of interdependency in bilingual development in that cognitive/academic abilities in the second language could be predicted from similar abilities in the first language.

*Requests for reprints may be sent to the author at Linguistics Department, Tilburg University, Post Office Box 90153,5000 LE Tilburg, Holland.

205

206 Language Learning Vol. 41, No. 2

The purpose of this study was to determine which factors predict bilingual proficiency in ethnic minority children by the time they begin formal instruction at school. Research on predictors of the early language proficiency of minority children has tended to focus upon three different domains: (a) the child, (b) the family, and (c) the institutional care.

CHILD CHARACTERISTICS

With respect to child characteristics, several studies fo- cused upon the child's cultural orientation. In general terms, cultural orientation can be defined as the attitudes minority children have developed toward their native language and culture and their attitude toward the majority language and culture. These attitudes determine the degree of identification with the two cultures (see Lambert, 1978).

The prediction of first-language maintenance from learner characteristics in young children has not been studied exten- sively. Research into the role of child characteristics on the process of bilingual development has focused mainly upon the prediction of success in second-language learning. Wong Fillmore (1976, 1979) presented evidence that some sort of integrative motivation influences the language development in minority children. She found that children displaying a great desire to belong to and identify with peers who speak the target language made the best progress in second-language learning. On the other hand, Genesee and Hamayan (19801, in a study of 6-year-old Anglophone Canadian children learning French, found no correlation between attitudinal variables and L2 proficiency. However, because in Canada both English and French are highly valued, the children did not express strong positive or negative feelings toward L2 speakers.

Another possibly relevant child characteristic is cognitive capacity. It can be assumed that in acquiring language profi- ciency one brings to bear a number of highly transferable skills that underlie many other intellectual abilities as well (see

Verhoeven 207

Oller, 1980). However, Genesee and Hamayan (1980) found that the correlation between intelligence measures and L2 proficiency was relatively low in younger learners.

FAMILY CHARACTERISTICS

Research on the relationship between family characteris- tics and minority children’s language proficiency has focused upon how and to what degree the family creates an effective in- home learning environment. A relevant characteristic in this respect is language exposure. Wells (1985) has clearly demon- strated that the rate of language development in children is associated with specific characteristics of adult speech. He found that the manner and extent to which adults adjust their speech to the immaturity of their conversational partners affects the ease with which children master the language system. In functional interactional terms he proposed four broad types of intention to be extremely relevant: (a) mainte- nance of intersubjectivity of attention, (b) expression of understandable propositions, (c) ensurance of successful com- munication, and (d) stimulation of further interactions.

Fantini (1985) showed that various channels of language input in the home environment, such as communication be- tween and with family members and communication with people outside the family may influence minority children’s language development. According to Tosi (1979), lack of rein- forcement of accepted language norms and exclusion from exposure to the standard language can be responsible for weakening Ll development. With reference to minority language speaking immigrants, Tosi (1984) pointed out that in the case of first-generation children, L1 will often be in a favorable position. Its development may originally benefit from a rich infrastucture in the family, but later on, if not reinforced by conditions of full exposure, the mother tongue may develop only in restricted domains. Accordingly, the tendency to shift to the majority language may grow stronger.

208 Language Lea m ing Vol. 41, No. 2

Schumann (1978) pointed out that contact of family mem- bers with target-language speakers is crucial for L2 development. Vermeer (1985) and Verhoeven (1987) indeed found that the degree to which Dutch is used in the family is related to the L2 proficiency of minority children in Holland. According to Wong Fillmore (1985)’ the interaction with target-language-speaking peers is a strong determinant of children’s L2 development.

Another relevant family characteristic is parental attitude toward L1 maintenance and L2 learning. Gardner (1968) has shown that parental attitudes have an effect on the attitudes and motivation of children. He found that Anglophone parents with positive attitudes toward French Canadian develop simi- lar attitudes in their children and have better success in L2 learning than do children of parents with unfavorable atti- tudes. Lalleman (1986) investigated the relationship between L2 proficiency in 6-year-old Turkish children in The Nether- lands and various attitudinal variables among their parents. She found significant correlations between children’s L2 pro- ficiency on the one hand, and the cultural attitudes of parents and the contact parents had with L2 speakers on the other.

INSTITUTIONAL CARE

The language development of children in modern indus- trialized societies increasingly occurs within an institutional context. Day care and kindergarten play an important role in children’s early language learning. Institutional care presents minority children with a set of new interpersonal relationships. The interaction with peers who speak L1 or L2 as a native language will enhance their language development. Wong Fillmore (1982) found that in classrooms displaying an open organization the presence of target-language-speaking peers supports minority children’s L2 development. Institutional contexts also give minority children the opportunity to use language in a meaningful way and to receive feedback from professional caretakers. In many studies it was found that the

Verhoeven 209

quality of interaction patterns is extremely relevant for minor- ity children’s progress in second-language development (see McLaughlin, 1985, pp. 145-163). Moreover, the extensive set of studies conducted in the Canadian context demonstrated that children participating in bilingual programs learn to express a healthy self-identity and appreciation for their own linguistic and cultural membership (cf. Harley & Lapkin, 1984; Cummins & Swain, 1986).

