BILINGUALISM IN

DEVELOPMENT
Language, Literacy, and Cognition

Ellen Bialystok
York University
published by the press syndicate of the university of cambridge
The Pitt Building, Trumpington Street, Cambridge, United Kingdom

cambridge university press

The Edinburgh Building, Cambridge cb2 2ru, uk
40 West 20th Street, New York, ny 10011-4211, usa
10 Stamford Road, Oakleigh, vic 3166, Australia
Ruiz de Alarcón 13, 28014 Madrid, Spain
Dock House, The Waterfront, Cape Town 8001, South Africa
http://www.cambridge.org

 Ellen Bialystok 2001

This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2001

Printed in the United States of America

Typeface Sabon 10/13 pt. System MagnaType [ag]

A catalog record for this book is available from the British Library.

Library of Congress Cataloging in Publication Data

Bialystok, Ellen.
Bilingualism in development / Ellen Bialystok.
p. cm.
Includes bibliographical references and index.
isbn 0-521-63231-5 – isbn 0-521-63507-1 (pb)
1. Bilingualism in children. 2. Language acquisition. 3. Language awareness in
children. 4. Cognition in children. I. Title.
p115.2 .b5 2001
401′.93–dc21 00-045520

isbn 0 521 63231 5 hardback

isbn 0 521 63507 1 paperback
Contents

Preface page xi
1 Faces of Bilingualism 1
Experiencing Bilingualism 1
Who Is Bilingual? 4
Methodological Complications 6
Where the Research Looks 9
Proficiency; or, When Is Enough Enough? 10
Measuring Proficiency 12
A Process Approach 13
Cognitive Dimensions of Language Proficiency,
14 • Toward a Definition, 18
2 Starting with One Language 21
Landmarks of Language Acquisition 23
Lexicon 24
Syntax 26
Phonology 28
Pragmatics 30
Approaches to Language Acquisition 31
Criteria for Distinguishing among Theories 32
Formal Theories 36
Government-Binding Theory, 38 • Learnability Theory,
38 • Minimalism, 40
Functional Theories 40
Cognitive Grammar, 42 • Construction Grammar,
44 • Connectionism, 46
Out of the Dichotomy 49
Contributions and Limitations 50
Divide and Conquer 52
3 And Adding Another 56
viii Contents

Mastering the System 58

Lexicon 61
Syntax 67
Phonology 70
When It Can Be Learned 71
Rationale for a Critical Period 73
Defining the Concept 74
Nature of the Evidence 76
What Does It Mean? 82
Conceptualization, 83 • Aging Cognition,
86 • Restricting the Range, 86
Learning Two Languages 88
4 Language in the Mind 90
In Search of Representations 91
Thinking Observed 91
Behavioral Evidence 95
Beyond Methodology 98
Adults and the Problem of Concepts 100
Types of Bilinguals 100
Storing Languages and Meanings 102
Implications of Complex Representation 103
Children, Language, and Thought 105
One Language, One Mind 106
Separated at Birth 108
Developing the Systems 112
Representing Two Languages 116
Interpreting the Evidence 117
Toward Distinct Representations 119

5 Thinking About Language 121

Defining “Metalinguistic” 122
Metalinguistic Knowledge 123
Metalinguistic Ability 124
Metalinguistic Awareness 126
The Relevance of “Meta” for Second-Language Acquisition 127
Metalinguistic: The Process 129
Identifying Processing Components 131
Demands of Metalinguistic Tasks 132
Is It Better with Two Languages? 134
Word Awareness 135
Syntactic Awareness 139
Phonological Awareness 141
Interpreting the Research 143
 Contents ix

Applying the Framework 145

Why the Difference? 150
6 Link to Literacy 152
First Steps 154
Background Preparation 156
Social Context of Reading 157
Cognitive Prerequisites 159
The Invariance of Symbolic Notations,
160 • Correspondence Rules for Understanding Print,
162
Early Reading 165
The Role of Phonological Awareness 165
Bilinguals and the Awareness of Sounds 168
Advantages in Phonological Awareness, 169 • The
Language Connection, 169 • It’s Just Easier, 171
Fluent Reading 173
Factors in Second-Language Reading 174
Decoding the Text 176
The Difference in Reading 179
7 Beyond Language 182
Bilinguals and Intelligence 183
What Is Intelligence? 184
The Processes of Intelligence 186
Languages and Cognition 188
Groundwork for Interactions 190
The Style of Cognition 192
Field Dependence-Independence 193
The Symbol Basis of Quantity 195
Numbers and Numerals: The Basis of Quantity 197
Computation and Quantity 199
Concepts and Creativity 203
Problem-Solving in Bilingual Children 204
Conceptualization in a Sorting Task 206
Piecing It Together 210
Pattern of Results 211
Toward an Explanation 213
The Source of Inhibition 216
8 The Extent of the Bilingual Mind 219
A Matter of Class 220
Socioeconomics and Learning 220
A Bilingual Disadvantage? 222
Time, Timing, and Timeliness 224
x Contents

