Bilingual Children’s Word
Awareness Revisited
Hui-tzu Min
National Cheng Kung University minhuitz@mail.ncku.edu.tw
Abstract
The purpose of this study was to examine if the bilingual advantage of word awareness reported in an earlier study can last to the second year. 30 bilingual and 30 monolingual preschoolers (mean age 5) participated in this study. Three word awareness tasks were used to examine the marginal bilingual and monolingual children’s word awareness skills: a word segmentation task (segmenting words and word count), word judgment task, e.g., word size judgment, “maomaochong” [caterpillar] (a long word but a creature small in size) vs. zhu [pig] (a short word but an animal large in size), and name manipulation task (declarative/procedural knowledge in label change). The results show that the bilingual group continued to significantly outperform their monolingual counterparts on the name manipulation task. In addition, they also outperformed their monolingual counterparts on word count and incongruous word size judgment tasks. These findings corroborate Bialystok’s (1987) proposal that the more sophisticated development of metalinguistic skills by children learning a second language does not hinge on comparable competence in the two languages, as proposed by Cummins (1987). Rather, it demonstrates that the child’s recognition of the coexistence of two separate language systems in representing the same set of concepts is the key to their enhanced word awareness (Bialystok, 1988).
Key words: word awareness, word size judgment, word segmentation, word/referent arbitrary relationship, bilingual children, monolingual children
INTRODUCTION
Numerous studies have been conducted to test the validity of the hypothesis that bilingual children are more advanced in terms of the development of metalinguistic awareness and to what extent they are more advanced. The empirical results are equivocal—some confirm the hypothesis, while others deny it. To date, the majority of the empirical research has focused on lexical awareness, with the intent to investigate whether bilingual children have heightened awareness of the arbitrary word-referent relationships as a result of their experience with two different ways of referring to the same object. A number of studies have reported advantages in favor of bilinguals in general (Ben-Zeev, 1977a; Ianco-Worrall, 1972) without specifying the degree of bilingualism, while others have reported positive effects of bilingualism with limited exposure to a second or foreign language (e.g., Bialystok, 1986a, 1986b, 1988; Eviatar & Ibrahim, 2000; Hakuta, 1987, 1988; Rubin & Tyler, 1989; Yelland, Pollard, & Mercuri, 1993). Still others have found that these effects were limited to children who have achieved high facility in both languages (Cummins, 1978, 1987; Goncz & Kodzopeljic, 1991; Ricciardelli, 1992). On the other hand, some studies found no quantitative differences between their monolingual and bilingual subjects, albeit qualitative differences were noted (Ben-Zeev, 1977b; Cummins, 1978; Nicoladis, 1992; Nicoladis & Genesee, 1996; Rosenblum & Pinker, 1983).
The majority of the research on the effects of bilingualism on children’s word awareness skills has been conducted primarily with
native speakers of English learning an Indo-European second language such as French (Bialystok, 1988; Nicoladis & Genesee, 1996; Rubin & Tyler, 1989), Spanish (Galambos & Goldin-Meadow, 1990; Galambos & Hakuta, 1988), Italian (Ricciardelli, 1992; Yelland et al., 1993), Swedish (Cromdal, 1999), and Serbo-Croatian (Goncz & Kodzopeljic, 1991), or a Semitic language like Hebrew (Rosenblum & Pinker, 1983). Only a rare few examine non-English speakers of other languages such as Russian-Hebrew (Eviatar & Ibrahim, 2000). There is a dearth of studies on the effects of foreign language learning on metalinguistic awareness, especially word awareness among Taiwanese children who are learning English as a foreign language. One might wonder if the beneficial effects frequently noted in Indo-European languages can also extend to the particular foreign language-learning context in Taiwan, in which Mandarin is the first language and English is the foreign one. From a psycholinguistic perspective, the accelerated metalinguistic awareness observed in various studies should be universally independent of the specific languages involved because Chinese learners of English also need to actively attend to language as an object in order to master the new language. In an observational study, Min (1999) documented numerous examples of Taiwanese preschoolers’ (3 years, 4 months) provision of Mandarin equivalents for newly learned English vocabulary items. The ability to supply a Mandarin equivalent for an English term was deemed to be evidence of those young children’s implicit awareness of the arbitrary relationship between word and object. Despite the convincing argument in diary studies (Leopold, 1949; Vygotsky, 1962) that this capability is a precursor to bilingual
children’s separation of sound and meaning (an essential component of word awareness), no research has ever been conducted to test the strength of such a link among Mandarin-English-speaking children (but see Min, 2001). The purpose of this study was to continue examining the beneficial effects of bilingualism on children’s word awareness in an earlier study (Min, 2001) and investigate whether enhanced word awareness can continue to the second year.
LITERATURE REVIEW
Researchers have long been intrigued by the question of who enjoys more advanced metalinguistic awareness: monolingual or bilingual speakers. The research body as a whole has yielded equivocal results for monolingual and bilingual children (Bialystok, 2001). Some studies declare the advantage to be in one or the other of the two groups (Ben-Zeev, 1972; Bialystok, 1986a, 1986b, 1988; Cummins, 1978, 1987; Feldman & Shen, 1971; Goncz & Kodzopeljic, 1991; Ianco-Worrall, 1972; Yelland et al., 1993), whereas others assert that there is no difference between the groups (Ben-Zeev, 1977; Nicoladis & Genesee, 1996; Rosenblum & Pinker, 1983). In order to account for the discrepant findings of bilingualism on children’s word awareness, Cummins (1979) proposed a dual threshold hypothesis as a partial account of inconsistencies in the literature. He maintained that bilingual children need to attain a lower threshold (as yet unspecified) in both languages to avoid detrimental effects of bilingualism on cognition and a higher level of competence in both
languages (as yet unspecified) to reap cognitive advantages.
On the other hand, Hakuta and Diaz (1985) and Diaz (1985), drawing on a series of empirical findings, argued that it is the relative knowledge of the two languages, rather than the absolute levels (fixed lower and higher threshold levels, whatever they may be) of first/second languages, that determines the effect of bilingualism on children’s performance on word awareness tasks. Employing causal modeling, they found that “the degree of balance (operationalized as the level of English proficiency) was a significant factor in determining performance on some of these measures for children who were overall very low in English skills” (cited in Bialystok, 1988, p. 560). Drawing on these findings, they argued that the bilingualism effect is most evident in beginning bilinguals whose language skills are less balanced (comparatively low in English skills) than on those who are more balanced (comparatively high in English skills). In other words, using Cummins’s term, only before a higher level of threshold of second-language ability (as yet unspecified) would bilingualism have a strong impact on cognitive ability.
