Bilingualism and cognitive development: three perspectives

7 Bilingualism and cognitive development: three perspectives
Kenji Hakuta, Bernardo M.Ferdman, and Rafael M.Diaz
The problem of researching the relation between bilingualism and cognitive development at once raises two thorny definitional issues. What do we mean by bilingualism, and what is it that develops in cognitive development? Much of the confusion in this area can be attributed to the lack of theoretical specificity in defining the intersection point of these component concepts. Our primary emphasis in this chapter is on the definition of bilingualism, with a secondary emphasis on cognitive development. The reason for the asymmetry is to be consistent with the traditional assumption that bilingualism is the independent treatment variable and Cognitive growth is the dependent outcome variable, even though, as we shall see, very few studies actually address the cause-effect issue. The major goal of this chapter is to demonstrate the great range of social and theoretical contexts in which the question has historically been asked and to argue for the importance of integrating the many disciplinary levels and perspectives that bear on the problem.
Defining the component concepts
The concept of bilingualism has been used in various ways by scholars and lay persons alike. It has been viewed as an individual-level mental concept - a characteristic of individuals who possess or who use two linguistic systems. It has also been viewed as a social psychological concept, still a characteristic of individuals, but of individuals who organize the social world in terms of the different groups and social situations associated with the two languages in which they interact. Bilingualism has also been used as a societal construct to describe the interactions between social groups and societal institutions, as well as among groups, in which the group and institutional boundaries correspond to linguistic boundaries. These different starting points for the definition of bilingualism have resulted in discrepancies in the kinds of statements that have been made about bilingualism and its relation with cognitive development.
The preparation of this chapter was supported in part by National Institute of Education Contract 400-85-1010for the Center for Language, Education and Research.
284

Bilingualism and cognitive development

285

When bilingualism is defined in the first way, as tl characteristic of an individual
who possesses two linguistic systems - we call it cognitive bilingualism - one
tends toward statements about the packaging problem of fitting two linguistic SySterns in the mind of an individual. It is a cognitive puzzle on the relation between language and thought and how these systems are represented neurologically and conceptually. Variables of obvious importance in cognitive bilingualism are the extent to which the individual has mastery of the two languages and the cognitive functions in which the languages are engaged.
Bilingualism defined in the second way, as a characteristic of the social condi-
tion and affect of the individual - we call it social psychological bilingualism -
tends toward social psychological accounts of the packaging of value systems within an individual. These emphasize not so much the linguistic aspects of bilinguals as the social correlates of the two languages. In this sense of the definition, the grammatical qualities of languages hardly matter. What really matters is the symbolism about group affiliation that the languages convey to the individual.
Bilingualism defined in the third way, as a characteristic of a societal unit - we
call it societal bilingualism - is concerned with between-group interactions in
which the two languages serve as a symbol over which interaction occurs. This perspective is not so concerned with individual differences within groups. As in the social psychological view of bilingualism, the extent of the vitality of the two
languages - vitality in the sense of the extent to which the grammar and form of the languages are maintained - is not so important in this view, though it can be made to be important depending on social conditions. What matters in this per-
spective is that language in some way signals membership in a group and serves to maintain the group’s cohesiveness and identity.
At the same time that there have been different levels of conceptualization of bilingualism, different theories of cognitive development have preoccupied psychologists of different generations. The earliest systematic attempts to document the relation were made at the beginning of the twentieth century. At that time, the primary definition of what we now call cognitive development was a psychometric one, based on the differential performance of individuals within a defined population on IQ tests. Subsequently, learning theory, skill theory, Piagetian operational thought, Chomskyan rationalism, and Vygotsky’s views of mind and society offered additional conceptions of what develops i n cognitive development.
Although a review of the various theories of cognitive development is far beyond the scope of this chapter, it would be important to consider the dimensions of theories that would or would not predict effects of bilingualism on cognitive development. One might think of bilingualism as an environmental “treatment,” to be compared with the alternative treatment of monolingualism.
As a first approximation toward appreciating the range of cognitive theories available, one can begin with commonly used typologies, particularly as relevant to bilingualism. These include nativism versus empiricism, modularity versus

2f6 K‘. HAKUTA,B . M . FERDMANA, N D R . M . DIAZ
commonality of functions, and context and cultural sensitivity versus independence.
With regard to the nativistic-empiricist dimension, any theory of cognitive development that subscribes to primarily innate factors, with respect to both the qualitativeaspects of cognition and differences among individuals, would not predict bilingualism to have any effect on the course of cognitive growth. This would include a Chomskyan orientation that attributes the characteristicsof our linguistic and other cognitive knowledge to our genetic makeup. It would also include a hereditarian interpretation of individual differences in intelligence, such as that espoused by Jensen ( 1980). In contrast, theories that emphasize the role of learning and the environment would easily accommodate influences of bilingualism on development. These would include traditional learning theory and skill theory, as well as Piagetian constructivism.
The second dimension of cognitive theories - modularity versus commonality of structures - will predict, given some effect of the bilingual treatment on cognitive development, how it would generalize to other domains of cognitive functioning. For example. Chomsky and Fodor’s extreme modular approach (see Piatelli-Palmarini, 1980). in which cognitive functions including language are considered to be analogous to structurally autonomous organs of the mind, would find minimal compatibility with broad-sweeping effects of bilingualism. The effects would be confined to the specific aspects of cognitive functioning that are influenced by the bilingual environment. For example, if bilingualism were to be defined strictly as a linguistic treatment rather than a social or societal one, the effects would be confined to linguistic aspects of cognitive functioning. In contrast. learning theory as well as theories of general intelligence and Piagetian operational theory would expect generalized effects since all cognitive functioning share a common source and are interrelated. However, it should be noted that Piagetian theory, though a theory of general intelligence, is characterized by its ascription of a marginal role for language in structuring intelligence.
The third dimension of cognitive theories, the cultural or context seslsitivity of theories, holds the strongest promise for relating cognitive development with the social psychological and societal levels of bilingualism. The theory best noted for its emphasis on culture is Vygotsky’s (1962), in which specific cognitive functions might exist in rudimentary form as part of the child’s genetic endowment, but the majority of the variance in cognitive growth can be explained by the ways in which society amplifies and interrelates these capacities. In contrast, both Chomskyan and Piagetian views on the role of culture are limited.
In this chapter, we make two general points centering on the definitionalconsiderations of bilingualism described above. First, we point to the importance of drawing clear distinctions among the definitions of bilingualism. Failure to do so can lead to misunderstandings about the role of bilingualism in cognitive development. Second, even though these various perspectives can and should be distin-