One at a Time? 225

The Learning Latency 230
Into the Future 232
Educating Children 234
Bilingual Education 234
Relationship to Achievement 238
Who Am I? 239
Lessons for Cognition 242
Language 243
Cognition 245
Brain 247
Finally 247
References 249
Index 281
1

Faces of Bilingualism

Picture a bilingual child. What languages does this child speak? What
kind of neighborhood does she live in? What are the educational arrange-
ments that either support or demand bilingualism? Are any of the child’s
languages spoken in the community outside the home? What were the
circumstances that led to her bilingualism? How long has the child been
living in the present country? By changing even a single answer to this
small sample of questions, the child being described is importantly
different from one who would have elicited a different answer. Is there a
common experience that unites this diversity of children? Is this common
experience reflected in some deeply rooted element of their intellectual
development? Does bilingualism in early childhood influence the nature of
children’s cognitive development?
These questions presuppose a more basic issue: How do we decide who
is bilingual? We all know shreds of other languages although we would
hesitate to include those imperfect systems as evidence for our bilingual-
ism. Children’s knowledge of any language is incomplete compared with
that of an adult. At what point does a child have enough command of two
languages to be declared bilingual? In part, the answer to that depends on
how the two languages were learned and to what purposes they are put.
But that does not solve the problem of deciding what is entailed by partial
knowledge of one language for a child whose linguistic knowledge of any
language is partial at best.

Experiencing Bilingualism
For adults, the idea of an “uncontaminated” monolingual is probably a
fiction. At the lowest levels of knowledge and awareness, exposure to
fragments of other languages is unavoidable. No language is immune to
2 Bilingualism in Development

intrusion from the barrage of words and phrases that rise out of one
language and through their universal appeal deposit themselves squarely
into the lexicons of another. The phenomenon can reflect cultural prestige
(blasé, rendezvous), the seat of power for commerce (computer, Big Mac),
the lineage to intellectual tradition (Zeitgeist, Angst), or the fascination
with a slightly exotic culture (chutzpah, schlep). But we would resist
describing the speakers who incorporate these terms into their conversa-
tions as bilingual. Often, many of these borrowed words are not even
recognized for their linguistic origins.
More language knowledge surely follows from the ubiquitous foreign
language requirements that most of us were required to complete at some
point in our lives. This experience may have left us with many things, but
fluent command of that language is probably not one of them. English-
speaking students of Spanish would hardly qualify for an educational
program conducted in Spanish, and Japanese-speaking students of En-
glish struggle to formulate the most rudimentary utterances in English.
Nonetheless, each of these instructional experiences leaves the student
with some facility in the language, perhaps more strongly developed for
one modality (reading or speaking, for example), and a level of comfort in
recognizing some forms and structures. Again, these students would be
unlikely to describe themselves as bilinguals.
Some people live in home environments where the language of the
extended family reveals an ethnic, cultural, or national background that is
different from that of the community. Here the adults can function in two
languages, and children born into these families may well learn some of
that heritage language through familial interaction. In some of these situa-
tions, home bilinguals are created by the deliberate decision of parents to
speak to the child in a different language, usually with one parent speak-
ing each language. In other cases, casual knowledge that the child picks up
in conversation can be supplemented by extra language classes – the
familiar Saturday or Sunday schools organized by various communities.
Often, however, there is little opportunity for formal study of this lan-
guage and little expectation that the child will learn much of it, apart from
that needed for ordinary domestic routines.
In some communities, bilingualism is simply expected. In these cases,
the social organization of language at home and at school gently and
irrevocably places children in the position of acquiring two languages.
Some of this type of bilingualism is a legacy of colonialism. Following
independence, many countries maintained the colonial language (notably,
English, French, Portuguese, Spanish, or Dutch) in most of their social and
 Faces of Bilingualism 3

official functions even though it was not a first language for the majority of
the population (Ellis, 1994). In Papua New Guinea, most children learn a
local language and Tok Pisin, the standard vernacular, before they enter
school where the language of instruction is English (Skutnabb-Kangas,
1981). In the Philippines, children may learn one of seventy languages in
the home before being immersed into English and Filipino instruction at
school (Galang, 1988). In Hong Kong where 99 percent of the population
is Chinese-speaking, English remains prevalent (Wong, 1988). English is
also the official language of Nigeria adopted from its colonial days, while
for Zaire, it is French (Ellis, 1994). These situations all demand that
children have high levels of proficiency in at least two languages.
Children who encounter another language in these ways experience
different kinds of interactions with each language, interact in different
types of social situations with each, encounter different opportunities for
formal study, and may also develop different kinds of attitudes to each
language. For these reasons, the various configurations that lead to bi-
lingualism leave children with different levels of competence in each of the
languages. When we think of bilingual children, we think of those who
appear to function equally in two languages, move effortlessly between
them, and adopt the appropriate sociocultural stance for each. Indeed, it is
an impressive sight to observe a young child, perhaps four or five years
old, engaged in conversation in different languages, controlling both, and
not struggling at the first sign that the language needs to change to accom-
modate some benighted monolingual in the group. Even these children,
however, may have experienced a broad range of circumstances that im-
portantly alter the nature of their bilingualism.
Romaine (1995) describes six patterns of home language bilingualism,
each one different, and each difference relevant. These patterns combine
values on social and linguistic dimensions, taking account of both the
minority or majority status of each language and the linguistic input
received by the child. The six types are:

Type 1: one person, one language

Type 2: nondominant home language/one language, one
environment
Type 3: nondominant home language without community support
Type 4: double nondominant home language without community
support
Type 5: nonnative parents
Type 6: mixed languages
4 Bilingualism in Development

In each case, she identifies the relevant differences and lists the major
research studies. This is a useful inventory of the circumstances children
experience in bilingual language acquisition. All these children become
bilingual at home, but all of them are learning their languages under
different conditions that undoubtedly lead to different levels of compe-
tence in each. These differences are not pursued here but considered as
variants of the experience of learning two languages in the home.
The proficiency achieved from different experiences can vary on other
dimensions besides absolute competence. Dopke (1992) distinguishes be-
tween productive bilinguals and receptive bilinguals. The first is the famil-
iar configuration of speaking two languages to some degree of compe-
tence; the second is the common but less often acknowledged
arrangement in which an individual can understand or possibly even read
a second language without being able to produce it.

Who Is Bilingual?
Who shall we include in the study of development in bilingual children? Is
there a formal criterion for proficiency that will point to the relevant
group of children? Academic speculation on this matter does not solve the
dilemma. Views vary from Bloomfield’s (1933) insistence that a bilingual
has full fluency in two languages to the more pragmatic assertion by
Grosjean (1989) that a bilingual is someone who can function in each
language according to given needs. We return to this problem of defining
proficiency below.
Begin with the formalities that are necessary for deciding about profi-
ciency in a language. There is less consensual agreement about the struc-
tural formalisms of language than we might wish. For example, as
speakers of English we feel confident that we understand the definitional
criteria for what constitutes a word. Yet speakers of some other lan-
guages, notably Chinese, have little understanding of what we mean by
word since that unit essentially does not exist in Chinese. When native
speakers of Chinese were asked to divide a Chinese sentence into words,
they first complained that the instruction made no sense and then pro-
duced a highly variable set of responses (Miller, Zhang, & Zhang, 1999).
This example illustrates that we cannot take for granted the absolute and
universal structure of language; our categorical and objective notions of
what languages look like are not necessarily accurate.
It is not only the problem of setting identifiable limits on speaker’s
proficiency that blurs the boundaries of a clear notion of bilingual. An-
 Faces of Bilingualism 5

other aspect of uncertainty is introduced by examining the particular

languages in the bilingual mix. We think of bilingual individuals as those
people who are able to speak two (or more) languages, to some level of
proficiency, but identifying what counts as a language is not a straightfor-
ward judgment. We take for granted that we know what languages are –
where one stops and the next one starts. That notion, too, is illusory: the
delineation of individual languages is often a matter of decree. The formal
differences that divide some languages, such as Dutch and Flemish or
Hindi and Urdu, are far smaller than those that divide dialects of the same
language, such as versions of Chinese or Arabic (Fabbro, 1999; Spolsky,
1998). In China, it is normal for people to know both an official language
and a dialect. These variants can be significantly different from each other.
Some languages, such as Arabic and Malasian, are diglossic. In these
cases, different styles of language are required in different settings, but the
differences between the styles can be as great as the differences between
acknowledged languages. In diglossia, one form of the language is used as
the vernacular for informal and social purposes and another is used as the
institutional form for formal, educational, or religious functions.
Children can also become bilingual by learning only one spoken lan-
guage. In these cases, children might learn a spoken language (perhaps
English) and a sign language (perhaps ASL), a system that is the same as a
natural language in every respect (Klima & Bellugi, 1979). It is normal for
hearing children of deaf parents to learn these two languages simultane-
ously in childhood, acquiring both in a completely natural manner from
their environment. Additionally, some children learn two or more sign
languages either with or without any spoken languages. Although there is
little research on these situations, the data that do exist confirm that the
bilingual acquisition of two languages, one signed and one spoken, by
young children has precisely the same in pattern and trajectory as that for
the acquisition of two spoken languages (Johnson, Watkins, & Rice,
1992).
Bilingualism also carries a psychosocial dimension that can itself pro-
foundly affect children. The language we speak is instrumental in forming
our identity, and being required to speak a language that is not completely
natural may interfere with the child’s construction of self. Children who
are bilingual because of relocation, particularly unwanted relocation, may
resent the new community language they have learned in spite of their
proficiency with it. Appel and Muysken (1987) describe how some of
these factors affect bilingual children by accounting for the attitudes to
the language and the role of language in establishing ethnic and cultural
6 Bilingualism in Development

affiliations. These factors undoubtedly have a strong causal role in deter-

mining how competent children ultimately become in each of their lan-
guages and the purposes for which they are eventually willing to use each.
The situations described above indicate some of the multidimensional
aspects of bilingualism and the complexity of defining the circumstances
that point uniquely and unambiguously to a set of bilingual children. All
the children in those examples are bilingual, even though their lives betray
very little of that common experience.