A somewhat different version supporting the argument that the relative knowledge of the two languages decides the effect of bilingualism is Bialystok’s model, which emphasizes the varying effects of bilingualism on different levels of metalinguistic awareness. This model argues for an advantage to bilinguals, regardless of levels of bilingualism, in only cognitive control of processing. Subsequent research has revolved around the debate over the levels of bilingualism and types of tasks from which bilingual children reap metalinguistic advantages over their monolingual counterparts.
Studies Supporting Cummins’s Threshold Hypothesis
In an attempt to address the effects of varying levels of bilingualism on children’s metalinguistic development, Goncz and Kodzopeljic (1991) asked two different bilingual groups—one composed of Serbo-Croatian preschool children studying in a French immersion program and the other constituted by Serbo-Croatian preschool children learning English as a curriculum subject—as well as one monolingual group to participate in name exchange and word size judgment tests. All groups were matched on age, sex, socio-economic status (SES), and general intelligence. The results showed that the bilingual children in the French immersion program performed more than twice as successfully as the English/Serbo-Croatian children, who were no more successful than their monolingual counterparts.
Goncz and Kodzopeljic’s (1991) findings lend support to Cummins’s claim that an analytic orientation towards language is more a result of actually using two languages (as in the case of Serbo-Croatian children in the French immersion program) than of language study per se (as in the case of Serbo-Croatian children learning English as a second/foreign language). One point worth mentioning is that the research design used a between-subjects design, which might have introduced unwanted individual variation into statistical analysis. In addition, the results were based not only on the correct answer but also on the correct explanation, which involves explicit analysis of knowledge. As we will see in Bialystok’s study (1986a, 1986b, 1988), the advantage enjoyed by partially bilingual children (see more details on p. 6) is limited to cognitive control of
processing (i.e., name exchange) rather than explicit explanation of their knowledge structure. Asking children to explain why names can be exchanged is equal to testing their verbal intelligence rather than their control of cognitive processing. It is thus unsurprising that Goncz and Kodzopeljic (1991) did not find any superiority in the performance of English/Serbo-Croatian children over their monolingual counterparts.
Studies Supporting Diaz’s Level of Bilingualism Hypothesis
Studies supporting Diaz’s (1985) level of bilingualism hypothesis include Hakuta (1987), Diaz (1985), and Galambos and Hakuta (1988). In a longitudinal study of the effects of bilingualism on cognitive ability, Hakuta (1987) observed children’s language and cognitive abilities over a three-year period. The subjects were approximately 200 Puerto Rican children registered in bilingual-education programs, where both the first (Spanish) and second (English) languages were used frequently as media of instruction. At the beginning of the study, the youngest subjects were attending kindergarten and at the end of the study, the oldest children were attending sixth grade. The study, therefore, provides valuable data about the effects of bilingual education throughout the elementary-school years. The most striking finding of Hakuta’s (1987) study is that bilingualism (defined as bilinguals’ ability in the second language, controlling for relative ability in the first language) predicts considerable portions of the variance in cognitive ability for younger children in kindergarten and first grade while the effects decrease for older children in grades four through six.
In a study of kindergarten and first-grade children from the same pool of subjects, Diaz (1985) examined the effects of bilingualism separately for children of high and low levels of second-language proficiency, within kindergarten and first-grade age groups. In both kindergarten and first grade, the degree of bilingualism predicted significant portions of the cognitive variance for children of low second-language (English) proficiency level, while the effects decreased or virtually disappeared for children of high second-language proficiency of the same age. Interestingly enough, within the first grade, bilingualism predicted 34 percent of the variance in Raven’s test scores for children of low English proficiency, while it predicted only 4 percent of the Raven variance for children of high English proficiency. This finding replicates, within an age group, the attenuation found by Hakuta (1987) across grades. Both Hakuta’s (1987) and Diaz’s (1985) studies demonstrate that bilingualism had a greater role in predicting metalinguistic performance for children whose language skills were less balanced, that is, for children in the earlier stages of second-language learning.
In an attempt to further understand the role of first and second languages as bilingualism on children’s metalinguistic awareness, Galambos and Hakuta (1988) examined low-income Puerto Rican children with low English proficiency. They found that native language (Spanish) proficiency exerted influence on children’s abilities to perform metalinguistic tasks. The effect of bilingualism, operationalized as level of English proficiency, was found to vary depending on the level of proficiency of the native language (Spanish) and the difficulty of items. At high levels of Spanish, the effect of
English mostly occurred on the hardest items, whereas at low levels of native language proficiency, the effect of bilingualism was more generally evident across item types. Galambos and Hakuta’s (1988) findings not only lend support to the argument that relative level of first language/second language proficiency is the relevant factor in bilingualism’s effect on children’s metalinguistic achievements, but also point out another very important fact: That the level of difficulty
of metalinguistic tasks mediates the effect of bilingualism.
Studies Supporting Bialystok’s Various Levels of Metalinguistic Awareness
In response to Galambos and Hakuta’s (1988) call for examining the effects of bilingualism on various metalinguistic tasks, Bialystok and Ryan (1985) embarked on classifying the specific skills required for metalinguistic tasks. According to Bialystok and Ryan (1985), there are two types of skill components in the processing requirements for metalinguistic tasks—analysis of linguistic knowledge and control of linguistic processing. Analysis of knowledge refers to the mental representation of knowledge within the linguistic domain. The process of analysis is responsible for “restructuring and recoding conceptual representations organized at the level of meanings…into explicit representations of structure organized at the level of symbols” (Bialystok, 1993, p. 221). Some tasks, such as concept of word, correction problems, or definition tasks, depend primarily on the child’s knowledge of linguistic structure. The solution to these problems depends on the child’s ability to detect, extract, and articulate some structural property of
language. Control of linguistic processing, on the other hand, is the executive component responsible for directing attention to the selection and integration of information. Metalinguistic tasks, such as Piaget’s (1929) sun/moon problem, sentence segmentation, symbol substitution, and repetition of deviant sentences, depend more on the child’s control of attention, because the child must carry out a simple task by deliberately ignoring the meaning of the sentences being manipulated.