Bilingualism and cognitive development

287

guished, attention should also be paid to the interactionsof variables across levels. Indeed, the question of bilingual cognitive development highlights the importance of maintaining multiple perspectives and cutting across levels of analysis in social science.
We make these points using the following structure. The first section takes a historical perspective in examining changes in the way bilingualism has been thought to influence intelligence in children. The section illustrates the importance
of maintaining clear distinctions among definitions of bilingualism, while at the same time pointing to the importance of the historical context of research. Then, we follow with a discussion of bilingualism and cognitive development as seen
from each of the three levels discussed above - cognitive, social psychological,
and societal. Obviously, the cognitive perspective has the most to say with regard to cognitive development, but the latter perspectives are important to the extent that social psychological and societal factors influence the degree of bilingualism that might be attained by the population of interest. In the concluding section, we trace the implications of this multilevel analysis of the problem toward a greater understanding of language, mind, and society, drawing from our own research efforts.

Some history
If one were to look at the literature on bilingualism and intelligence over its long history, it would at first seem that the early literature showed that bilingualism had negative consequences, whereas the more recent literature, improving on the earlier methodologies, showed the opposite, that bilingualism could have a positive influenceon cognitive development.Consider the contrast to be found in following two accounts of the relation between bilingualism and intelligence. Conclusions from the early literature can be summarized by the following statement that a p peared in George Thompson’s (1952) American textbook on child psychology: There can be no doubt that the child reared in a bilingual environment is handicapped in his language growth. One can debate the issue as to whether speech facility in two languages is worth the consequent retardation in the common language of the realm. (p. 367)
A rather brighter portrait is drawn by Elizabeth Peal and Wallace Lambert (1962) in reporting a study of bilingual children in Montreal. They describe their typical subject as
a youngster whose wider experiences in two cultures have given him advantages which a monolingual does not enjoy. Intellectually his experience with two language systems Seems to have left him with a mental flexibility, a superiority in concept formation, a more diver-
sified set of mental abilities . . . . In contrast, the monolingual appears to have a m m un-
itary structure of intelligence which he must use for all types of intellectual tasks,(p. 20)
These statements and their inherent contradictionscan be interpreted as a dramatic
example of misunderstandings that resulted from failure to distinguish between different levels of definition of bilingualism.

288 ”K. HAKUTA.B. M. FERDMANA,N D R .M.DIAZ
Clearly, if the goal of a study were to establish whether the extent of bilingualism in children had an effect on individual-level cognitive development. one should define bilingualism in terms of their abilities in the two languages. What one should nor do is to use a societal definition of bilingualism. Yet the earlier
literature primarily used a societal definition - bilinguals consisted of newly ar-
rived immigrants to the United States - whereas the more recent literature has tended to use a cognitive definition. In part, this discrepancy in definitions and
findings can be attributed to improvements in methodological controls. For example, the more recent studies attempt to control for the socioeconomic status
(SES)of the comparison groups, whereas the older studies did not. However, a historical perspective enables us to appreciate why the earlier literature used the
societal definition and essentially ignored what are now considered obvious confounds, such as SES.
In order to comprehend the early literature and what the debate was all about, one must view them against the backdrop of the concerns of Americans at the turn of the century (see Gould 1981; Hakuta 1986). At that time, there raged a social debate over the quality of the new immigrant groups from southern and eastern Europe, a fear that was expressed forcefully by Francis Walker, president of MIT and a prominent spokesperson for immigration restriction:
These immigrants are beaten men from beaten races, representing the worst failures in the struggle for existence. Europe is allowing its slums and its most stagnant reservoirs of degraded peasantry to be drained off upon our soil. (Quoted in Ayres, 1909, p. 103)
The various measures of intelligence, particularly in the tradition of Goddard’s translation of Binet’s IQ test, came to play a major role in this debate, for the immigrants’ performance on these tests seemed to confirm the worst fears of restrict ion ist s I i ke Walker.
In explaining the poor performance of the new immigrantson intelligencetests, the battle line was drawn between those who believed in genetic versus those who believed in experiential explanations. Researchers in those days - including luminaries in the field such as Lewis Terman, Florence Goodenough, and George Stoddard - debated whether bilingualism was or was not a handicap in the measurement of intelligence.
The hereditarians, who believed that IQ test performance was attributable largely to genetic factors, accounted for the poor test performance of the new im-
migrants - those primarily from southern and eastern Europe - in terms of selec-
tive migration. The data were considered to support the general fear about the quality of the new immigrants. The strongest data in support of the hereditarian position were the results of the testing of U.S. Army recruits in World War I, conducted by Robert Yerkes and synthesized and popularized by Carl C. Brigham (1922).The most compelling bit of evidence, in the eyes of hereditarians, was the decreasing intelligence test scores as a function of recency of immigration. Brigham’s explanation was as follows:

Bilingualism and cognitive development

289

Migrations of the Alpine and Mediterranean races have increased to such an extent in the last thirty or forty years that this blood now constitutes 70% or 75% of the total immigration. The representatives of the Alpine and Mediterranean races in our immigration are intellectually inferior to the representatives of the Nordic race which formerly made up about 50% of our immigration. (p. 197)
The alternative explanation, of course, was that those who had immigrated most recently had learned less English and that inadequate proficiency in English resulted in poor test performance. This possibility of a language handicap in test taking was recognized by proponents of the hereditarian position, such as Lewis Tennan (1918). He and his students began a full-scale assault of the possibility that the bilinguals might be taking the tests under a language handicap and attempted to show that the differences existed even despite it (Young, 1922). Such
heroics notwithstanding, however, it became clear that the recent immigrants -
the bilinguals - were operating under a handicap. For example, Terman’s own student Darsie (1926) showed that bilinguals performed particularly poorly on the subtests of the Binet scale that required language.
Despite evidence of this sort, the hereditarians did not change their position on the genetic quality of the new immigrants. Florence Goodenough (1926). for example, turned the argument around and wrote that “those nationality groups whose average intellectual ability is inferior do not readily learn the new language” (p. 393).
In contrast to the hereditarians, psychologistswho emphasized the environmental factors associated with intelligence test scores, spearheaded by George Stoddard and Beth Wellman of the Iowa Child Welfare Research Station, were trying to explain the poor performance of immigrants using experiential factors (Stoddard & Wellman, 1934). Rather than question the validity of the IQ tests for this
particular population, they arrived at the conclusion that bilingualism - an experiential factor - must cause some kind of mental confusion, resulting in the poor
development of verbal skills. Madorah Smith, who received her doctorate at Iowa, figures prominently in this
history. For her dissertation, she had pioneered a method of analyzing free speech utterances of young monolingual children to obtain quantitative indices of language development. Later, she moved to Hawaii, where she began applying her method to the Speech of bilingual children from a wide variety of language backgrounds (Smith, 1939). A comparison of these statistics with her Iowa samples showed that bilinguals were inferior to the monolinguals, leading her to the conclusion that “an important factor in the retardation in speech found in the preschool population is the attempt to make use of two languages” (p. 253). (There are many alternative explanations of her data, a discussion of which can be found
in Hakuta, 1986.) The twists and turns of this research area can be recapitulated as follows. The
backdrop of the initial research was concern with the new immigrants, who per-

2%) K‘: HAKUTA,B. M.FERDMANA,ND R. M.DIAZ
formed poorly on tests of intelligence. The hereditarians argued that this poor performance reflected inferior genetic stock and attempted to argue against a language handicap in test taking. The evidence mounted, however, that bilinguals were operating under a handicap. The hereditarians then interpreted this handicap to be the result of innately inferior intelligence. In contrast, the environmentalists took the language handicap of bilinguals to be the result of experience, the most salient experience to them being exposure to two languages.
What is remarkable about this debate is that the language handicap of bilin-
gualism, initially construed as a test-taking factor associated with a group trait namely, foreignness and recency of immigration - soon became an alleged char-
acteristic of a supposed mental state - in our terminology, cognitive bilingualism. How were these early studies of bilingualism and intelligence conducted?They
were primarily comparisons of two groups of students, one labeled “bilingual” and the other “monolingual,” on the various tests of intelligence (including the Stanford-Binet) that were becoming increasingly popular in those days. And how was bilingualism defined? Societally. For example, studies were conducted in which children were classified as bilingual if they had a foreign last name. What was relevant for these researchers was that bilinguals were from certain ethnic backgrounds and were recent immigrants to the United States. We do not know whether the bilinguals in these studies were actually cognitively bilingual or only societally bilingual. It is quite possible that children participating in some of these studies actually were proficient only in their native, non-English language. What these studies suggest to us is that societal bilingualism, being a label in this historical context for individuals who are low on the societal totem pole, can be detrimental to performance on tests of intelligence that are used as the basis for predicting success in the educational system. What they do not suggest is that cognitive bilingualism could be detrimental to the mental development of children, since the extent to which they were cognitively bilingual is uncertain.
Indeed. as we argue in the following section, if we adopt a cognitive definition of bilingualism, as recent studies of bilingualism and cognitive development have done, there emerges a relatively consistent picture of a positive relation. In these studies where bilingualism is defined cognitively rather than societally, the criterion has often been to include only those children who are equally proficient in the two languages.
In general, this shift in definition of bilingualism from a societal to a cognitive one has gone hand in hand with a shift in the type of subject population studied. Earlier studies tended to look at immigrants and minorities in the process of language shift from their native language to English. The more recent studies, though not all, have tended to look at subjects who live in societal circumstances where equal proficiency in two languages is possible and advantageous, such as in Canada, and who tend to come from middle-class populations. Thus, in order to appreciate the full range of studies conducted on the topic of bilingualism and cog-

Bilingualism and cognitive development

29 1

nitive development, it will become necessary to delve into the societal correlates of different types of bilingualism. First, however, we turn to a fuller consideration of the cognitive perspective.

Cognitive-level bilingualism
In this section, we review two types of studies conducted strictly at the cognitive level of bilingualism, where subjects are defined in terms of their relative abilities in the two languages rather than on a social or societal basis. The first type of study looks at cognitive performance in balanced bilingual children; the second type relates children’s degree of bilingualism to cognitive ability. The section concludes by documenting the present search for a model at the cognitive level that explains how bilingualism might affect the development of children’s intelligence.
The concept of the “balanced” bilingual child was conceived by Peal and Lambert (1962) in an attempt to distinguish “pseudobilinguals” from truly bilingual children. In our terminology, they shifted the definition of bilingualism from a societal to a cognitive one. Peal and Lambert were responding to the long history of bilingual research, just described, that failed (from the cognitive perspective) to take into account the actual language proficiency of bilingual samples. In their famous monograph, the investigators argued that, in order to understand the effects of bilingualism on children’s intelligence, the first thing that is needed is truly bilingual subjects or, in their new term, a sample of “balanced” bilingual children. Furthermore, they argued that previous negative findings could be attributed to careless sampling procedures, under which subjects’ bilingual proficiency was questionable. Several formal definitions of balanced bilingualism have been formulated through the years, some more rigid than others. For the purpose of the present review, we assume the idealization that a balanced bilingual child is a child who can function, age appropriately, in his or her two languages.
When Peal and Lambert compared their sample of French-English balanced bilingual fourth graders with a group of comparable monolinguals on a battery of intelligence tests, the results were surprisingly in favor of the bilingual children. The study had a significant impact on the field, on two different counts. First, the positive findings questioned the validity of a long string of studies that had employed the societal definition of bilingualism and had concluded that bilingualism had a negative influence on a child’s language and cognitive development. Second, the study was perceived as a methodological breakthrough. Peal and Lambert’s research paradigm (Le., a comparison of balanced bilinguals with monolin-
guals, controlling for SES,parental education, years of schooling, and other
relevant variables) promised to be a sure way to document empirically what linguists’ case studies (e.g., Leopold, 1949; Ronjat, 1913) had been claiming for years. The new paradigm, as evidenced by the studies reviewed below, fulfilled its promise.
In a detailed account of his daughter Hildegard’s bilingual upbringing, Leopold