Methodological Complications
The intention of this volume is to examine how bilingualism influences the
linguistic and cognitive development of children. As we have seen, how-
ever, the designation of the subjects of study, namely, bilingual children, is
not straightforward. Criteria are needed, but there is a potential cir-
cularity in specifying what those criteria might be. Consider, for example,
that a decision is made to include all children who have even very limited
competence in a second language as bilingual, and then study their
development in contrast to absolute monolinguals. Using this approach, it
may emerge that bilingualism has little impact on children’s intellectual
growth. That conclusion, however, may be too heavily weighted by the
children whose second-language competence was restricted. If the re-
search showed that limited competence in a second-language does not
lead to the same consequences as more balanced mastery, then that would
be grounds to eliminate marginally bilingual children from the inquiry.
The problem is that we could not know that until the data were examined.
Conversely, an early decision to apply restrictive criteria to the definition
may exclude some children who were nonetheless affected by their lin-
guistic background. Using a different approach, it may be discovered that
even modest control over another language adjusts the dimensions of
children’s development. This conclusion would have been forfeited by an
overly restrictive set of criteria. The problem is that the decision about
who to include as bilingual must precede the evidence for what effect
bilingualism has on these children, a situation that is logically reversed.
Nevertheless, restrictions must be made. The earlier examples of situa-
tions that lead to some small measure of control over another language for
children illustrate the complexity of determining the parameters for child-
hood bilingualism. Children become bilingual for many reasons: immigra-
tion, education, extended family, temporary residence in another country,
dislocation, or simply being born in a place where it is assumed that
 Faces of Bilingualism 7

bilingualism is normal. These precipitating conditions are often associated

with a set of correlated social factors, such as education level of parents
and parental expectations for children’s education, degree and role of
literacy in the home and the community, language proficiency in the domi-
nant language, purposes for which the second language is used, com-
munity support for the second language, and identity with the group who
speaks the second language. Children’s development is affected by all of
these factors. The constellation of social, economic, and political circum-
stances of life have a large bearing on how children will develop both
linguistically and cognitively. If bilingual children differ from each other
in these dimensions, as they surely do, then they will also differ in the way
that their bilingualism has interacted with the highly variable dimensions
of their linguistic and cognitive development. Therefore, any averaging of
relevant developmental indices across the conditions for becoming bi-
lingual will be confounded with an array of hidden factors that crucially
influence development.
Another factor that complicates the equation is that languages can be
used for different purposes. It is reasonable to suppose that, all else being
equal, the uses for which a child must employ the second language will
influence the way in which it impacts on cognitive development. Grosjean
(1996) notes that this issue is partly responsible for the fact that bilinguals
rarely develop equal fluency in their languages. He discusses how different
causal factors, such as migration, nationalism and federalism, education,
trade, and intermarriage, lead to different uses of each language in each
setting. The proficiency that the child develops in each language, therefore,
is a specific response to a set of needs and circumstances. Some of these
specific functions become embedded as immutable aspects of language
proficiency: bilingual adults routinely count and pray in the language in
which they first learned these behaviors (Grosjean, 1996; Spolsky, 1998).
To understand the role of bilingualism in children’s development,
therefore, not only must bilingualism be defined precisely but also must it
be separated from the myriad of social conditions with which it is corre-
lated and linguistic contingencies with which it is confounded. This is not
easily done. The procedure for discovering how bilingualism impacts on
development is to engage in controlled research, but these definitional
ambiguities mitigate against the creation of a clean empirical design.
In constructing research designs, the attempt is to identify the factors,
or independent variables, that create potentially relevant groups. Some-
times the factors are included in order to dismiss their role in behavior,
such as including gender in a design when it is hoped that no gender
8 Bilingualism in Development