Drawing on previous findings of her own research (Bialystok, 1986), Bialystok (1988) made three hypotheses. First, fully competent bilingual children (Anglophone children educated in French at school and English at home) and partially bilingual children (Anglophone children attending French immersion program) would be more advanced than their monolingual counterparts in their level of controlled processing. Second, fully competent bilinguals would be superior to their partially bilingual and monolingual counterparts on tasks demanding high levels of analysis of knowledge. Finally, children who have only scant knowledge of a second language would not show an advantage over monolingual children when solving metalinguistic problems requiring high levels of analysis of knowledge. With these in mind, Bialystok (1988) embarked on testing these propositions. In the first part of the study, she asked three groups of Grade 1 children matched on SES and IQ—monolingual English-speaking children, partially French-English bilingual children, and fluent French-English bilingual children—to perform three metalinguistic tasks. Among the three tasks, two were related to word awareness—arbitrariness of language (i.e., the traditional sun/moon
name exchange which is high demand on control processing) and concept of word (high on analysis of knowledge). The results were in the predicted direction for the first two hypotheses. Both fully and partially bilingual children performed better than their monolingual counterparts on the sun/moon name exchange tasks demanding high levels of control of processing. In terms of task performance on defining word, which requires high levels of analysis of knowledge, the fully bilingual group, as expected, was superior to both the partially bilingual group and the monolingual group. The monolingual group, however, did not perform better than the partially bilingual group on tasks demanding high levels of analysis of knowledge as predicted, despite their highest level of English proficiency on the Peabody Picture Verbal Test.
In the second part of her 1988 study, Bialystok used a more stringent research design—within-subject group design—with another sample of Italian-English children from a working-class background. Using the children’s scores on the Italian Peabody Picture Verbal Test, Bialystok (1988) divided them into high bilingual and low bilingual groups. A series of statistical analyses confirmed the two hypotheses. The high bilingual group was superior to the low bilingual group on tests of analysis of knowledge. With regard to tasks requiring high levels of control of processing, there was no difference. Bialystok’s research not only corroborated Galambos and Hakuta’s (1988) findings that the level of difficulty of different task demands can impact bilingual children’s metalinguistic performance, but also further classified the different task demands into those tapping analysis of knowledge and cognitive control and successfully proved
differing effects of bilingualism on different metalinguistic tasks. Her research shows that it is the relative levels of first/second language proficiency that determine the effects of bilingualism on metalinguistic achievements, rather than the absolute levels of first language/second language proficiency as posited by Cummins’s threshold hypothesis (1979). Bilingual children, regardless of levels of bilingualism, enjoy an advantage in tasks that require the control of cognitive processing. The difference in levels of bilingualism plays a role only in tasks that require high levels of analysis of knowledge.
Another research reporting a bilingual advantage of word awareness with a minimal level of bilingualism was conducted by Yelland, Pollard, and Mercuri (1993). On the basis of Bialystok’s (1988) conclusion that bilingualism facilitates bilingual children’s control of cognitive processing, Yelland et al. (1993) hypothesized that even a minimal level of bilingualism would facilitate children with very limited second language proficiency to outperform their monolingual counterparts on word awareness tasks. Two age groups—preparatory grade and grade 1—matched on SES participated in this experiment. Each age group consisted of a marginal bilingual group and a monolingual subgroup. The marginal bilingual group was composed of English speakers learning Italian as a second language for a very short period of time. The two groups were required to perform the word size judgment task twice respectively—one when the marginal bilingual children had only three hours of second language instruction and the other when they had received 24 hours of classroom instruction. For the younger group (the preparatory grade), the marginal bilingual group started to
outperform their monolingual counterparts only in the 2nd test session. It is unsurprising given that the marginal bilingual group, receiving only three hours of Italian instruction before the 1st test, was not different from the monolingual group in the 1st test session.
Concerning the older group (Grade 1), the difference was obvious from the first test session, with the bilingual group outperforming their monolingual counterparts. However, during the second test session, there was not much of a difference. Drawing on these findings, Yelland et al. (1993) concluded that the metalinguistic advantages enjoyed by bilinguals in the literature extend to children who have only very limited contact with a second language. They also reported that the word awareness advantage found among the preparatory grade children in a second language program is relatively short-lived. Yelland et al.’s (1993) finding seems to corroborate Diaz’s hypothesis that bilingualism facilitates children more during their early stages of second language acquisition.
Summary and Critique
Three competing models have provided explanations for the effects of bilingualism. Cummins (1979, 1987) predicted the positive effects only for those who have reached beyond the threshold (as yet to be specified). In contrast, Diaz and Klingler (1991) argued that the bilingualism effect is most evident on beginning bilinguals and then gradually declines as the children’s second language proficiency increases. In other words, only before, not after a certain (as yet unspecified) threshold of second-language ability would bilingualism have a strong impact on cognitive ability. On the other hand,
Bialystok (1986a, 1986b, 1988) hypothesized a cognitive advantage in the control of processing for bilinguals, regardless of the level of bilingualism (operationalized as the second language level).
The Cummins’s hypothesis was formulated as a meta-analysis of seemingly contradictory findings from early studies that compared bilingual and monolingual children. However, his dual threshold hypothesis fails to account for why some of the bilinguals in transitional programs (characteristic of subtractive situations) experience some positive cognitive effects (Diaz, 1985; Hakuta, 1987; Hakuta & Diaz, 1985) or why children with limited foreign language proficiency can reap beneficial effects of bilingualism (Yelland et al., 1993). By contrast, Diaz (1985) and Bialystok’s hypotheses, on the other hand, were formulated to reflect the differential explanatory power of second-language proficiency in predicting cognitive variance within groups of different levels of bilingual proficiency. However, Diaz’s (1985) level of bilingualism hypothesis appears as ill-defined as Cummins’ (1979) threshold hypothesis due to its lack of specification of language levels. The two components Bialystok (1986a, 1986b, 1988) proposed in metalinguistic tasks, on the other hand, have been tested empirically (Ricciardelli, 1993). In terms of the explanatory power, Bialystok’s hypothesis appears superior, given its verified construct validity and considerable empirical support.
None of the research has ever found a blanket advantage for bilinguals on all tasks—it is normally confined to certain tasks, and sometimes to certain parts of some tasks. It is not uncommon for studies employing a battery of tasks to report quite different results for each. The inconsistencies are primarily due to methodological
issues. Most researchers comparing bilinguals and monolinguals did not attempt to make the two groups comparable on relevant variables, such as SES, parental education, years of schooling or any other possible confounding variables that could make the two groups different for reasons other than bilingualism such as verbal intelligence (Rosenblum & Pinker, 1983). In addition, researchers, except for Bialystok (1988) and Galambos and Hakuta (1988), did not employ a within-subject group design when comparing fully competent and partially bilingual groups. Failure to use a more sensitive within-subject design to investigate the differences in performances among bilingual children with different levels of bilingualism might introduce individual subject differences, thus masking the crucial differences between the subjects.