292 k.HAKUTAB, . M. FERDMAANN,D R. M.DlAZ
( 1949) not only reported adequate language development and minimal confusion between the child’s two languages, but also suggested that bilingualism seemed to be an advantage in his daughter’s mental development. Leopold noted Hildegard’s special objective awareness of language, proposing that bilingual children, forced to make an early separation of word and referent, would develop an early awareness of the abstract and symbolic nature of language. According to Leopold, such awareness would free the child’s thinking from the concreteness and “tyranny” of words. At present, such objective awareness of language is commonly referred to as “metalinguistic awareness.”
A large number of studies have shown that, when compared with monolinguals, balanced bilingual children show definite advantages in measures of metalinguistic awareness. Ianco-Worrall (1972) showed that children raised bilingually outranked monolinguals in the capacity to compare words along semantic rather than phonetic dimensions. Cummins (1978) found that Irish-English and UkranianEnglish bilingual children outperformed monolinguals on several measures of metalinguistic awareness, including the capacity to evaluate tautological and contradictory sentences. More recently, in a study of Spanish-English bilingual children in El Salvador, Galambos (1982) found that bilinguals had a stronger “syntactic orientation” than both English and Spanish monolingual children when judging grammatical and ungrammatical sentences in both languages. Syntactic orientation was defined as the ability “to note errors in constructions, to use syntactic strategies in the correction of these constructions, and to offer syntactically rather than semantically oriented explanations for the ungrammaticality noted”
(p. 2). A study done with Hebrew-English balanced bilingual children (Ben-Zeev,
1977) clearly shows bilinguals’ awareness of linguistic rules and structure. The investigator gave children a “symbol substitution” task, measuring children’s ability to substitute words in a sentence according to the experimenter’s instructions. For example, the children were asked to substitute the word “I” for the word “spaghetti.” The children were given correct scores when they were able to say sentences like “Spaghetti am cold” rather than “Spaghetti is cold” or a similar sentence that, although grammatically correct, violated the rules of the game. Basically, in the symbol substitution task, the children were asked to violate the rules of grammar, and hence the task demonstrated their control over the somewhat automatic production of correct sentences. Needless to say, this task required an unusual awareness of and attention to linguistic features and detail. Through their performance on this and other related tasks, the balanced bilingual children showed a greater objective awareness of language than their monolingual peers.
Bialystok (1984; Bialystok & Ryan, 1985) increased the sophistication of the conceptualizationof metalinguistic awareness by hrguing that the skill consists of two components: access to the knowledge about language, and the ability to con-

Bilingualism and cognitive development

293

trol linguistic processes and apply them to a problem situation. She argued that bilingualism would influence the latter, but not the former. To support her point, she demonstrated that bilingual children were superior to monolingual controls specifically on items with anomalous meanings that were nevertheless grammatically correct. Bialystok argued that these items recruited controlled processing of linguistic knowledge, since the subject has to overlook the meaning and focus on the grammatical form. Bialystok further related her findings to the attainment of biliteracy, since of the different groups of bilinguals that she tested, the strongest effect was observed among students who had developed the ability to read in both languages. Presumably, the positive effects of bilingualism are most likely to occur in situations where the use of both languages in the literate, decontextualized functions (Snow, in press) is emphasized.
The paradigm comparing balanced bilingual to monolingual children has also been used to assess bilingual advantage on measures other than metalinguistic awareness. Balanced bilingual children outperform their monolingual peers on measures of concept formation (Bain, 1974; Liedtke & Nelson, 1968). divergent thinking skills and creativity (Torrance, Wu, Gowan, & Alliotti, 1970), and field independence and Piagetian conservation concepts (Duncan & De Avila, 1979)as well as in their capacity to use language to monitor cognitive performance (Bain
& Yu,1980).With unusual consistency, the findings suggest that bilingualism has a positive effect on a child’s developing intelligence.
Despite consistent positive findings, the methodology adopted in the studies of balanced bilingual children has been criticized (see Diaz, 1985a;Hakuta & Diaz, 1985; MacNab, 1979). The foremost criticism is that bilingual and monolingual groups are not comparable groups. Children are not randomly assigned to bilingual or monolingual upbringings and, more often than not, childhood bilingualism co-occurs with variations in a wide range of socioeconomic, cultural, educational, and ethnic variables. Regardless of experimenters’ efforts to match the groups on relevant variables, good experimental science tells us that cognitive differences between bilinguals and monolinguals could ultimately be explained by differences other than proficiency in a second language. A second criticism of this line of research concerns its exclusive focus on balanced bilingual children. These children are not representative of the majority of children who are exposed to two languages at an early age or who are educated bilingually. The findings, therefore, cannot be generalized to most populations of interest. Finally, the conclusion that bilingualism has a positive effect on children’s cognitive development has been criticized because of its gross inference regarding causality. The finding that balanced bilinguals outperform their monolingual peers can also be interpreted in the reverse way: that only the most intelligent children become truly balanced bilinguals. Research comparing balanced bilinguals and monolinguals cannot distinguish between these two alternative explanations. Of course, a third explanation is that other factors are related to both balanced bilingualism and cognitive ability.