differences emerge. In a balanced design, the differences between groups

lead to simple inferential conclusions. If the difference between groups is
not statistically reliable, then the conclusion is that the two groups are the
same. For example, a set of results indicating that a group of male and
female participants who are otherwise comparable scored the same on a
concept formation task would lead to an acceptance of the null hypoth-
esis, namely, that gender (the independent variable) is irrelevant to perfor-
mance. Sometimes, as in the case of gender, it is desirable to be in a
position to accept the null hypothesis.
The more interesting aspect of research design is in constructing the
independent variables that we wish to use as a basis for rejecting the null
hypothesis. We want a difference to emerge between groups and we
would like statistical evidence that the difference is reliable. Usually, this is
relatively straightforward. We may believe, for example, that a significant
change in performance occurs at a certain age, or in a specific instructional
program, or for speakers of a particular language. The procedure is then
to assign participants to levels or groups determined by these independent
variables, such as age, program, or language. The reason for this is proba-
bly due more to limitations of analytic sophistication than conceptual
imagination. In other words, the procedures for analyzing the data require
this categorical assignment to groups, even if our conceptual notions are
more graduated. Simple judgments about membership in a level of the
independent variable are central to carrying out the statistical analysis, an
indispensable step toward making empirical conclusions.
There are two problems when this empirical model is applied to the
study of bilingual children. The first is that bilingualism is not a categori-
cal variable. Any assignment of children to a group labeled either bilingual
or monolingual is an obfuscation of the complexity of the concept of
bilingualism and a diminishment of the intricacy of children’s language
skills. Bilingualism is not like age, or gender, or grade, or any of the usual
variables we use to classify children in developmental research. At best,
bilingualism is a scale, moving from virtually no awareness that other
languages exist to complete fluency in two languages. At what point on
this scale do we declare children to be bilingual? How do we conduct
research on the impact of a variable that we struggle to define?
In the ideal research design that compares performance across groups,
the two groups are exactly the same except for the single independent
variable we have chosen to study. This clear divide between the groups is
necessary if we are to interpret any performance differences that emerge
between them. With everything else being equal and controlled, signifi-
 Faces of Bilingualism 9

cant differences in performance that may emerge can be attributed only to

the single dimension that divides the otherwise equivalent samples. It is
then a straight line from examining the data to interpreting the meaning
conveyed by those results.
The second problem is in the equivalency of the groups, even if
categorical placements can be achieved. Bilingual children are never ex-
actly the same as an otherwise comparable group of monolingual children
except for the number of languages they speak. In some inevitable sense,
bilingual children live different lives than their friends and neighbors who
may be socially, economically, and politically similar but speak only one
language. Bilingual children may have different home arrangements, per-
haps being cared for by an extended family member who speaks another
language. Bilingual children may travel more than monolinguals, making
family visits to some other homeland. Bilingual children may spend more
time than monolinguals in formal schooling, attending after school or
weekend classes in their other language. Any of these differences that
come with the bilingual experience may itself have an impact on aspects of
language and cognitive development, aside from the bilingualism per se.
This situation presents an immense challenge to research. Controlled
investigation of the impact that bilingualism might have on children’s
development requires that bilingual children are compared with equiv-
alent monolinguals on specific aspects of performance. In the absence of a
truly ideal control group, every effort must be made to assure that the
experiences encountered by the two groups of children in the study are as
comparable as possible. Additionally, it is imperative that an assessment
of broad intellectual functioning take place to provide empirical confirma-
tion that, on important developmental indices, the two groups are operat-
ing at the same cognitive level. The approach to handling this design
problem is to make every effort to minimize the effect of extraneous
variance by being scrupulous in designing the research studies. The only
alternative would be to maintain an excessively purist attitude and refuse
to participate in research that did not conform to the most rigorous defini-
tion of design control. That option, which would paralyze any scientific
examination of the development of bilingual children, seems to be both
unnecessary and indefensible.

Where the Research Looks

Research on bilingualism probably fails to reflect its diverse reality. The
considerations that constrain research studies normally prevent many
10 Bilingualism in Development

types of bilingual children from being included in scientific inquiry. An

important source of evidence for how bilingual children develop language
and cognitive skills comes from carefully recorded diary studies. Such
accounts were among the first evidence for the nature of bilingual
development, beginning with Ronjat (1913), reaching an important level
with Leopold (1939– 49), and proliferating greatly after that (Arnberg,
1979; Fantini, 1985; Hoffman, 1985; Saunders, 1982; Taeschner, 1983;
Vihman, 1985; Volterra & Taeschner, 1978). These studies form an essen-
tial part of the database. They cover diverse languages, different home
arrangements, and together include a reasonably large number of chil-
dren. The majority of these accounts, however, reflect a single reality: an
educated middle-class family that has made a conscious decision to raise
the children with two languages. Although this does not undermine the
reliability of the descriptions produced by these studies, it does leave open
a question about their generalizability to other social contexts. This issue
is discussed in Chapter 8.

Proficiency; or, When Is Enough Enough?