With respect to the institutional care of minority children there is considerable discussion on the effectiveness of monolin- gual and bilingual programs. Evaluation reports showed that in many instances an exclusive use of the majority language in education brought about an enormous drop in academic achievement of minority children (Cummins, 1979; Tosi, 1984). However, the discussion was complicated by extremely contra- dictory results in different educational programs dealing with second-language instruction: immersion versus submersion programs. Immersion programs have the following character- istics: the children usually belong to the prestigious and dominant group; their home language is respected; all other childrenin the classroom are from the same language background; their parents are supportive of the program; teachers have high expectations for children’s achievement, and the mother tongue is brought in as a second medium of instruction during the course of the program. Submersion programs, on the other hand, are usually set up for children speaking a low-status language, and follow a monolingual “sink or swim” method without much support from the teacher or the parents. In immersion programs the home-school switch seemed to have no detrimental effects, whereas in submersion programs such a switch proved to lead to poor academic achievement and an inadequate command ofboth L1 and L2. According to Skutnabb- Kangas (1984) this paradox can be explained as follows. In immersion programs the mother tongue of children, having a high prestige, is not threatened, so the acquisition of a second language can occur with no loss of L1. However, in submersion

210 Language Learning Vol. 41, No. 2

programs the mother tongue has a low prestige and L2 may represent a threat to L1. As a consequence, the social environ- ment will create ambivalence on the part of the minority language and the majority group, in addition to insecurity toward the minority language and the minority group. As a result, the learningofL2 will reflect the loss or stagnation of L1. Skutnabb-Kangas claims that in a submersion environment minority language education is necessary to develop the children’s L1 and that this is in turn a prerequisite for a successful acquisition of L2. In support of this claim, Cummins (1981) has put forward the “interdependency hypothesis,” which assumes that if the outside environment provides sufficient stimulus for L1 maintenance, intensive exposure of L2 in the school leads to a rapid bilingual development with no negative effects on L1. In cases in which the L1 is not sufficiently developed outside the school, high exposure to an L2 in the school will hamper the continued development of L1. In turn, that L1 remains poor will prove a limiting factor in the development of L2.

Although the interdependency hypothesis has never been tested with minority children at the preschool level, it can be expected that the use of L1 in institutional care may benefit children’s bilingual development in two ways. First, adequate feedback in L1 will enhance children’s L1 proficiency which in turn will support L2 development. Second, the use of L1 may enhance children’s self-respect to discover that their home language is respected by the institution. As a conse- quence, their motivation to learn L2 will also increase.

In sum, it is shown that the level of proficiency minority children develop in L1 and L2 is related to various variables originating from the child, the family the child belongs to, and the institutional care the child has gone through. However, individual differences in bilingual proficiency may well have multiple causes. The possibility of more complex multiple- factor theories must therefore be admitted. Okamura-Bichard (1985) conducted a study of multiple predictors of the first and

Verhoeven 21 1

second language proficiency of 12-year-old Japanese children in the United States. She found the children’s attitudes and the extent of language use to be the most relevant predictors of proficiency in either language. Parental attitudes proved to be an additional predictor of proficiency in L1; years of schooling in the U.S. proved to be another predictor of proficiency in L2.

However, studies taking into account a multiple-factor design in predicting the level ofbilingual proficiency in younger children are generally lacking. In the present study such a design is taken as a starting point in predicting the bilingual proficiency of 6-year-old Turkish children living in Holland. These children form part of the second generation of Turkish immigrants who came to The Netherlands during the past decades. Their parents are predominantly Turkish speaking with only limited competence in the second language (Dutch). The children, most of whom were born in Holland, participate in a linguistic network that is quite complex. Their early language input is Turkish, but soon the Dutch language enters into their lives by way ofDutch playmates and day care. By the age of 6 years, these children can often be seen as bilinguals whose Turkish and Dutch language systems are in a state of flux. Before going into the ultimate goals of this study, one important question will be considered: How can children’s first and second language proficiency be defined?

BILINGUAL PROFICIENCY

With respect to the nature of language proficiency there is little consensus among researchers. According to the early views of Oller (1978) and many others, there is only one global language proficiency factor that accounts for the variance in a wide range of language measures. However, as has been indicated by Bialystok (1981, 1982) the monolithic concept of language proficiency can be questioned on theoretical grounds. It seems that a multidimensional concept of language pmfi- ciency, distinguishing linguistic knowledge from language use

212 Language Learning Vol. 41, No. 2

is more adequate. For instance, Chomsky (1980) makes a distinction between grammatical competence, including all linguistic aspects of meaning, and pragmatic competence, re- ferring to the ability to use linguistic knowledge along with the conceptual system to achieve certain language purposes. A similar distinction was proposed by Palmer (1979). Other authors define the construct of language proficiency in terms of a greater set of abilities, including components, such as sociolinguistic competence, strategic competence, and illocutionary force (Canale & Swain, 1980; Bachman & Palmer, 1982; Canale, 1983; Bachman, 1988).

Cummins (1979, 1980) found empirical evidence for a pluralis tic concept of language proficiency. He found perfor- mance by learners on a variety oflanguage tasks to be clustered into two sets. These sets couldbe labeled as cognitivdacademic language proficiency and basic interpersonal communicative skills. The former are general cognitive or academic skills, such as the range of vocabulary and the knowledge of complex syntax; the latter skills are closely related to the use of lan- guage and interpersonal communication. In later publications Cummins integrated the distinction between basic interper- sonal and cognitive/academic language skills in a new theoretical framework(Cummins, 1984; Cummins & Swain, 1986). In this framework two continuums are distinguished to conceptualize language proficiency: (a) a horizontal continuum relating to the range ofcontextual support for the expression and understanding of meaning and (b) a vertical continuum addressing the degree of active cognitive involvement for appropriate performance on a task. Departing from this framework, a task is defined as more academic as the context-reduction and the cognitive demands increase.

With reference to bilingual education, Cummins claims that surface aspects, such as oral fluency, develop separately in L1 and L2, but that an underlying academic proficiency is common across languages. This common underlying profi- ciency is said to facilitate the transfer of cognitivdacademic

Verhoeven 213

abilities from one language to the other. As such, a theoretical foundation for the interdependency hypothesis, as mentioned above, was given. However, Snow (1987) argues against the idea that conversational skills in L1 and L2 are separately developed. She claims that apart from language specific achievements, nonlinguistic interactive skills must underlie conversational skills in L1 and L2.