In addition to previously mentioned methodological issues, an equally important, if not more important issue—the duration of positive bilingual advantage enjoyed by bilinguals—has been overlooked. Most researchers examining the effects of bilingualism focused on comparing bilingual/monolingual performances on different types of word awareness tasks at a certain point in time with different samples of populations. Very few addressed this beneficial effect in a longitudinal manner. For those who did examine the duration, the effect appears to be short-lived (Diaz, 1985; Hakuta, 1987; Yelland et al., 1993). The positive bilingual effect seems to be mediated by the maturation factor. In other words, the initial advantage enjoyed by bilingual children seems to disappear when the children grow older. The interaction between bilingualism and maturation on metalinguistic awareness needs to be further examined
with other samples of population. Our understanding of the differences in word awareness among monolinguals/bilinguals is, by far, quite limited in that it only points out the differences at one point in time, not in a continuous manner.
Given Bialystok’s hypothesis that bilingual children, irrespective of their levels of bilingualism, enjoy an advantage in the control of cognitive processing, children learning a foreign language also can reap this bilingual advantage. Very few researchers attempted to extend this hypothesis to foreign language situations (but see Yelland et al., 1993), especially in Taiwan where most researchers are interested in bilingual children’s phonological awareness (Hu, 2003) and phonemic awareness (Chan, Hu, & Wan, 2005). Therefore, the purpose of this study aims to investigate the duration of this bilingual advantage and see if this advantage will be mediated by maturation effect; in other words, whether the beneficial effect is long-term or short-lived.
Research Questions
Two questions were examined in the current study: (1) Was the EFL group able to outperform the monolingual group in word segmentation (word string repetition and word count) and word size judgment during the 2nd year? (2) Did the bilingual advantage in name manipulation found in the first year last through the second year? By investigating only preliterate preschoolers, the researcher seeks to control systematically for the possible confounding effects of literacy (Nicoladis & Genesee, 1996).
METHOD
Research Sites
The research sites were two kindergartens—one monolingual and the other EFL (English as a foreign language)—located in the same district in Tainan. Both kindergartens emphasized the importance of environmental input. The classroom walls were decorated mostly with pictures and occasionally words, with the monolingual group displaying Chinese characters and the EFL group both Chinese characters and English letters. However, both the teacher’s assessment and the researcher’s vocabulary test (see below) showed that neither group could read either Chinese or English.
The medium of instruction in the monolingual group was Mandarin, whereas the medium of instruction in the EFL group was mostly Mandarin mixed with English vocabulary in content courses and English in English class. None of the children in the monolingual kindergarten received any form of formal instruction in English, either at home or school. By contrast, all of the children in the EFL kindergarten started receiving instruction in English when they entered the program. The foci of the English classes were mainly topic-based, usually taking the form of rhymes, stories, pictures, songs and games. Although the EFL group had English classes three times a week, with each lasting for 30 minutes, the actual time for foreign language acquisition might have exceeded the official amount of time. Prior to the study, the EFL group had learned English for a total of 96 hours at school. Parental report revealed few English learning activities at home except school learning activities introduced by children.
Selection of Participants
The researcher employed several measures to screen children for participating in this study. All children were tested for volubility, reading ability, and nonverbal intelligence. Given the possibility that talkative children might perform better on some word awareness tasks (Rosenblum & Pinker, 1983), the researcher asked the children to verbally describe five different stuffed animals (a lion, a cat, a puppy, a crocodile, and a bear) and audio-recorded and transcribed their description. The total number of content words (i.e., nouns, verbs, adjectives, etc.) in their oral description was interpreted as an indicator of their volubility.
To test children for reading ability, the researcher administered a battery of 30 words collected from children’s picture books in Chinese, slightly modified to render them more natural for prereaders’ emerging reading abilities. If a child could read any word accurately, he or she was eliminated from the study. The EFL children were also required to read words in their English textbooks. None of the children could read in either Chinese or English, albeit they could retell the gist of the stories based on their memory or the pictures in the storybooks. Nonverbal intelligence was evaluated by a standardized measure—Raven’s Colored Progressive matrices. Socio-economic status was computed according to the formula established in Goncz and Kodzopeljic (1991): “(parent’s educational level * 2) + (parent’s occupation * 3), where both the educational level and occupation were divided into four categories, from incomplete primary schooling to completed doctorate for the former, and unskilled laborer to professional status for the latter, with one to four points allocated to each category” (p. 143).
Word Awareness Task
In accordance with Bowey and Tunmer (1984), there are three requirements for full awareness of a word: Awareness of the word as a unit of language and as an arbitrary phonological label, as well as comprehension of the metalinguistic term word. Given the comparatively late development of defining the metalinguistic term word, only the first two types of word awareness were examined in this study. Three word awareness tasks—word segmentation, word size judgment, and name manipulation—were administered to the monolingual and bilingual groups. All demanded high control of attention (Bialystok, 1988, 1991), on which EFL children were expected to have better performance. The purpose of administering these tasks was to find out if the expected advantage still held during the second year. That is, is there a maturation effect that mediates the effect of bilingualism?
Word segmentation task. The ability to identify word
boundaries is one important aspect of word awareness. The measure used to assess children’s concept of a word unit is word segmentation. In addition to asking the children to segment anomalous speech stream into words, as was done in the first year, the researcher also asked them to count the number of words, a typical task employed to assess bilingual children’s “awareness of the segmentational process that isolates words in utterances” (Bialystok, 2001a, p. 171). The counting task is high on cognitive control of processing because the children need to ignore the odd meaning of the word strings and attend to the word boundary. This task is a modification of Bialystok’s (1986) study. In contrast to Bialystok who asked the kindergarten
children and 1st graders to count the number of words unscrambled in a sentence to produce a meaningless string, the researcher said to the preschool children a number of three-word lists comprising content words given their much younger age. The word lists (a total of 9) consisted of adjectives, nouns, and verbs (Appendix A). The number of characters in each word string was varied so as to be either congruent (e.g., ku/xiao/tian [cry/laugh/sweet]) or incongruent (e.g., meirenyu/maoyi/zhu [mermaid/sweater/pig]) with the number of words. After listening to the researcher, the children were asked to segment each word list into single words by repeating each word and pausing between words, and also to tell the number of words they had just repeated. The researcher gave the children two practice items, each with corrective feedback provided, which was considered to be sufficient to explain to them how to do the task, without actually teaching them how to do it (Tunmer, Herriman & Nesdale, 1988). One point was given to each correctly separated word stream. The maximal score is 9.