The problem of knowing who is bilingual conceals a more basic question:
how much is enough? Who among us does not know pieces of some other
language – words or phrases, perhaps a rule or two, and some social
routines for greeting, toasting drinks, or asking directions? These frag-
ments hardly count as competence in the language, but how much more is
required before some implicit threshold is reached? Accepting the stan-
dard assumption that no bilingual is ever equally competent in both lan-
guages, how much language is needed before we agree that a person is
bilingual?
The answer depends on how we define language proficiency. We talk
about language as though it had concrete existence and could be measured
by scientific instruments. We describe the acquisition of language as
though we move irrevocably from a state of innocence to one of mastery
along a predictable path. We identify language impairment, language
delay, and language precocity without ever specifying the standard against
which these cases are to be judged. We use “language” in research designs
as both a dependent and an independent variable, choosing fragments to
serve as stimuli but concluding truths that define the domain. But what is
the norm for language competence? What do we mean by language profi-
ciency? What are its components and what is the range of acceptable
variation? Although these questions may seem to be prior to any use of
 Faces of Bilingualism 11

language as a research instrument or conclusion about language ability in

individuals, they rarely if ever are explicitly addressed.
There is an intuitive sense in which the question of judging proficiency
seems trivial: we find the task of rating speakers along various dimensions,
for example, for their pronunciation or grammatical correctness, to be
straightforward and meaningful. These judgments, however, tend to be
based on holistic impressions, the details of which elude us. What features
of language are most important? Joseph Conrad obviously had masterful
command of English grammar but reputedly appalling phonology. Skilled
mimics can sound like a native speaker of anything while producing vir-
tual gibberish. How can a term that has such wide application be so
lacking in definitional rigor?
For a theoretician, a definition is specific to an epistemology, a set of
assumptions and principles about what language is, how it is learned, and
what is entailed when it is used. Different classes of linguistic theories are
derived from different epistemologies and therefore lead to different con-
ceptions of language proficiency. The two major perspectives in linguistic
theory are formal and functional approaches. These theories differ in
many ways, but their assumptions regarding the mechanisms for language
acquisition and the nature of language proficiency are strikingly disparate.
In general terms, formal theories posit endogenous mechanisms for lan-
guage acquisition that lead to uniform and universally prescribed rule
systems. Functional theories posit exogenous factors for acquisition based
on social interactions that lead to specific linguistic forms being extracted
from these encounters and building up over time into more formal linguis-
tic rules. A more detailed comparison of these theoretical orientations is
presented in Chapter 2. For the present purpose, the important point is
that each perspective considers different linguistic dimensions to be essen-
tial. Therefore, for each orientation, the criteria for determining language
proficiency are rooted in different domains.
Summarizing the conception that arises from each position, we have
the following. For formal linguistic theories, language proficiency is the
reflection of circumscribed and specialized knowledge that is an elabora-
tion of an abstract template. This template is part of the inheritance of
humans and develops with little need for social manipulation, provided
children are placed in normally functioning environments where a com-
munity language is heard. It is frequently related to cognitive theories in
which many such specialized modules coexist, each with their own dedi-
cated knowledge and procedural specifications. For functional linguistic
theories, language proficiency is the reflection of cognitive processes that
12 Bilingualism in Development

extract regularities from the environment and record those generalities as

knowledge. In this sense, linguistic knowledge is no different from other
kinds of knowledge of the world – knowledge about the nature of objects
and categories, for example – and becomes part of the child’s knowledge.
The mental representations for language are equivalent to the mental
representations for any other aspect of children’s knowledge of the world.
Interaction is crucial for the accrual of this knowledge.
There is no doubt that these are different conceptions of language
proficiency. Does one provide the correct description? For formalists,
language is defined by its structure; for functionalists, it is defined by its
meanings. Is there a way of measuring one or both of these aspects that
will yield an objective assessment of language proficiency?

Measuring Proficiency
If language proficiency is entailed by both of these perspectives, then it
may result in a concept so broad that it serves only to deposit a night-
marish disarray at the door of those whose responsibility is language
testing and assessment. However, the strategy of excluding parts of the
story leaves researchers in a worse situation. Should the criteria for profi-
ciency be based on broad abstractions or concrete details? To recast the
question in Chomskian terms, are the correct criteria the abstract formal-
isms of competence or the actual rules revealed through performance?
Part of the resolution to the measurement problem is in the difference
between criterion-referenced and norm-referenced assessment. In criter-
ion-referenced evaluations, performance is compared with a known stan-
dard having objective levels and limits. Progress is easily tracked as
learners move gradually along the scale toward the objectives that signal
higher levels of competence. Higher scores mean higher competence, and
there is usually a level of perfection that additionally defines mastery.
Using a trivial example, height can be measured on a criterion-referenced
scale showing growth from childhood into adulthood, the eventual goal.
The approach also has wide application in language research. The objec-
tive judgments about a language learner’s mastery of the standard rules of
structure, morphology, and pronunciation are part of the observable pro-
gress that signals the process of language learning.
The “nativeness” of the learner’s speech is another matter and is not so
easily captured by lists of features and sets of criteria. Assessments that are
outside the boundaries of purely formal prescriptions cannot be con-
ducted through consultation to a set of rules. One of the problems is that
 Faces of Bilingualism 13