In an earlier study (Verhoeven, in press) we found empiri- cal evidence for a multidimensional concept of the bilingual proficiency of 72 six-year-old Turkish children. In both L1 and L2, pragmatic conversational skills could be distinguished from context-reduced grammatical skills. Moreover, it was found that the two types of language skills in L1 were significantly related to corresponding skills in L2. In the present study, it will be examined which child, family, and institutional variables relate to the children’s proficiency levels in Turkish and Dutch. A multiple-factor approach will be followed to find the best predictors of children’s bilingual proficiency. Moreover, it will be determined to what extent there is evidence for inter- dependency in the process of first- and second-language development. Special attention will be given to the role of first language use in day-care institutions.

DESIGN OF THE STUDY

SUBJECTS

A group of 72 Turkish children was selected from 10 kindergarten classes in fairly big cities in the eastern part of Holland. All subjects belong to the large group of Turkish immigrants who came to The Netherlands during the 1970s. The parents of the children originated in villages in Central Anatolia and the Black Sea region in Turkey. In Holland, they all worked in factories or were owners of small shops.

The group of subjects consisted of 36 boys and 36 girls,

214 Language Learning Vol. 41, No. 2

their ages ranging from 6.1 to 6.9 years with a mean age of 6.4 years. All the children have Turkish as their native language. Fifty-eight of them were born in The Netherlands, the others had been living there for at least four years. The mean period of residence was 5.5 years with a standard deviation of 1.1 years. All informants had attended nursery school in Holland.

Earlier studies have shown that the proficiency level of Turkish children at kindergarten level in either language is lower than that ofmonolingual peers. Verhoeven & Boeschoten (1986) found that the first language progress of Turkish chil- dren in The Netherlands in the age range from 4-8 years was slower than that of peers living in Turkey. Verhoeven & Vermeer (19851, on the other hand, found that the rate of acquisition of Dutch by Turkish children in the same age range clearly lags behind that of their Dutch peers.

LANGUAGE PROFICIENCY MEASURES

An attempt was made to operationalize pragmatic con- versational and decontextualized grammatical aspects of the children’s first and second language proficiency. The assess- ment of pragmatic language proficiency started with the recording and transcription of spontaneous speech using three different elicitation contexts: (a) givinga spatial description, (b) describing a series of events, both on the basis of pictures, and (c) some free conversationbetween the child and the interviewer. For Turkish the interviewer was a native speaker of Turkish, for Dutch the interviewer was a native speaker of Dutch. Several parameters of language use were assessed from the transcriptions. As measures of lexical variety in discourse, the numbers of different content words and function words occur- ring in 75 utterances (25 random utterances from eachelicitation context) were computed; utterances being defined as in Hunt (1970). To measure the use of morphosyntactic devices in the two languages, the mean number of morphemes in the longest utterances (10% of each individual corpus) was computed.

Verhoeven 215

Moreover, the percentage of subordinated clauses over 150 utterances (50 random utterances from each elicitation con- text) was determined.

The assessment of decontextualized grammatical proficien- cy involved the administration of tests measuring phonological, lexical, and morphosyntactic subskills in Turkish and Dutch (cf. Verhoeven, 1987). The phonological tests requirechildren to make similar-different distinctions between word pairs that differ in one phoneme, in the position or number of phonemes. To test both productive and receptive lexical skills, children are required to label pictures and to select the correct referent out of four pictures for a spoken word. Morphosyntactic abilities in the two languages were measured by a 24-sentence imitation task. The children’s reproductions were scored for correct imitation of function words, word final markers, and clause linking.

PREDICTOR MEASURES

CHILD VARIABLES

Three child characteristics that were thought to be pos- sible predictors of children’s first and second language proficiency were assessed. As an index of cognitive capacity the Progres- sive Matrices Test (Raven, 1965) was administered.

To evaluate the cultural orientation of the children, two attitudinal measures were taken. The orientation toward cultural life in Turkey and Holland was independently as- sessed by means of a structured interview with the child. In the interview the attitudes children had toward 10 different as- pects of cultural life in Turkey and The Netherlands were examined: school, residence, food, language, play, music, books, television, friends at home, and friends a t nursery school. The child’s orientations were rated along a 4-point-scale, ranging from extreme like to extreme dislike, based upon their answers to questions (e.g., ‘Would you like to be in school in Turkey?”). The interrater reliability turned out to be high (.96).

216 Language Learning Vol. 41, No. 2

The other attitudinal measure taken was the children’s attitudes toward stereotypic symbols, using a five-point scale similar to the one used by Zirkel and Green (1976). Attitudes toward Turkish and Dutch symbols were independently mea- sured. Children were asked to react to pictures of symbols of everyday life, such as food, clothes, famous persons, etc., by selecting one out of five faces arranged on a happy-sad Likert scale. Turkish and Dutch items were presented in a random order, in a booklet. Previous research had shown that the present cultural attitude scale is sufficiently reliable with KR20 coefficients ranging from .75 to .91 for different samples of children (de Bot, Broeder, & Verhoeven, 1985).

FAMILY VARIABLES

Three family variables were under consideration. The first two family variables, assessed in a semistructured interview in Turkish, were the language interaction in the family context and the cultural attitudes of parents. Questions about the family interaction pattern concerned use of Turkish and Dutch within the family, contact with Turkish and Dutch speakers in the home context and the use of Turkish and Dutch media. Questions about the parents’ attitudes concerned the celebra- tion of Turkish and Dutch holidays and the ideas they had about their children’s education in Turkish and Dutch. The third family variable was parents’ cultural behavior. In a questionnaire for the kindergarten caretaker, 10 questions were asked about the child rearing practice of parents in cultural domains that show up in school (i.e., dress, food for lunch, and the celebration of holidays).