Word size judgment task. The ability to distinguish the physical
size of a word and that of the referred object is another crucial indicator of word awareness. The task that assesses this capability is word size judgment task, which is modified after those used by Templeton and Spivey (1980) and Bialystok (1986). In this task, the children were asked to distinguish big words from little ones in word pairs that were read to them (see Appendix B). The word pairs comprised two types—the bigger word referring to the bigger object (congruous pair such as changjinglu [giraffe]), and the bigger word representing the smaller referent (incongruous pair such as
maomaochong [caterpillar]). The big words were described as those that take “a long time to say” and little words as those that take “very little time to say” (e.g., mao [cat]) (Yelland et al., 1993). There were 14 pairs of words, with word size and object size varied systematically. Half of the pairs contained congruous word and object size, and the other half contained incongruous ones. Similar to the word segmentation task, the children were given two practice items with corrective feedback, followed by the actual test stimuli without feedback. One point was given to each correctly judged word pair. The maximal score is 14.
Name manipulation task. The last important aspect of word
awareness is recognition of the arbitrary nature between referent and meaning (Bialystok, 2001a). The measure to assess children’s awareness of this particular relationship is name manipulation, a modification of Ben-Zeev’s (1977) symbol substitution task. There were two parts in this task. In the first part, the researcher asked the children if, in principle, the names of two objects could be changed, either to a nonword (e.g., “Can we call a bumblebee TiouTiou?) or to a real word (e.g., “Can we call a bee GoGo [Puppy]?”) Then the researcher asked the children to justify their answers for both names. One point was given to each correctly answered question. The maximal score is 6.
In the second part, the researcher told the children to manipulate an object with two changed names (the new name being a nonword and a real word) and then to answer five questions about the following story. “TiouTiou (a bumblebee) is going out to have some fun. (Does TiouTiou fly or run?) After playing for a while, TiouTiou
feels hungry and wants to eat something. (What does TiouTiou like to eat?) While he is eating, someone is trying to take away his food.
TiouTiou is angry and tries to make some sound to scare that person away. (What sound does TiouTiou make?) But that person is not afraid of TiouTiou’s sound and is still trying to take away TiouTiou’s food. (What is TiouTiou going to do to fight for his food?) After playing, eating and fighting, TiouTiou feels tired and wants to go home. (Where does TiouTiou live?)” After the first version of the story, the researcher retold the story with a changed name GoGo (Puppy) and repeated the questions about the attributes of the “bumblebee” (e.g., Does GoGo (Puppy) fly or run? What does GoGo (Puppy) like to eat? What does GoGo (Puppy) sound like? How does GoGo (Puppy) fight for his food? Where does GoGo (Puppy) live?) One point was given to each correctly answered question. The maximal score is 6.
Procedure
The three tests began toward the end of May, approximately a year after the first year’s study, and lasted three weeks. Both groups of children were tested individually in a quiet corner of their classrooms during the formal experiments. In the first session, the children were screened for participation in the study. The tests for volubility, reading screening, and nonverbal intelligence were administered during this session. In the second session, the lexical awareness tasks—word segmentation (word string segmentation and word count), word size judgment, and name manipulation—were administered. Owing to the extremely short attention span of preschoolers, the researcher administered only one test at a time so
that the children could fully stay on the task. Each question was repeated twice and no correction was made when the word awareness test began. After the whole group completed one word awareness task, a second test began. As an incentive to motivate the children to stay on-task, each child was rewarded with a sticker after completing each task.
Data Analysis
The researcher employed a 2*2 factorial design, with language as the between-groups variable (monolingual vs. bilingual) and age as the within-groups variable (1st year vs. 2nd year). The age effect will be reported in another study. For each of the individual task, the researcher ran separate one-way ANOVAs. The results are the same as those produced by t-tests given that t-tests are a special case of the one-way ANOVA (Minium, King, & Bear, 1993, p. 378).
The study reported here is the second-year results of a longitudinal study. The results of the first year (Min, 2001) show that the EFL and monolingual groups did not differ significantly in two of the three word awareness (word segmentation and word size judgment) tasks although the EFL group outperformed the monolingual group in both tasks by a small margin. With regard to the third (name manipulation) task, the EFL group significantly outperformed the monolingual group, demonstrating its superior awareness of the word-object relationship.
RESULTS
Screening Measures
To ensure that the two groups were comparable on relevant factors such as nonverbal intelligence and volubility, the researcher ran a set of t tests to compare the two groups. Given that there was no attrition in either group or change of parental SES, no t tests were run to compare these two factors. The monolingual class still consisted of 30 children (14 boys and 16 girls, M = 5 years 2 months, SD = 3.58 months), with a range in age from 4 years 4 months to 5 years 7 months. The bilingual class also comprised 30 children (15 boys and 15 girls, M = 5 years 1 month, SD = 3.89 months), with a range in age from 4 years 5 months to 5 years 7 months. A t test of the age difference between these two groups (t = -1.168, df = 58, p > .01) shows no significant difference in age. Table 1 presents the different means of nonverbal intelligence and volubility, which indicate no
significant differences between the monolingual and bilingual groups. Table 1
Comparison of the Monolingual and Bilingual Children in Relevant Factors Groups Nonverbal Intelligence (Mean) Volubility (Mean) Monolingual 13.57 14/16 Bilingual 13.20 15/15
The mean for the monolingual children on the volubility test was 29.5; for the bilingual children, it was 30.1, t = .61, df = 58, p > .05. For all statistical tests, the decision criterion was p < .05.
Word Segmentation Task
There were two tests for word segmentation—word string repetition and word count. The results show that there was no significant difference (Table 2, F = 1.289, p > .05) in the repetition task between the monolingual and bilingual groups. In fact, both groups’ performance reached ceiling (the full score was 9, and the mean score for monolinguals and bilinguals was 8.8 and 8.9, respectively). This is not surprising, given the almost ceiling performance (M = 8.7 for monolinguals, and 8.73 for bilinguals) of both groups during the first year (Min, 2001). This lack of difference in repeating speech stream indicates that both groups were able to capture almost 100% of the words in the speech string without difficulty during the second year. The comparable performance also demonstrates their equal ability in receptive vocabulary.