the prescriptive rules provide an incomplete and somewhat inaccurate

account of what native speakers really do with language. “Performance”
is not simply different from “competence”; it is systematically different.
Speech communities adopt characteristic styles of speech and accept spe-
cific digressions from the rules. The only option for evaluating language
learners on these grounds is to make subjective comparisons that estimate
how close the learner comes to behaving as a native speaker. These judg-
ments must be made by norm-referenced evaluation. For this, there are no
absolute criteria for performance that exist outside the use of the system
by a group. Norm-referenced evaluation compares an individual with the
other members of a particular class, group, or community. Using height
again, adults can be described by their height relative to other adults.
How do we arrive at the standards that define the norm-referenced
criteria? Consider the problem of deciding about standard native-speaker
English. The easiest way to do this is to construct a mental model of the
prototypical native speaker and use that mold as the template for judging
other speakers. But will different judges create the same template? The
received pronunciation versions of English that are broadcast on the
nightly news by the BBC in Britain, the CBC in Canada, NBC in the
United States, and ABC in Australia, to name just a few, are identifiably
different from each other. The problem becomes even more complex
when we move from the acknowledged purveyors of linguistic standards,
namely, news anchors, to the great unwashed territory of real people.
Looking only at England, the disparity between the language spoken by
the BBC broadcaster, a clerk in Yorkshire, a farmer in Devon, and a
teacher in East London is undoubtedly greater than that among the inter-
national examples listed above. What is the appropriate yardstick against
which we should measure learners of English as a second language? At
what point would we decide that a learner’s proficiency is so advanced
that they have achieved native-speaker competence? These are difficult
questions but neglecting to address them does not eradicate the problem.

A Process Approach
We need to make the construct of language proficiency stand still long
enough to be a meaningful measure of the knowledge and skill individuals
have with language. We need to establish fixed criteria that supersede the
theoretical squabbles and point to critical landmarks in language mastery.
These are lofty goals, but without some framework for evaluating pro-
gress it is impossible to produce meaningful descriptions of the acquisi-
14 Bilingualism in Development

tion of language, let alone its potential impact on development. The inten-
tion here is not to solve the problem but rather to simply point to ap-
proaches that may eventually provide a fruitful resolution.
Ultimately, language proficiency must include both formal structure
and communicative application; it must evolve from a prepared mind and
be nurtured by a supportive context; it must set clear standards of use and
include disparate (but systematic) variations of the rules. We need a way
of organizing this multiplicity into a coherent statement about the human
potential to learn and use language. If there is no agreement about what is
included in language proficiency, then any explanation that attempts to
probe some of the more profound mysteries of language will be incom-
plete.
One way to take account of these perspectives is to adopt a process-
oriented approach to language proficiency based on identifiable cognitive
operations. This method would ideally set the boundaries of proficiency,
acknowledge variability, but still provide some metric for gauging a
learner’s position, preferably on a number of dimensions.

Cognitive Dimensions of Language Proficiency

A framework proposed in earlier work for considering the relation
among different uses of language and their underlying cognitive require-
ments may provide a starting point (e.g., Bialystok, 1991a). This frame-
work is described more fully in Chapter 5. Two cognitive processes, analy-
sis of representational structure and control of attention, are set out as
orthogonal axes which define a Cartesian space indicating their degree of
involvement in each quadrant. Language tasks, or language use situations,
can be located in the Cartesian space to indicate their relative reliance on
each of these cognitive processes.
The cognitive process of analysis refers to the level of explicit structure
and organization that is represented with knowledge. A significant change
with development is that mental representations of knowledge become
more explicit and more structured. Children increasingly are able to know
not only unrelated facts but also the relationship among various concepts
and ideas. Other theoretical perspectives have described this development
in different terms; for example, Karmiloff-Smith (1992) points to repre-
sentational redescription as the fundamental cognitive change for chil-
dren, and Zelazo and Frye (1997) describe cognitive complexity as the
mechanism for cognitive development. Both of these perspectives describe
a process similar to the role that analysis of representational structures
plays in building up mental representations. Control refers to the level of
 Faces of Bilingualism 15