INSTITUTIONAL CARE VARIABLES

Five variables related to the institutional care the children had gone through were assessed. First, the experience of institutional care was determined by computing the total num-

Verhoeven 217

ber of months each child had been taken care of in an institu- tional setting, such as a day-care center or a kindergarten (starting at age 4). In Holland there is no compulsory education for children younger than 5 years old. Until that age parents can bring their children to a day-care center on a voluntary basis. Furthermore, it should be noted that the atmosphere in kindergarten and day-care centers is highly comparable, in that in either condition a formal curriculum is avoided in favor of play situations.

Second, the presence of peers speaking L1 and L2 as a native language was examined by computing the average proportion of LlLZspeaking peers during the child’s period of institutional care.

Third, the extent of peer interaction in the two languages was measured by means of a questionnaire for the caretaker in which various aspects of the children’s contact with Turkish and Dutch peers during day-care time had to be rated.

Fourth, the extent of caretaker interaction in the two languages was assessed by computing the average number of hours per week of direct caretaker-child interaction in Turkish and Dutch. These numbers were based on week-to-week estimates by the caretakers in question. In Dutch day-care centers caretaker-child interaction in ethnic minority lan- guages is quite infrequent. Given the general policy of assimilation, the Dutch authorities question the usefulness of first language interaction in day-care centers. Accordingly, the facilities for such interaction, ifgiven a t all, are very small. For the day-care centers in our study, caretaker-childinteraction in Turkish was absent in about 40% ofthe cases. In the remaining cases, it made up only a small part of the timetable (4-8 hours a week). With respect to child-caretaker interaction in Dutch, it is worth noting that in some kindergarten classes there were extra caretakers available to help slow L2 learners.

Fifth, the extent of parental involvement during the time of their children’s institutional care was evaluated by means of a questionnaire for the caretaker. In it, questions were asked

218 Language Learning Vol. 41, No. 2

Table 1 Taxonomy of Predictor Measures

Predictor Measure Methodhstrument

Child Variables Cognitive capacity Raven Progressive Matrices Orientation cultural life Child report (questionnaire) Orientation cultural symbols Cultural Attitude Scale

Family Variables Family interaction in L l L 2 Parents’ cultural attitudes Parents’ cultural behavior

Parent report (questionnaire) Parent report (questionnaire) Caretaker report (questionnaire)

Institutional Care Variables Period of institutional care Number of months in care centers Presence of LUL2-speaking peers Proportion L1&2 speakers Peer interaction in L l L 2 Caretaker report (questionnaire) Caretaker interaction in LLIL2 Number of hours per week Parents’ involvement in day care Caretaker report (questionnaire)

about the parents’ presence at day care, their involvement in child activities and their contact with caretakers.

All of the questionnaires had been used in an earlier study on the acquisition of literacy by ethnic minority children (Verhoeven, 1987). In the former study as well as in the present one, the internal consistency of all the questionnaires turned out to be reasonably good: KR20 coefficients yielded values ranging from .71 to 39. Table 1 presents a summary of the predictor variables along with an abbreviated description of the methods and instrumentation used to derive each measure.

PROCEDURE

To investigate the relationship between predictor vari- ables and language proficiency in Turkish and Dutch a number of statistical procedures were followed. By means of factor

Verhoeven 219

analysis, it was determined to what extent the eight measures for proficiency in the two languages can be reduced to a smaller number of factors (with eigenvalues greater than 1.0). Follow- ing the theoretical framework proposed by Cummins (1984)’ it was hypothesized that two factors underlie the children’s pro- ficiency in either language: (a) referring to context-embedded cognitively undemanding pragmatic skills and (b) context- reduced cognitively more demanding grammatical skills.

In exploring the relationship between language profi- ciency and predictor variables, the children’s factor scores obtained in each language would be taken as the criterion variables. First of all, the correlations between the language scores and the predictor variables referring to the child, the family, and the institutional care were computed. For those predictor measures yielding two separate scores, the score referring to the minority languagekulture was related to pro- ficiency in L1, whereas the score referring to the majority languagekulture was related to proficiency in L2. An exception was made for the measure caretaker interaction. Given that attention to L1 in institutional contexts may have a beneficial effect on L2 development, the extent of caretaker interaction in L1 was taken as an additional predidor measure ofL2 proficiency.

To find the best predictors of children’s first and second language proficiency, stepwise multiple regression analyses were conducted. The notion of interdependency was taken into account by exploring to what extent the children’s skills in L1 and L2 are interrelated. If a significant correlation between the two types of skills was the case, an additional multiple regres- sion analysis would be conducted with L1 proficiency as an extra predictor of children’s L2 proficiency.

220 Language Learning Vol. 41, No. 2

Table 2 Factor Loadings of the Turkish Proficiency Measures

Factor 1 Factor 2

Use of Different Content Words Use of Different Function Words Utterance Length Proportion of Subordinated Clauses Phoneme Discrimination Receptive Vocabulary Productive Vocabulary Sentence Imitation

.54

.73

.87

.52

.I4

.06

.05

.24

.01

.09

.18

.21

.29

.74

.92

.58

RESULTS

FACTORS UNDERLYING L1L2 PROFICIENCY

Common factor analysis for the measures of Turkish language proficiency yielded a two-factor solution, explaining 36 and 23% of the total amount of variance, respectively. Additional factors yielded no substantial increase in amount of variance explained (eigenvalues lower than 1.0). Table 2 presents the factor loadings of the eight proficiency measures on the two factors after avarimax rotation had been carried out. It can be seen that the language measures with high loadings on the first factor refer primarily to language use in a natural manner. The measures of lexical variety, utterance length, and proportion of subordinated clauses all concern the unconscious use of linguistic knowledge in a communicative situation. The language measures with high loadings on the second factor, on the other hand, all refer to a context-reduced situation in which the focus of the learner is to perform a linguistic task con- sciously. The phoneme discrimination task, the vocabulary tasks, and the sentence imitation tasks all demand a conscious orientation on the part of the child.