‘I
294 .,KH.AKUTA,B . M.FERDMAN,A N D R . M . DIAZ
Degree of bilingualism and cognitive ability
A second group of studies, more modest in number than the studiesjust reviewed, have attempted to deal with current methodological criticisms by studying the effects of bilingualism using a “within-bilingual” design. The effort is directed at relating. within a group of bilingual children. the degree of a child’s bilingualism to his or her cognitive abilities. The claim is that, by using a within-bilingual design, a study not only will avoid the bilingual-monolingual comparison, but also will necessarily include children who are nonbalanced bilinguals. In addition, the inclusion of a longitudinal component in some of these studies has allowed for some analysis of the direction of causality between bilingualism and cognitive variables.
In one of the first attempts to use a within-bilingual design for assessing the cognitive effects of childhood bilingualism, Duncan and De Avila ( 1979) studied children from four Hispanic populations who differed in their relative abilities in English and Spanish. On the basis of their scores on the Language Assessment Scale, the children were assigned to one of five language proficiency groups: proficient bilinguals, partial bilinguals, monolinguals. limited bilinguals, and late language learners, where proficient bilinguals had the highest scores and late language learners the lowest scores in both languages. Subjects were given several tests of cognitive ability, including two measures of field independence and a measure of Piagetian conservation concepts.
Duncan and De Avila reported two major findings. First, proficient bilinguals ranked higher than any other proficiency group on all cognitive measures; second, no differences were found between partial bilinguals, limited bilinguals, and monolinguals on the same measures. Specifically, the data ranked the five proficiency groups in the following order: ( 1 ) proficient bilinguals; (2) partial bilinguals, monolingual~a,nd limited bilinguals; and (3) late language learners.
The investigators pointed out that the lack of a significant difference between partial bilingual, limited bilingual, and monolingual groups brings into question the “usual view of limited-English speakingchildren as being intellectually inferior to their monolingual peers” (p. 16). In addition, supporting Cummins’s ( 1976) threshold hypothesis, they concluded that, after a certain threshold of proficiency in the two languages, bilingualism is clearly related to positive cognitive gains.
A major problem in interpreting Duncan and De Avila’s (1979) data is that the observed rank ordering of proficiency groups could be attributed simply to group differences in intellectual ability or IQ rather than to differences in degree of bilingualism. Since the authors did not control for group differences in a measure of basic ability. it is possible that the proficient bilinguals and the late language learners represent the opposite tails of the IQ distribution. This IQ or basic ability confound, to which within-bilingual designs are vulnerable, was dealt with by Hakuta

Bilingualism and cognitive development

295

and Diaz (1985), Diaz and Padilla (1983, and Diaz (1985a)by the use of multiple regression techniques, as explained in the remainder of this section.
The multiple regression approach advocated by the present authors proposes that the effects of bilingualism on cognitive ability can be assessed by estimating the variance explained by second-language proficiency, once the variance ex-
plained by first-language ability and other relevant variables (such as age and SES)
is partialed out from the analysis. Specifically, the following hierarchical regression equation is proposed for the analysis of the data (the two steps in the regression are separated by a slash),
Cognitive ability = first-language proficiency + age + SESI + second-languageproficiency
where the outcome variable is any measure of cognitive ability appropriate for the age of the sample, the measure of first-language proficiency is considered a measure of “basic ability,” and the measure of second-languageproficiency is entered
last in the equation. The claim is that any changes in the variance explained (R2)
by the inclusion of second-language proficiency as the last variable in the equation is a good estimate of the effects of bilingualism on a child’s cognitive ability.
Three recent studies have taken the multiple regression approach (Diaz, 1985a; Diaz & Padilla, 1985; Hakuta & Diaz, 1985) to examine the effects of bilingualism in preschoolers, kindergarten children, and first-grade children who were, at the time, attending bilingual education programs. The measures of cognitive ability included measures of analogical reasoning, metalinguistic awareness, and visual-spatial skills for kindergarten and first-grade children and measures of classification, story sequencing, and block designs for preschoolers. Overall, the multiple regression analyses indicated significant contributions of second-language proficiency to most of the cognitive abilities measured. As reported by Hakuta and Diaz (1985) and Diaz (1985a)the findings were particularly strong for the effects of bilingualism on the Raven’s Progressive Matrices, a commonly used measure of nonverbal intelligence.
Hakuta and Diaz (1985)and Diaz ( 1985a)reported several analyses of direction of causality between bilingualism and cognitive abilities. The analyses were done on short-term longitudinal data with measures of language proficiency and cognitive ability at two points in time. Even though causality cannot be appropriately determinedfrom correlational data, longitudinal designs allow for an examination of the direction of causality between two sets of variables. Using both multiple regression and path analyses techniques, the authors reported stronger relations between language variables at Time I and cognitive variables at Time 2 than vice versa. Recognizing the limitations of their correlational data, the authors argued that, if bilingualism and intelligence are causally related, bilingualism is most likely the causal factor.
Two additional findings, reported in Diaz (1985a). are worth noting. First, in contrast to Cummins’s (1976) threshold hypothesis, which predicts positive ef-

I

296 K.HAKUTAB,. M. FERDMANA,N D RM. .DIAZ

fects of bilingualism at high levels of second-language attainment, these data suggest that degree of bilingualism may have a stronger effect on cognitive abilities for children who are at the beginning stages of second-language learning. When Diaz (1985a) examined the regression equations for groups of relative high and low second-language proficiency separately, the variance explained by degree of bilingualism was significant and substantial for the low group on most cognitive measures but was weak and nonsignificant for the high group on the same measures. These findings suggest that some effects of bilingualism might occur as a result of the initial struggles and experiences of the beginning second-language learner. This does not rule out the possibility that there are additional effects at the high threshold level.
A second important finding is that groups of high and low second-language proficiency are significantly different on measures of SES, suggesting an SES-bilingualism confound even within a somewhat homogeneous group of Spanishdominant children who are learning English in the context of bilingual education I programs. It is for this reason that SES should be controlled for in the hierarchical regression equation. We address the problem of how to interpret this confound in the section on societal bilingualism.