Table 2
Comparison of Repeating Word Strings Between the Monolingual and Bilingual Children
Sum of Squares df Mean Square F Sig Between Groups .600 1 .600 1.289 .261 Within Groups 27.000 58 5.718 Total 27.600 59
With regard to the second task, word count, a significant difference (Table 3, F = 7.582, p < .05) was found between the two groups, signaling that the bilingual group was more advanced in awareness of “word” as a unit. This result also shows that the ability to capture and repeat aberrant aural input differs from that of counting words. Repeating anomalous word strings is much easier than counting the number of words in them, as revealed in the marked difference in the mean scores between word repetition and word counts for both groups (for the monolingual group: 8.8 for word repetition, and 4.46 for word counts; for the bilingual group: 8.9 for word repetition, and 6.17 for word counts; the full score for both tasks was 9).
Table 3
Comparison of Correct Word Counts Between Monolingual and Bilingual Children
Sum of Squares df Mean Square F Sig Between Groups 43.350 1 43.350 7.582 .008* Within Groups 331.633 58 5.718 Total 374.983 59 Note.* p < .05
Word Size Judgment Task
Another one-way ANOVA was run to compare the performances on word size judgment between the monolingual and bilingual groups. Although the result did not show a significant
difference between the two groups in overall word size judgment (F = 3.776, p > .05), a significant difference was found in the incongruous pairs (F = 4.423, p < .05). This suggests that the bilingual group was superior to their monolingual counterparts in distinguishing word size from object size in more challenging contexts in which word size contrasts with object size (i.e., incongruous pairs).
Table 4
Comparison of Word Size Judgment Between Monolingual and Bilingual Children
Sum of Squares df Mean Square F Sig Overall Performance Between Groups 30.817 1 30.817 3.776 .057 Within Groups 473.367 58 8.161 Total 504.183 59 Incongruous Pair Between Groups 13.067 1 13.067 4.432 .040* Within Groups 171.333 58 2.954 Total 184.400 59 Congruous Pair Between Groups 1.350 1 1.350 .682 .412 Within Groups 114.833 58 1.980 Total 116.183 59 Note. * p < .05
In order to assure that both groups’ performance on the word size judgment task was not merely the result of chance, the researcher ran an additional t test to examine the chance factor. The results show that the percentage of correct word size judgments on the congruous items was above chance (50%) for both the monolinguals (t (29) = 9.727, p < .05) and the marginal bilinguals (t (29) = 8.998, p < .05). In a similar vein, the percentage of correct word size judgments on the incongruous items was also above chance level for both the monolinguals (t (29) = 5.132, p < .05) and the bilinguals (t (29) = 8.998, p < .05), albeit the monolingual group’s performance (M = 4.93) on the incongruous items was significantly inferior to that of the bilinguals (M = 5.87).
Name Manipulation Task
The researcher compared the number of assents that children gave to the question of whether the name of a bumblebee could be changed into a nonword (TiouTiou) or a real word (Puppy). Almost all the children in both groups agreed that the bumblebee could be called
TiouTiou, with the exception of one child in the monolingual group. All the children were able to manipulate the objects with changed nonword labels. Almost all replied that the bumblebee liked to eat honey, flew in the air, buzzed and stung people, and lived in the tree but few of them knew the word “beehive.” By contrast, fewer children consented to changing to the real word Puppy. Eight children in the monolingual group refused to change the name, whereas only one in the bilingual group found such a name change inappropriate. One-way ANOVA indicates a significant difference in these drastically different replies (Table 5, F = 6.932, p < .05). This marked
difference indicates a lack of mature awareness of the arbitrary and thus separable relationship between word and its referent among some of the monolingual children, whereas most of the bilingual children were ready to acknowledge such a relationship.
Table 5
Comparison of Agreement on Naming the Bumblebee “Puppy”
Between Monolingual and Bilingual Children
Sum of Squares df Mean Square F Sig Between Groups .817 1 .817 6.932 .011* Within Groups 6.833 58 .118 Total 7.650 59 Note. * p < .05
In the second name manipulation task, the researcher replaced the name TiouTiou with Puppy and asked the children to retell the story about the bumblebee named Puppy. Despite some initial resistance from eight children in the monolingual group and one in the bilingual group, the researcher was able to coax most of them to play along. Three children from the monolingual group, however, vehemently opposed this name change, claiming that a bumblebee could not be called Puppy. Table 6 demonstrates a significant difference between the two groups in the number of correct answers about the attributes of the bumblebee with a changed label Puppy. The drastic difference again implies a more sophisticated awareness of the arbitrary relationship between words and objects among the bilingual group.
Table 6
Comparison of Correct Answers
About the Attributes of the Bumblebee “Puppy”
Between Monolingual and Bilingual Children
Sum of Squares df Mean Square F Sig Between Groups 18.150 1 18.150 5.060 .028* Within Groups 208.033 58 3.587 Total 226.183 59 Note. * p < .05
DISCUSSION
The purpose of this study was to investigate if the bilingual advantage found in the first year’s study can last to the second year without being set off by the age effect. Specifically, it intended to examine whether the significant difference noted in the first year (i.e., name manipulation) between the two groups could still be found in the same word awareness tasks during the second year. The results show that during the second year, the two groups still differed significantly in name manipulation as in the first year (Min, 2001). Again like the first year, there was no significant difference between the two groups on overall word size judgment. However, the bilingual group significantly outperformed their monolingual counterparts on word count (a new test in word segmentation) and incongruous word size judgment tasks. In other words, in addition to superior awareness
of the arbitrary relationship between a word and its referent in name manipulation, which was found in the first year, there were two additional significant differences emerging during the second year—word count and name exchange.
All these tasks demanded high control of attention (Bialystok, 1988, 1993), on which EFL children were expected to have better performance. These results lend support to Bialystok’s findings and hypothesis that bilingual children, regardless of their levels of bilingualism (in this case, 96 hours of English language instruction) enjoyed an advantage in tasks requiring cognitive control of processing. The current findings do not support Cummins’s hypothesis that bilingual children need to reach a higher level of competence in both languages to reap cognitive advantages.