Factor analysis on the measures of Dutch proficiency also

Verhoeven 221

Table 3 Factor Loadings of the Dutch Proficiency Measures

Factor 1 Factor 2

Use of Different Content Words Use of Different Function Words Utterance Length Proportion of Subordinated Clauses Phoneme Discrimination Receptive Vocabulary Productive Vocabulary Sentence Imitation

.a3

.70

.96

.45

.06

.19

.23

.41

.21

.22

.23

.10

.51

.79

.a3

.77

yielded a two-factor solution, explaining 50 and 19% of the variance in proficiency scores, respectively. The relative gain in variance explained by additional factors was very small (eigen- values lower than .go>. In Table 3 the factor loadings of the eight Dutch proficiency measures on the two factors are given after a varimax rotation had been carried out. The pattern of factor loadings highly corresponds to the one found for the Turkish data. Language measures with high loadings on the first factor refer primarily to interpersonal communicative skills, whereas the ones with high loadings on the second factor draw back rather on context-reduced conscious linguistic knowledge.

Thus, it can be concluded that the children’s proficiency in both Turkish and Dutch can be seen as two-dimensional con- structs. In either case, the underlying dimensions can be easily interpretedin terms ofcummind( 1984) theoretical framework in which context-embedded unconscious pragmatic skills are distinguished from context-reduced conscious grammatical skills. In subsequent analyses the children’s factor scores vrdl be used as the relevant indices for proficiency in the two languages. For the balance of this paper, acronyms will be used to indicate the underlying dimensions in the L1 (PRAGMlI GRAMMl) and the L2 (PRAGMW GRAMMB).

222 Language Learning Vol. 41, No. 2

Table 4 Correlations Between Predictor Variables and Turkish Proficiency

Measures

PRAGMl GRAMM1

Cognitive Capacity .01 Orientation Toward Turkish Life .26** Orientation Toward Turkish Symbols -.05 Family Interaction in L1 -. 19 Parent Cultural Attitudes .13 Parent Cultural Behavior -. 02 Period of Institutional Care -. 05 Presence of L1-Speaking Peers .22* Peer Interaction in L1 .08 Caretaker Interaction in L1 .05 Parental Involvement .22*

.14

.26**

.19

.07

.17

.25* -. 03

.08

. ll

.23*

.15

*p<.05; **p<.Ol

PREDICTORS OF L1 PROFICIENCY

The correlations between predictor measures on the one hand, and the children’s proficiency levels in Turkish on the other are presented in Table 4. Three measures are signifi- cantly related to PFUGM1: (a) children’s orientation toward the Turkish way of life, (b) the presence of Turkish-speaking peers in institutional care, and (c) parental involvement in institutional care. Again three predictor measures turn out to be significantly related to GRAMMl: (a) children’s orientation toward the Turkish way of life, (b) the cultural behavior of their parents, and (c) the extent of caretaker interaction in Turkish.

Table 5 presents the results of the multiple regression analyses with PFUGMl and GRAMMl as criterion variables. It can be seen that Turkish pragmatic skills are significantly predicted by three measures: (a) children’s orientation toward the Turkish way of life, (b) the presence of Turkish-speaking peers in day care, and (c) parental involvement in day care.

Verhoeven 223

Table 5 Predictors of Turkish Language Profiiency

Criterion Variable Predictor Variable 0 R

PRAGMl Orientation Toward Turkish Life .23 Presence of L1-Speaking Peers .29 Parental Involvement .25 .42**

GRAMMl Orientation Toward Turkish Life .25 Parents’ Cultural Behavior .27 Caretaker Interaction in L1 .23 .43**

**p<.o1

Taken together, these measures explain about 18% of the variance in PRAGM1 scores. The criterion variable GRAMMl is significantly predicted by the following measures: (a) children’s orientation toward the Turkish way of life, (b) the cultural behavior of parents, and (c) the extent of caretaker interaction in Turkish. About 19% of the variance in GRAMMl scores is explained by these measures.

PREDICTORS OF L2 PROFICIENCY

Table 6 presents the correlations between predictor mea- sures and children’s proficiency levels in Dutch. The extent of caretaker interaction in L1 was taken as an additional predic- tor measure, given the idea that attention to L1 may have a beneficial effect on children’s L2 development. Three predictor measures turn out to be significantly related to PRAGM2: children’s orientation toward the Dutch way of life, the extent of family interaction in L2, and the extent of caretaker interac- tion in L1. Not less than nine predictor measures yield a significant correlation with GRAMM2. The correlations are positive in eight cases: children’s cognitive capacity, parental attitudes toward the Dutch culture, the extent of family inter- actionin L2, the cultural behavior ofparents, the peer interaction

224 Language Learning Vol. 41, No. 2

Table 6 Correlations Between Predictor Variables and Dutch Proficiency

Measures

PRAGM2 GRAMM2

Cognitive Capacity Orientation Toward Dutch Life Orientation Toward Dutch Symbols Family Interaction in L2 Parent Cultural Attitudes Parent Cultural Behavior Period of Institutional Care Presence of L2-Speaking Peers Peer Interaction in L2 Careteker Interaction in L1 Caretaker Interaction in L2 Parental Involvement

*pc.05; **pc.Ol; ***pc.OOl

.04

.25* -.11 .21* .19 .13 .12 .09 .19 .42***

-. 19 .ll

.26*

.19 -. 05

.46***

.20*

.27*

.20*

.04

.45***

.32** -.21*

.23*

in L2, the extent of caretaker interaction in L1, and the extent of parental involvement in day-care. For the measure caretaker interaction in L2 the correlation is negative. The finding that both pragmatic and grammatical skills in Dutch yield negative correlations with the extent of caretaker interaction in L2 can be explained by the fact that in some groupings of day care, extra caretaker interaction was supplied in cases of L2 learning problems.