A review of explanatory hypotheses
The positive relation between cognitive bilingualism and children’s other cognitive abilities is well replicated. Beyond the issue of causality, a major gap in our knowledge is the lack of an explanation for this positive relation. That is, if bilingualism affects children’s intelligence, how does it do so? As Diaz (1985b) has suggested, “The gap in our knowledge is due in part to the fact that research has focused mostly on outcome rather than process variables” (p. 19). Such a focus on outcome variables does not clarify such issues as whether bilinguals solve cognitive tasks differently from monolinguals or whether the positive effects are explained by a higher rate of cognitive development fostered by the bilingual experience. Nonetheless. regardless of the scarcity of process data, several hypotheses have been formulated to explain the positive results.
The code-switching hypothesis. Code switching refers to the observation that bilinguals can move from one language to the other with relative ease. As an explanatory hypothesis, code switching was proposed first by Peal and Lambert (1962)when explaining their pioneer findings. The investigators believed that the possibility of switching linguistic codes while performing cognitive tasks gave bilingual children a flexibility that monolingual children did not enjoy. In their own words:
[the)hypothesis is that bilinguals may have developed more flexibility in thinking . . . .
lBlilinguals typically acquire experience in switching from one language to another, pos-

Bilingualism and cognitive development

297

sibly trying to solve a problem while thinking in one language and then, when blocked. switching to the other. This habit, if it were developed, could help them in their performance on tests requiring symbolic reorganization since they demand a readiness to drop one hypothesis or concept and try another. (p. 14)
More often than not, errors in cognitive and academic tasks are caused by children’s perseveration on the wrong hypothesis. Bilingual code switching might,
indeed, facilitate the development of a more flexible “mental set” to approach cognitive tasks (Duncan & De Avila, 1979). Furthermore. when a bilingual child is frustrated or blocked when performing a task verbally, he or she has the option of switching to the second language, starting the problem once again with a fresh and different perspective.

The objectification hypothesis. In a large number of studies, bilingual children have shown a special objective awareness of language. The second hypothesis claims that bilinguals’ objectification of language is conducive to higher levels of abstract and symbolic thinking.
As suggested by Lxopold (1949), bilingual children have two words for each referent and, early on, are forced to realize the conventional nature of language. The separation of word from referent is seen as one of the major milestones in the development of symbolic thinking. Furthermore, as Vygotsky (1962) suggested, since bilinguals could express the same thought in different languages, a bilingual child tends to “see his language as one particular system among many, to view its phenomena under more general categories, and this leads to an awareness of his linguistic operations” (p. 110). In other words, according to this view, learning more than one language leads not only to knowledge of a second language but to a knowledge of “language.” Through this objectification process, the hypothesis suggests, children are able to bring their concepts to a higher level of symbolism and abstraction.

The verbal mediation hypothesis. Cognitive development in the preschool years is heavily influenced by children’s increasing reliance on language as a tool of thought (Luria, 1961; Vygotsky, 1962). The use of language for self-regulatory functions, commonly referred to as “private speech,” appears shortly after the onset of social speech and gradually becomes subvocal to constitute inner speech or verbal thinking. The internalization of private speech forms the basis for the capacity to use covert verbal mediation. The origins, development, and internalization of private speech have been documented elsewhere (see, e.g., Frauenglass & Diaz, 1985; Zivin, 1979).
Several investigators(Bain & Yu,1980;Diaz, 1983;Diaz & Padilla, 1985) have
suggested that the unique linguistic experience of bilingualism and the accompanying awareness of language might lead to an increasing reliance on verbal mediation in cognitive tasks. In fact, bilingual advantage on some nonverbal measures

298 * K . HAKUTA,B. M . FERDMANA.ND R . M . DIAZ
(e.g., the Raven’s test) has been explained in terms of bilinguals’ increasing reliance on covert verbal or linguistic strategies when solving the tasks (Hakuta & Diaz, 1985). It is possible, as the hypothesis suggests, that the bilingual experience and the resulting metalinguistic awareness foster a more efficient and precocious use of language as a tool of thought. Bilinguals’ improved performance on so many different tasks could be explained by this efficient reliance on self-regulatory language.
Evaluating and integrating the models
No single study has tested a model of the process by which bilingualism might
affect a child’s cognitive development. Nonetheless, the data from several studies can be pooled and integrated, first, to examine the validity of the hypotheses reviewed above and, second, to outline some empirical constraints on the development of an explanatory model of the relation between bilingualism and cognitive ability.
In a study of the self-regulatory private speech of bilingual preschoolers, Diaz and Padilla (1985) reported two major findings that shed light on the verbal mediation and code-switching hypotheses. First, the study reported a positive relation between degree of bilingualism and production of task-relevant private speech utterances. Children in this sample with a relatively higher degree of bilingualism not only emitted more self-regulatory utterances than the other children but also used a higher number of task-relevant language functions such as labeling and description of materials, transitional utterances, guiding, and planning statements. This first finding gives some support to the hypothesis that bilingualism fosters an increased and more efficient reliance on language in cognitive tasks.
The study also examined the patterns of language switching in the private speech protocols. If the code-switching hypothesis were correct, three observations would be expected: (1) Within a given task bilingual children should switch or use more than one language, (2) the incidence of language switching should increase with tasks of increasing difficulty, and (3) the frequency of language switching should be positively related to children’s performance on the tasks. The findings, however, supported none of the three predictions. The observed frequency of language switching in private speech was minimal (less than 2%),even for those children who could easily switch languages in social situations. The findings suggest that, at least in bilingual preschoolers, language switching is a social and not an intrapersonal cognitive phenomenon.
To summarize the preceding discussion and review, a process model should take into account the following research findings:
I. Bilinguals show consistent advantages in metalinguistic awarenessand in the use of language as a tool of thought.