Word Segmentation
In the repeating task, both groups, as expected, were able to successfully repeat after the researcher commented on the ease of such repetition of meaningless word strings. One child said: “Zhe hao jiandan.” (This is very easy.) In fact, all children repeated accurately after the researcher. In addition to commenting on the simplicity of the first task, some children also expressed their feelings about the oddness of such word strings by employing a questioning tone after each repetition. However, none attempted to rationalize or contextualize such oddity by offering explanations. This contrasts sharply with the children’s tendency to rationalize aberrant speech stream observed during the first year (Min, 2001). The children’s apparent questioning intonation after repeating anomalous word
strings and the lack of an obvious attempt at contextualizing them appeared to convey a tacit acceptance of arranging words in a meaningless way. This acceptance can be construed an emerging awareness of treating language as an object without being bound to meaning. The recognition that language can be sometimes used for purposes other than meaningful communication helps the children focus on more abstract linguistic form and thus develop metalinguistic awareness.
In the second task requiring children to count words in a speech stream, most monolingual children found strings made of nouns and verbs easier to count than those composed of adjectives. Approximately 26% (n = 8) of the monolingual children employed a morphosyllabic strategy when they tried to segment word strings made of adjectives (holding up a finger when pronouncing a character), while 30% (n = 9) of them guessed wildly (without holding up fingers as they did in counting noun strings). The result was an accurate morpheme count, a miscount due to too many morphemes involved, or a wild guess, all of which were incorrect answers. By contrast, most of the children in the bilingual group were able to successfully count the number of words in each word string, although 17% (n = 5) were confused about words with morphemes and employed a morphosyllabic strategy in counting word strings made of adjectives (by holding up a finger for each character).
Why are word strings composed of adjectives more difficult to separate? It might be due to their lack of concrete identity, in other words, their abstractness. Nouns and actions are more salient to children because the former usually refer to distinct entities and the
latter can be easily distinguished from one another. In contrast, adjectives are not as salient and easily identifiable as objects and actions because they are usually abstract concepts. When three abstract notions are put together in a word string, it requires a high cognitive control of processing on the children’s part to first visualize and then to disentangle them. The results show that the bilingual group was better able to complete this demanding task than their monolingual counterparts. This is a strong piece of evidence that these bilingual children are more advanced than their monolingual counterparts in awareness of a word as a unit, especially involving adjectives.
The comparable performance of the monolingual children to that of the bilingual children on counting nouns in word strings but inferior performance on segmenting adjectives also points out the influence of the level of task difficulty on the children’s performance. As discussed in previous paragraphs, Galambos and Hakuta (1988) emphasized the impact of different levels of task difficulty on performance. It appears that the monolingual children’s performance was somewhat adversely influenced by the more difficult task. Although they were able to successfully count nouns in word strings, they were unable to do so in counting adjectives. This lends support to Bialystok’s (1988) hypothesis that bilingualism enhances bilingual children’s control of cognitive processing, thus helping bilingual children ignore the abstract nature of adjectives while segmenting them into individual words.
Word Size Judgment
There were two parts of word size judgment—congruous pairs (a big word denoting a large-size object) and incongruous pairs (a big word denoting a small-size object). Three separate ANOVAs were run to compare the differences between the two groups. Similar to Bialystok’s (1986a, 1986b) findings, the results show that the two groups did not differ significantly in their overall performance and in congruous pairs. However, there was a significant difference in incongruous pairs between the two groups. One might wonder why the monolingual group performed equally well as their bilingual counterparts on congruous pairs but significantly poorer on incongruous pairs.
The key lies in the interaction between difficulty of tasks and ability of cognitive control (Akbulut, 2007; Bialystok, 1986a, 1986b). By nature, congruous pairs are easier than incongruous ones because words referring to a bigger-size object also represent a bigger word size. The object size had a less inhibiting effect on the children, who did not need to struggle hard to distinguish the object size from the word size. In contrast, incongruous pairs are more difficult given the contrasting sizes between word and object. While making a judgment on an incongruous pair, the children needed to direct their attention to the word size only and ignore the competing influence of object size, or their judgment would be adversely impaired.
The fact that the monolingual children’s performance on incongruous pairs was above chance level on both congruous and incongruous pairs suggests that they already had some form of word awareness. However, this fledgling word awareness fluctuated with
the cognitive demands of tasks. When the word judgment task required less cognitive control, as in the judgment of congruous pairs, the monolingual children performed as well as their bilingual counterparts (Bialystok, 1988). When the word judgment tasks required a high-level control of cognitive processing, as in the case of judging incongruous pairs, they might encounter problems. The significant difference between monolingual and bilingual children on incongruous pairs suggests that the monolingual children were less developed in terms of controlling their attention to the required task. In other words, they were less able to direct their attention as the task difficulty of cognitive demands increased, although they may have had some rudimentary form of word awareness.
It should be mentioned that the researcher did not make any change to the syllabic structure of the test items in both congruous and incongruous pairs during the second year. Most of the test items were still composed of bisyllabic and trisyllabic words despite the less distinguishable contrast in time duration between the bi- and tri-syllabic test pairs. Although facing a more difficult word size judgment task than those in other studies (e.g., Yelland et al., 1993), the monolingual children performed better than the children of comparative age who still performed at around chance on the incongruous items (Yelland et al., 1993). Unfortunately, due to a lack of specific information about SES, volubility, and non-verbal intelligence of the monolingual children in Yelland et al.’s study, the researcher cannot reach a definitive conclusion from this difference.
Name Manipulation
Both groups exhibited great flexibility concerning transforming the bumblebee’s name into TiouTiou but displayed markedly different attitudes towards renaming the bumblebee Puppy. This finding is partially in concert with Rosenblum and Pinker’s study (1983) in that bilingual children in both studies were more context-oriented in their justification of a name change. Bilingual children in this study tended to accept the name change as a result of playing games (“zhe zhishi zai wan yiuxi [this is just a game]), while monolingual children tended to justify their explanation by mentioning the bumblebee’s physical attributes (“yinwei ta shi xiaomifeng,” [because it’s a bumblebee]). With regard to the name change to Puppy, the bilingual group was more willing to separate the name from the object (F = 6.932, p < .05).