Table 7 presents the main statistics ofthe multiple-regres- sion analyses with PRAGMB and GRAMM2 as the criterion variables. The results reveal that Dutch pragmatic skills are significantly predicted by two predictor variables: (a) caretaker interaction in L1 and (b) parental attitudes toward the Dutch culture. Taken together, these measures explain 25% of the variance in PRAGMZ scores. The criterion variable GRAMMZ is significantly predicted by four measures: (a) family interac- tion in L2, (b) peer interaction in L2, (c) cognitive capacity, and

Verhoeven 225

Table 7 Predictors of Dutch Language Proficiency

Criterion Variable Predictor Measure P R

PRAGMB Caretaker Interaction in L1 .46 Parents’ Cultural Attitudes .26 .50***

GRAMMB Family Interaction in L2 .27 Peer Interaction in L2 .37

Caretaker Interaction in L1 .19 .63*** Cognitive Capacity .19

***p<.oo1

(d) caretaker interaction in L1. About 40% of the variance in GRAMM2 scores is accounted for by these measures.

ROLE OF LINGUISTIC INTERDEPENDENCY

To test Cummins’ (1979) claim of interdependency of linguistic skills in L1/L2, the relationship between Turkish and Dutch proficiency measures was explored. A significant posi- tive correlation was indeed found between the children’s GRAMM scores in the two languages (r=.26, p<.05). For that reason, another multiple-regression G W M 2 score was con- ducted with GFtAMMl as an additional predictor measure. The results are given in Table 8. It can be seen that GRAMMl is a substantial predictor of GRAMM2. It seems that GRAMMl as a predictor measure has taken the place of the measure care- taker interaction in L1. The results also reveal that the inclusion of GRAMMl as a predictor measure leads to an increase of the proportion of explained variance in GRAMMB scores. Taken together, the predictors family interaction in U, peer interaction in L2, GRAMM1, and cognitive capacity ex- plain 46% of the variance in GRAMM2 scores.

Another significant correlation was found between the children’s PRAGM scores in the two languages (r=.49, p<.O1).

226 Language Learning Vol. 41, No. 2

Table 8 Regression of GRAMMB and PRAGM 2 With GRAMMl and

PRAGM 1 as Additional Predictor Measures

Criterion Variable Predictor Measure I3 R

GRAMMB Family Interaction in L2 .35 Peer Interaction in L!2 .40 GRAMM1 .30 Cognitive Capacity .16 .68***

PRAGM2 Caretaker Interaction in L1 .42 PRAGMl .43 Parental Cultural Attitudes .15 .65***

***p<.oo1

Thus, an additional multiple-regression analysis was conducted with PRAGMl as an additional predictor measure of PRAGM2 scores. These results are also presented in Table 8. PRAGMl turns out to be a substantial predictor of PRAGM2. The propor- tion of explained variance in PRAGM2 scores increased to 42%.

DISCUSSION

The present study suggests that multiple predictors ac- count for individual differences in the bilingual proficiency of minority children by the age of 6 years. According to the results of the study, the prediction of children’s first and second language proficiency is complicated in various ways. First of all, there is clear evidence that the proficiencies children have developed in L1 and L2 cannot be conceived of as monolithic traits. The present data show that in either language, context- embedded unconscious pragmatic skills can be distinguished from context-reduced conscious grammatical skills. This result is consistent with the concept of language proficiency as a multidimensional construct. Furthermore, the present study

Verhoeven 227

makes clear that there are different predictors of children’s L1 versus L2. The prediction of first language proficiency turned out to be relatively low. This finding can at least partly be explained by the fact that first-generation immigrants do not substantially differ in respect of their attitudes toward the home culture and their first language interaction patterns (cf. Tosi, 1984). The present study also gives evidence that the two underlying dimensions in both L1 and L2 have at least partially distinct antecedents. However, some common relationships between predictor variables and children’s first and second language skills were uncovered as well.

There is some indication that certain sociocultural factors are positively related to all proficiency levels under consider- ation. The finding that the children’s orientation toward cultural life is substantially related to their bilingual profi- ciency level lends support to the claim that the degree of identification of minority children with both the minority and the majority culture is a crucial factor for L1 and L2 develop- ment (see Lambert, 1967; 1978). The cultural attitudes of parents also turned out to be positively related to the children’s L1L2 proficiency level. This finding is in line with Fishman’s (1977) claim that parents play animportant rolein the children’s acculturation and in their language development. There is also some evidence that the involvement of parents in institutional care is positively related to the children’s L1 and L2 develop- ment. This result is consistent with the outcome of earlier studies examining the relationship between the involvement of parents in school and the language and academic performance of their children (see Stevenson & Baker, 1987).

Highlighting the relationship between multiple predictor measures and the language proficiency levels in ethnic minor- ity children provides a better understanding of the factors underlying the course of bilingual development. With respect to L1, it was found that attitudinal measures related to the child and to his or her parents account for much of the variation in proficiency scores. The proportion of L1-speaking peers in day

228 Language Learning Vol. 41, No. 2

care turnedout tobe an additional measure predictingchildren’s conversational skills, whereas the extent of caretaker interac- tion in L1 was an additional measure predicting their grammatical skills. Apparently, the opportunity to speak with peers is sufficient to develop the former kinds of skills, whereas feedback from a professional caretaker is needed to promote the latter kinds of skills.

With respect to L2, it was found that the children’s commu- nicative skills are primarily predicted by their orientation toward the Dutch way of life and the extent of caretaker interaction in L1. The importance of feedback in L1 for L2 development lends support to Skutnabb-Kangas’ (1984) argu- ment that the motivation to learn an L2 increases as societal institutions pay more attention to the native language and culture of minority children. The children’s grammatical skills in L2 were related to both social and cognitive factors. Contact with L2 in the family and among peers turned out to be highly relevant measures.