Bilingualism and cognitive development

299

2. There is no evidence for the suggestion that bilinguals switch languages spontaneously while performing cognitive tasks.
3. If bilingualism affects a child’s cognitive development, the effects can occur at the beginning stagesof second-languagelearning as well as at the more advanced stages of balanced bilingualism.
4. Bilingual environmentsin which the languages are used for functionsthat require controlled cognitive processing lead to stronger effects on metalinguistic awareness.
5 . The positive effects are found in bilingual additive situations(i.e., contexts where the second language is acquired without loss of the mother tongue) that
involve a somewhat systematic use of the two languages.

Taking into consideration present findings on bilingual cognitive development, we offer the following integrative hypothesis: The systematicexposureto two languages found in bilingual additive situations will give children a unique advantage in the objectification of language. Such objectification of language, in turn, will foster an increased and more efficient use of language for self-regulatory functions. These effects will be more pronounced in contexts where the decontextualized functions of language engaged in inforrnation-processing tasks, rather than conversational functions of language, are emphasized.

Cognitive bilingualism in perspective
To obtain clear answers to cognitive questions, studies must be designed with a cognitive perspective on bilingualism in mind. However, a selective focus on individual cognitive effects, when properly studied, is made at the expenseof losing contact with social psychological and societal aspects of bilingualism. Remember that what properly designed cognitive studies attempt to do is to control for societal background characteristics such that the “pure” effects of bilingualism can be discerned. Searching for such controls may be a futile and unrealistic endeavor.
Researchers concerned with the cognitive effects of bilingualism have often made methodological points regarding the proper design of studies to answer such questions. McLaughlin (1984) describes the ideal study as one that would include the random assignment of children to bilingual and monolingual groups, as well as longitudinal testing and control of relevant variables, such as intelligence. In the same book in which McLaughlin’s chapter appears, Eurly bilingualism and childdevelopment, Lebrun and Paradis (1984) title their introduction “Tobe or not to be an early bilingual?”
Although such experiments and such questions are important to pursue, one must question their ecological validity. To whom would the findings ofa study with random assignment be applied? To randomly assigned children? A focus on the social psychological and societal aspects of bilingualism highlights the way in which bilingualism is distributed in the population in a nonrandom fashion. For many children, it is not a matter of individual preference whether “to be or not to

. 300 ' K . HAKUTAB, . M . FERDMANA, N D R . M . D I A Z
be an early bilingual," and in any case, it is not a decision made by families and children on purely cognitive grounds. Moreover, the presence of two languages in an individual's environment may affect a variety of other variables that in turn may be responsible for any cognitive effects. McLaughlin (1984) does point out that the family environments of children raised monolingually are probably different from those of children raised bilingually, and that therefore it is impossible to separate the environmental effects on linguistic and cognitive variables from those of bilingualism itself. MacNab (1979) makes similar points in discussing the limitations of many of the cognitively oriented studies in this area.
In this section, we reviewed constraints on models of the relation between bilingualism and cognitive development and proposed an integrative hypothesis. This is not sufficient, however, because so far we have treated individual-level cognitive bilingualism as the independent variable and have not paid attention to factors associated with the social environment in which bilingual children develop. A more complete model must consider the context of bilingual cognitive development. In the following section, we consider perspectives that take into account the social psychological and societal correlatesof bilingualism and then discuss their implications for our models of the relation between bilingualism and cognitive development.
Social psychological and societal bilingualism
The issues of language and cognition aside, bilingualism has captured the interest of social scientists precisely because of its correlation with social psychological and societal phenomena of interest to them. Ethnographers such as John Gumpen (1982) take interest because of the roles that language plays in regulating social order by serving as a symbol of group identification and societal status. Sociologists such as Joshua Fishman (1971) take interest because language is correlated with the traditional institutional categories of the sociologist, such as the domains of society where language can be used. These other perspectives on bilingualism are important for the student of bilingualism and cognition because they grapple with the question of the determinants of the distribution of bilingualism. Even though we may establish that certain types of cognitive bilingualism are related to mental development, these types of cognitive bilingualism are not characteristics randomly distributed in the population. Bilingualism is rooted in a set of social conditions that lead particular individuals to particular outcomes.
Aside from trying to arrive at a "pure" assessment of the relation between bilingualism and cognition, then, we must consider the conditions under which various types of bilingualism might obtain and how these might be related to the cognitive models elaborated in the previous section. Investigations of the cognitive effects of bilingualism must be accompanied by an investigation of the parameters within which bilingualism occurs. Fishman (1977) makes this point quite well:

Bilingualism and Cognitive development

30 1

My own socio-historical perspective leads me to doubt that answers . . .can be found by
better controlledexperiments,which in essence. cannot explain shiftsin socialclimate that
take place across a decade or more. 1 would predict that every conceivable relationship between intelligence and bilingualism could obtain, and that our task is not SO much the determinationof whether there is a relationshipbetween the two but of when (i.e., in which
socio-pedagogicalcontexts) which kind of relationship (positive,negative, strong, weak,
independent or not) obtains. (p. 38; emphasis in original)