Similar to the monolingual children reported in precious studies (Bialystok, 1988; Ricciardeli, 1992), the monolingual children in this study were “more wedded to the familiar meanings of words than were their bilingual peers” (Bialystok, 2001a, p. 171). When asked to justify why Puppy was an inappropriate name for the bumblebee, most referred to the object’s attributes (“…yinwei ta you chibang, guoguo meiyou” […because it has wings and puppies don’t have wings]). Others referred to the uniqueness of the name (“ruguo ta ye jiao guo de hua, na bieren jiu fen bu qingchu ta daodi shi guo haishi mifeng” [if its name is Puppy, people don’t know whether it’s a puppy or a bumblebee]). For those who consented to a name change to the nonword TiouTiou but were opposed to the change to a real word Puppy, name was part of the attributes of an object. It is as
though the meaning is inherent in the word, an inseparable part of it. The nonword name TiouTiou was accepted because no particular object had such a name attribute. Therefore, it was free to be attached to any object, as long as the children could tolerate the weirdness and silliness of such a name. In contrast, the real word
Puppy already had an associated object, and thus could not be used with another object. The justification of the refusal that people would not know whether it is a puppy or a bumblebee implies the perceived inseparable relationship between word and its referent. To one fourth (n = 8) of the monolingual group, renaming the bumblebee as Puppy violated the relationship between names and their accompanying attributes, which was deemed to be indivisible. This explanation, in fact, was a display of a lack of awareness of the divisible relationship between a word and its referent.
In contrast, the bilingual group in this study, like those reported in previous studies (Bialystok, 1988; Ricciardeli, 1992), was more willing to “accept that the meaning of word was more convention than necessity; more agreement than truth, and felt free to break the agreement if they so chose” (Bialystok, 2001a, p. 171). Such willingness is a precursor to an awareness of the separable relationship between words and objects, and a manifestation of the beneficial effect of exposure to a second foreign language.
The results of this study have two important bearings. First, the bilingual advantage found in name manipulation during the first year did not attenuate with time. In other words, the bilingual children still significantly outperformed their monolingual counterparts on this task. Second, additional beneficial effects emerged during the second year.
Apart from their superior recognition of the divisible relationship between words and their referents, the bilingual children were more advanced than the monolingual children in counting words and in real word name exchange (Puppy for TiouTiou). These significant differences demonstrate that the bilingual children were more sophisticated in their comprehension of word as a unit and of the arbitrary relationship between objects and their phonological realizations (Bialystok, 2001a, 2001b).
Unlike Yelland et al.’s (1993) finding that the bilingual benefits “had eroded by the time the children were roughly 6 years old” (approximately one year after the positive effects of bilingualism had been reported), this study found that the beneficial bilingual effects extended to the second year. Before attempting any interpretation, there are two important differences between these two studies that need to be addressed. First, in Yelland et al.’s study, the monolingual and bilingual Grade 1 children received literacy instruction, but in this study the monolingual and bilingual children did not receive any formal literacy instruction, albeit they were constantly read to by their respective homeroom teachers. It is perhaps the effect of literacy that offset the bilingualism effect in Yelland et al.’s study.
Second, the language learning environments in both studies were quite different. The bilingual children in Yelland et al.’s study attended a mainstream traditional program in which the majority language children were instructed in their first language (English) and exposed to a foreign language (Italian) as a curriculum subject. No description of the amount of Italian the children used to converse with one another was provided. It is the researcher’s assumption that the
bilingual children did not use much Italian outside of classroom contexts given its limited language currency in eastern Melbourne in Australia.
By contrast, the bilingual children in the current study attended a program that could be deemed as a combination of mainstream traditional program and immersion program. This type of program might increase the bilingual children’s second language exposure and thus perhaps modify the process of metalinguistic development in the preschool period in comparison to the development of these processes in the monolingual children, despite their use of Mandarin in conversing with each other. This is probably why the bilingual children in the present study could still outperform their monolingual counterparts on word awareness tasks during the second year. On the other hand, without a facilitative foreign language-learning environment, it is less surprising that the literate monolingual children caught up with their bilingual counterparts a year later in Yelland et al.’s study.
CONCLUSION
The findings lend support to Cummins’s (1978) argument that the metalinguistic benefits of learning a second language will accumulate if bilingual children continue developing both their languages. Although this study found that the beneficial bilingual effect on word awareness could extend to the second year, the findings can only be applied to preschool children who do not receive
any literacy education and who attend a bilingual program that approximates an immersion program. The findings do not predict a positive bilingual effect on other aspects of metalinguistic awareness or during the third year, when formal literacy education begins.
As is often the case with research, this study raises more questions than it was designed to address. It only focused on investigating the bilingual effect on the cognitive control of processing ability of the bilingual and monolingual groups in word awareness tasks; it did not examine its impact on the two groups’ capability in analysis of knowledge. Bialystok (1988) hypothesized that fully bilingual children would be superior to partially bilingual and monolingual groups in terms of their analyzed knowledge. Future research needs to use a within-subject design to divide a bilingual group into more and less advanced bilingual subgroups and compare their performance on this component of word awareness with a monolingual group. Another question worth probing is the bilingual children’s relative metalinguistic skills regarding the two languages. According to Cromdal (1999), young bilingual children demonstrate more sophistication in performing metalinguistic tasks in their weaker language than in their stronger language. This argument is based on the premise that bilingual children constantly engage themselves in attending to the formal structure of the second language, thus frequently refining their linguistic knowledge. Further research needs to design age-appropriate and level-appropriate word awareness tasks to examine the validity of this hypothesis.
REFERENCES
Akbulut, Y. (2007). Bilingual acquisition and cognitive development in early childhood: Challenges to the research paradigm.
Elementary Education Online, 6(3), 422-429.
Ben-Zeev, S. (1972). The influence of bilingualism on cognitive
development and cognitive strategy. (ERIC Document Reproduction Service No. ED104535).
Ben-Zeev, S. (1977a). The influence of bilingualism on cognitive strategy and cognitive development. Child Development, 48, 1009-1018.
Ben-Zeev, S. (1977b). The effects of bilingualism in children from Spanish-English low economic neighborhoods on cognitive development and cognitive strategy. Working Papers on
Bilingualism, 14, 83-122.
Bialystok, E. (1986a). Children’s concept of word. Journal of
Psycholinguistic Research, 15(1), 13-32.
Bialystok, E. (1986b). Factors in the growth of linguistic awareness.
Child Development, 57, 498-510.
Bialystok, E. (1988). Levels of bilingualism and levels of linguistic awareness. Developmental Psychology, 24, 560-567.
Bialystok, E. (1991). Metalinguistic dimensions of bilingual language proficiency. In E. Bialystok (Ed.), Language processing in
bilingual children (pp. 113-140). London: Cambridge University Press.
Bialystok, E. (1993). Metalinguistic awareness: The development of children’s representations of language. In C. Praff & A. Garton