Furthermore, the children’s abilities in L2 could be pre- dicted from similar abilities developed earlier in L1. As such, empirical evidence was found for Cummins’ ( 1984) interdepen- dency hypothesis. With respect to grammatical abilities, it is interesting to note that besides GRAMMl, cognitive capacity appeared as a predictor of GRAMM2. Apparently, there is information that is of predictive value in GRAMMl that is not in the measure of general cognitive abilities. Thus, it can be tentatively concluded that the nature of the interdependency hypothesis is at least partly linguistic. Another interesting finding is that there was even more evidence for transfer at the level of pragmatic skills. This result is in accord with Snow’s (1987) claim that there must be nonlinguistic interactive skills underlying conversational skills that are easily transferable from one language to the other.

With respect to the present findings, a few words ofcaution are in order. First, the direction of effect between predictor and criterion variables can be questioned. It can be argued that

Verhoeven 229

several important relationships, for which causality from socio- cultural factors to child language is assumed, can just as easily be reversed. However, scenarios arguing for the importance of the child's language proficiency in leading to higher scores on sociocultural measures seem unlikely in the acculturation process of first- or second-generation immigrants (cf. Schu- mann, 1978). Another question that remains concerns the generalizability of data from minority children in The Nether- lands to minority children in other countries, such as the U.S. There are obvious differences in demographic trends of ethnic minorities in Holland and the U.S., resulting in different patterns of language use. Although the linguistic background ofminorities in theU.S. is generally more stabilized thanis that of minorities in The Netherlands, important cross-national similarities can be pointed out as well. In both countries, most minorities have a low socioeconomic status, determined by a low level of education and profession, or employment. More- over, ethnic minority languages usually have a low status. These similarities make sure that in both countries most minority children acquire language in a second language sub- mersion environment. As has been pointed out by Cummins (1979), it is quite reasonable to assume that research data on predictors of early language proficiency in a second language submersion context can easily be generalized from one situa- tion to the other.

The findings from the present study have important impli- cations for theinstitutional care ofminoritychildren. Caretaker interaction in L1 proves to be crucial for the development of cognitive/academic skills in both L1 and L2. Policy makers should account for this finding by supporting bilingual pro- grams at the preschool level. It can be hypothesized that the caretaker interaction in L1 stimulates L2 development in at least two ways. First, there may be a direct influence in that the attention to the native language of minority children will enhance their self-respect. It will make children discover that their mother tongue, in which they have invested several years,

230 Language Learning Vol. 41, No. 2

is respected by the day-care center. There can also be an indirect effect: sufficient support to the first language may enhance children’s cognitive/academic skills in that language, which in its turn may have a positive effect on their cognitive/ academic skills in the second language. Besides caretaker interaction, the opportunity for children to use L1 and L2 in interaction with peers appears relevant as well. Not only should caretakers create opportunities for language use, they should also help children gain peer acceptance in the new environment for the child. Pursuing more extensive positive contacts prior to day care may be helpful in this respect. In addi tion to implications concerning the institutional care, the present data suggest the need to consider factors such as the child’s cultural orientation and the involvement of parents in day care. In terms of children’s cultural orientation, it may be helpful to intervene during preschool to modify negative atti- tudes toward the minority culture or majority culture. With regard to parental involvement, it may be beneficial to inform parents about the child’s behavior and to invite parents to engage in day-care activities.

Naturalistic studies, like the present one, can be seen as a first step toward answering questions about determinants of bilingual development in a second language submersion envi- ronment. A more complete answer can be expected from longitudinal studies in which both language data and detailed observations concerning the sociocultural and linguistic back- ground of informants are collected. In that case, an explanation for individual differences in acquisitional patterns could be sought.

REFERENCES

Bachman, L. F. (1988). Problems in examining the validity of the ACTFL oral proficiencyinterview. Studies in SecondLangungeAcquisition, 10,149-164.

Bachman, L. F., & Palmer, A S. (1982). The construct validation of some components of communicative competence. TESOL Quarterly, 16,449-465.

Verhoeven 231

Bialystok, E. (1981). The role of linguistic knowledge in second language use. Studies in Second Language Learning, 4,3145.

Bialystok, E. (1982). On the relationshipbetween knowing and usinglinguistic forms. Applied Linguistics, 3, 181-206.

Canale, M. (1983). On some dimensions oflanguage proficiency. In J. W. Oller, Jr. (Ed.), Issues in lnnguage teaching research (pp. 332-342). Rowley, Massachusetts: Newbury House Publishers.

Canale, M., & Swain, M. (1980). Theoretical bases of communicative ap- proaches of secondlanguage teachingand testing.Applied Linguistics, 1,147.

Chomsky, N. (1980). Rules and representations. New York, New York: Colum- bia University Press.

Cummins, J. (1979). Linguistic interdependence and the educational develop- ment of bilingual children. Review of Educutioml Research, 49,221-251.

Cummins, J. (1980). The cross-lingual dimensions of language proficiency. TESOL Quarterly, 14,175-187.

Cummins, J. (1981). The role of primary language development in promoting educational success for language minority students. In Schooling and &an- guage minority students: A theoretical jhamework (pp. 349). Los Angel-, California: Department of Education, Evaluation, Dissemination and As- sessment Center.

Cummins, J. (1984). Wanted A theoretical framework for relating language proficiency to academic achievement among bilingual students. In C. Rivera (Ed.),Lunguagepmficiency and academic achievement (pp. 2-19). Clevedon, Avon: Multilingual Matters.