In one of the early attempts to account for the contradictory findings on the effects of bilingualism, Lambert (1975)proposed a distinction between additive and subtractive bilingualism. The distinction between these terms hinges on the context in which bilingualism develops and thus effectively integrates a social PSYchological perspective into the question of the effects of bilingualism. The concepts were developed to explain the divergent findings of studies that looked at immigrant or minority children from those looking at majority children in immersion programs. Additive bilingualism is said to occur when an individual acquires a second language at the same time that all abilities in the first language are maintained. In such situations, there is no threat of loss of the first language. This is the type of bilingualism most often seen in situations where children of the dominant ethnolinguisticgroup in a society learn the minority language at school, such as the case of Anglophones learning French in Canada. It can also be found in situations where the maintenance of language minority children's first language, although societally subordinate, is strongly promoted at school.
Subtractive bilingualism (also termed replacive bilingualism) refers to situations in which the group shifts in the direction of the second language while losing its ethnic language.The language situation of immigrantchildren is characterized by this type of bilingualism, in which they never fully develop their abilities in their home language while they are instructed at school in a new language,that of the host culture. In this subtractive situation, it is likely that children will be less proficient in each of the two languages than would monoglot native speakers (Cummins, 1984a).
Rather than describing the characteristics of the individual, these terms are better seen as describing the social milieu in which an individual develops his or her language abilities. The effects of each of these types of bilingualism cannot be understood in isolation from an analysis of the environment of the individual. Additive bilingualism occurs when the society values both languages and sees acquisition of the second language as a positive aspect of the child's development. This type of bilingualism occurs in situations where the linguistic and cultural systems represented by the two languages exist in a complementary fashion. In contrast, subtractive bilingualism exists where these two systems are in competition or conflict. Schooling for ethnolinguistic minorities in a society may be available only in a language different from the home language. The society may not value the minority's language, and upward mobility may be possible only when the majority language is acquired. Such acquisition may be associated with a loss

302 K . HAKUTA,B . M . FERDMANA, N D R . M . DIAZ
of the original home language. More significantly, a social milieu of subtractive bilingualism is likely to be associated with quite‘differentcharacteristics in terms of home support for language development than an additive situation. In sum, these variant social conditions are seen as leading to different types of individuallevel cognitive bilingualism.
Cummins (1976. 1981, 1984a)developed the threshold hypothesis cited earlier in order to explain why these different situations might influence bilingual children’s cognitive development.This view explains the effects found in additive and subtractive situations in linguistic and cognitive terms by seeing the development of children’slevel of proficiency in each language as a variable mediating the cognitive consequences of bilingualism. Different types of social environment in which children acquire language lead to different types of cognitive bilingualism, which in turn affect cognitive development by resulting in different levels of proficiency in each language. What is important about Cummins’s theoretical framework is that it explicitly recognizes the way in which linguistic and cognitive development must be understood as occurring within a sociocultural context. It is the differences among these types of societal bilingualism that lead to the variety of cognitive findings.
Also important to know are the conditions that lead to each of these types. By considering bilingualism, or, more precisely, degree and/or type of bilingualism, as a dependent variable, one can ask what social conditions lead to different characterizations of bilingual proficiency, at both the group and the individual level. We first discuss individual-level social psychological variables accounting for bilingualism. Then we discuss group-level factors. In both cases, however, we attempt to look for precursors to individual degree of bilingualism.
Social psychological perspectives
Robert Gardner ( 1983) addresses this question from the perspective of social psychological variables at the individual level. Subjects in his research come mostly from the English-speaking parts of Canada and thus are primarily speakers of the majority language learning a second language in a social milieu where there is little contact between the two language groups. Gardner has used primarily paperand-pencil attitude measures and correlates them with various measures of second-language acquisition.
Gardner accounts for the findings of his many studies through a socioeducational model (based in part on Carroll, 1962, and Lambert, 1967)that emphasizes four elements involved in second-language acquisition: the social milieu of learning. individual difference variables (including attitudes, motivation, and language aptitude), the contexts for language acquisition, and outcomes.
Gardner hypothesizes that the cultural beliefs developed in a particular social milieu influence the development of attitude variables, which include integrative-

Bilingualism and cognitive development

303

ness - referring to positive affect toward the other language community - and
attitudes toward the learning situation - referring to the individual’s evaluative feelings about the learning context. These two types of attitudes, in turn, influence the individual’s motivation. The integrative motive is the composite of these three
variables. This notion of an integrative motive was developed from Lambert’s ( 1967) distinction between an instrumental orientation toward learning a second
language - when the language is being learned primarily for utilitarian reasons -
and an integrative orientation - when the language is acquired because the indi-
vidual wants to learn more about the language group or even join it. Another hypothesis of the model is that motivation and language aptitude, two
individual difference variables, interact with the context of language acquisition formal or informal - to influence the development of language proficiency and the
outcomes of second-languageacquisition, which include both linguistic and nonlinguistic effects. In formal acquisition contexts, such as classrooms, both aptitude and motivation are seen as being important, whereas in informal contexts, motivation becomes predominant because it affects whether the learner will take advantage of the available opportunities. The outcomes need not be just linguistic
- that is, language knowledge and skills - but can also be nonlinguistic - for example, the degree to which the individual wishes to learn more of the language,
and his or her attitudes toward the second-language community. Gardner’s model is important because it clearly links cognitive variables to so-
cial ones such as attitudes. It addresses some of the complexity inherent in the development of bilingualism by viewing second-language learning as a dynamic process affected by a variety of factors acting on each other.
Unfortunately. however, much of the research supporting Gardner’s model has been done only in situations in which language majority children are studying a second language in school. In these contexts, the model has received a good deal of empirical support. As Gardner ( 1983)points out, little work has been done linking the social milieu to the individual difference variables. Although acknowledged within the model, this connection is left in a general and unelaboratedstate. When bilingualism is seen from a societal perspective, this is a crucial link to
elucidate theoretically in our view. Because Gardner’s model has not been tested in situations involving a variety of intergroup conditions, we do not know in what range of contexts it will be valid. An example of its limitations as a tool for understanding the situation of language minority children is that Gardner’s model says nothing about the role of the individual’s first language. Clearly, this takes on different importancein situations of language minority children learning the dominant language than in situations where majority children are learning a foreign language.
Fred Genesee (1984;Genesee, Rogers, & Holobow, 1983), in attemptingto expand Gardner’s model to bilingual, cross-cultural contexts by including intergroup factors in the model, has examined the role of the second-languagelearner’s