Cummins, J., & Swain, M. (1986). Bilingualism in education. London, United Kingdom: Longman.

de Bot, C., Broeder, P., & Verhoeven, L. T. (1985). Sociaalculturele orientatie en taalvaardigheid bij allochtone kinderen [Socialcultural orientation and language proficiency of ethnic minority children]. Toegepaste Taalkun.de in Artikelen, 22,3349.

Fantini, A. E. (1985)Languageacguisition ofa bilingualchild. Clevedon, Avon: Multilingual Matters.

Fishman, J. (1977). The social science perspective. In Center for Applied Lin- guistics(Ed.), Bilingualedmtion: Currentperspectives (pp. 149).Arlington, Virginia: Center for Applied Linguistics.

Gardner, R. C. (1968). Attitudes and motivation: Their role in second language acquisition. TESOL Quarterly, 2, 141-150.

Genesee, F., & Hamayan, E. (1980). Individual differencesin secondlanguage learning. Applied Psycholinguistics, I, 95-110.

Harley, B., & Lapkin, S. (1984). The effects of early bilingual schooling on first language development. Toronto, Ontario: Ontario Institute for Studies in Education.

232 Language Learning Vol. 41, No. 2

Hunt, K. W.(1970). Syntacticmaturityin school childrenandadults.Monograph

Lalleman, J. A. (1986). Dufch languagepmfiiency of Turkish children born in

Lambert, W. (1967). A social psychology of bilingualism. Journal of Social Is-

Lambert, W. (1978). Psychological approaches to bilingualism, translation and interpretation. In D. Gerver & H. W. Sinaiko(Eds.),Language interpretatinn and communiccrtion (pp. 131-144). New York, New York: Plenum Press.

McLaughlin7B.(1985).Secondlanguageacquisitioninchildhood. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Okumura-Bichard, F. (1985). Mother tongue maintenance and second lan- guagelearning: AcaseofJapanese children.LanguageLearning735, 63-89.

Oller, J. W., Jr. (1978). The language factor in the evaluation of bilingual education. In J. W. Alatis (Ed.), GURT 1978 (pp. 410422). Washington, DC: Georgetown University Press.

Oller, J. W., Jr. (1980). Language is intelligence? Language Learning, 31,465- 492.

Palmer, A. (1979). Compartmentalized and integrated control: An assessment of some evidence for two kinds of competence and implications for the classroom. Language Learning, 29,169-180.

Raven, J. (1965). Guide to using the Coloured Progressive Matrices. London, United Kingdom: H. K. Lewis.

Schumann, J. (1978). The acculturation model for secondlanguage acquisition. In R. Gingras (Ed.), Second language acquisition and foreign language teaching (pp. 27-50). Arlington, Virginia: Center for Applied Linguistics.

Skutnabb-Kangas, T. (1984). Bilingualism or not: The e d m t w n of minorities. Clevedon, Avon: Mu1 tilingual Matters.

Snow, C.E.( 1987).Beyondconversation: Secondlanguagelearners’acquisition of description and explanation. In J. P. Lantolf & A. Labarca(Eds.), Research in second language hrning: Focus on the classroom (pp. S16). Norwood, New Jersey: Ablex.

Stevenson, D. L., & Baker, D. P. (1987). Family-school relation and the child’s school performance. Child Development, 58,1348-1357.

Tosi, A. (1979). Mother tongue teaching for the children of migrants. Language Teaching and Linguistics: Abstracts, 12,213-231.

Tosi, A (1984).Zmmigmtion and bilingualeducation. Oxford, United Kmgdom: Pergamon Press.

Verhoeven, L. T. (1987).Ethnicrninoritychildren acquiringliteracy. Dordrecht, Holland Foris Publications.

Verhoeven, L. T. (in press). Assessment of bilingual proficiency in ethnic

of the Society for Research in Child Development, 35, 1.

The Netherlands. Dordrecht, Holland: Foris.

SUB, 23,91-109.

Verhoeven 233

minority children. In L. T. Verhoeven & J. H. A. L. de Jong (Eds.), The construct of language proficiency. Amsterdam, Holland John Benjamins.

Verhoeven, L. T., & Boeschoten, H. E. (1986). First language acquisition in a second language environment. Applied Psycholinguistics, 7,241-256.

Verhoeven, L.T., & Vermeer, A. (1985). Ethnic group differences in children’s oral proficiency of Dutch. In G. Extra & T. Vallen (Eds.), Ethnic minorities and Drrtch as a second language (pp. 105-132). Dordrecht, Holland: Foris Publications.

Vermeer, A. (1985). The influence of social factors on tempo and structure ofL2 acquisition. In G. Extra & A Vallen (Eds.), Ethnic minorities andDutch as a second language (pp. 49-64). Dordrecht, Hol1and:Foris.

Wells,G. (1985).Language~velopmentinthepreschoolyears. Cambridge,United Kingdom: Cambridge University Press.

Wong Fillmore, L. (1976). The second time around: Cognitive and social strut- egies in second language acquisition. Unpublished doctoral dissertation. Stanford, California, Stanford University.

Wong Fillmore, L. (1979). Individual differences in second language learning. In C. J. Fillmore, W. S. Wang & D. Kempler (Eds.), Individual differences in krnguageahility and language behavior (pp. 203-228). New York, New York: Academic Press.

Wong Fillmore, L. (1982). Instructional language as linguistic input: Second language learning in classrooms. In L. Cherry-Wilkinson (Ed.), Cornrnuni- cationintheclassroorn(pp. 283-’296).NewYork,NewYork: AcademicPress.

Wong Fillmore, L. (1985). Second language learning in children: A proposed model. In J. Provenzano (Ed.), Issues in language development (pp. 33-42). Rosslyn, Virginia: National Clearinghouse for Bilingual Education.

Zirkel, P., & Green, J. (1976). Cultural attitude scales. In A Simoes (Ed.), The bilingual child (pp. 3-16). New York, New York: Academic Press.