attention and inhibition recruited during cognitive processing. These

mechanisms of attention regulate the access to and activation of the men-
tal representations that are involved in performing various tasks. It bears
some resemblance to the notion of agency described by Russell (1996).
As a first rough measure, three broad domains of language use – oral,
literate, and metalinguistic – can be positioned in terms of their cognitive
demands, giving the impression of a linear development from the first to
the last. Each subsequent domain roughly requires higher levels of both
analysis and control. This progression is shown in Figure 1.1. The linear
impression is created because the underlying cognitive processes increase
equivalently with each transition. The domains are only broad categories
and conceal great variability within each, so a more detailed array can be
constructed for each domain.
Imagine now a closeup of the domains individually. Under a larger
magnification, a single domain could be expanded to fill the space of the
whole matrix, because it too consists of graduated variability along both
cognitive dimensions. For example, the oral domain could be spread out
with various oral language activities dispersing across the space. Oral
language uses include highly skilled functions such as lecturing and more
common activities such as casual conversation. This diversity is illustrated
in Figure 1.2. The cognitive demands, and hence the degree and nature of
language proficiency involved, are strikingly different in these two cases.
Similarly, children’s conversations, consisting of short utterances situated
in the “here-and-now,” make the lowest demands on cognitive processes,
but conversations in a second language require both more formal knowl-
edge and highly skilled attention to perform at a reasonable level of
proficiency.
The same exercise can magnify and expand the variation concealed by
each of the other two broad domains. For literacy tasks, reading and
writing at different levels of competence (for example, beginning vs. fluent
reading), for different purposes (for example, studying vs. skimming), or
in different genres (for example, fiction vs. poetry) are based on different
levels of these underlying cognitive processes and convey different levels
of language proficiency. These are shown in Figure 1.3. Finally, metalin-
guistic tasks, often used as the quintessential evidence for language profi-
ciency, also differ among themselves in these ways. Some examples of
these tasks and their relative placement in this matrix are shown in Figure
1.4.
A framework still needs to be integrated with a set of formal criteria.
The procedure would be to include some of the product-oriented sets of
Figure 1.1. Three domains of language use indicating values on analysis and
control.

Figure 1.2. Tasks included in oral uses of language indicating their demands for
analysis and control.
Figure 1.3. Tasks included in literate uses of language indicating their demands for
anlysis and control (L2 = second language).

Figure 1.4. Tasks included in metalinguistic uses of language indicating their

demands for analysis and control.
18 Bilingualism in Development

abstract linguistic features that define formal linguistic theories in this

process-oriented account. In that way, the absolute, formal, or criterion-
referenced standards would be stated separately for situations that share a
functional and processing context. Research questions, then, could be
made precise enough for the appropriate measures of proficiency to be
constructed. For example, research that has ostensibly examined the
viability of the critical period hypothesis for second-language acquisition
(see discussion in Chapter 3) has usually examined only one or two as-
pects of language proficiency but made claims about the question at large.
A more differentiated approach to explaining proficiency would allow us
to say what it would mean to function like a native speaker across several
domains and then to evaluate the success with which language learners
approximate those performances. This approach would likely eliminate
simplistic conclusions that force a choice between two opposing positions
and compel us to consider a more complete set of factors in our concep-
tion of language learning. Allowing for a more nuanced explanation by
removing false dichotomies and incorporating degrees of variation along
specified dimensions will advance theorizing around many questions
(Bialystok, 1998).

Toward a Definition
Where does this leave a definition of language proficiency? The primary
consequence of this process-oriented approach is that it eliminates the
possibility of constructing an overriding definition that includes the range
of skills arrayed on these matrices. There can be no single statement that
captures the multidimensionality conveyed in this view. Instead, the
definition needs to set constraints and limits, pointing to critical areas.
Language proficiency is the ability to function in a situation that is defined
by specific cognitive and linguistic demands, to a level of performance
indicated by either objective criteria or normative standards. The profi-
ciency displayed by children as they learn language is just as valid as the
proficiency of a highly skilled native speaker engaged in a formal debate,
but the demands of each situation are different in ways that can be system-
atically captured.
This sets out two agendas for assessment. The first is to establish the
criterion-referenced achievements that provide a guideline by which we
can assess the proficiency of language learners. The second is to embed
these descriptions into a context that is sensitive to the age, proficiency
level of the learner, and the linguistic functions they are required to carry
out. This embedded inventory is the norm-referenced protocol. These
 Faces of Bilingualism 19

tasks are complementary, and it is difficult to think that one could be

achieved independently of the other. Advances in testing sophistication,
developmental psycholinguistics, and sociolinguistic frameworks are the
background against which these assessments will emerge. Ultimately, the
aim is to have an objective means by which we can judge the bilingual
proficiency, as opposed to the monolingual proficiency, of young children.
Those pronouncements about the mandate for test developers are in-
tended to set out a long-term project that will introduce rigor into the
description of language proficiency. It would allow us to know, for exam-
ple, that the reading comprehension of a ten-year-old who has been study-
ing English in school for three years should be at a specific level, or the oral
conversational fluency of an eight-year-old with a particular background
in English should be at some other level. But that is a different point from
the need to have formal criteria for deciding on who is to be included in
the category of bilingual. Most research takes a rather casual approach to
this question. In case study research that tracks the development of chil-
dren’s two languages, the amount of input in each and consequently the
amount of language acquired in each is very different. In cross-sectional
designs that compare bilinguals and monolinguals, parental report is often
the only indicator of the children’s language competence. More carefully
designed studies include some proficiency testing in both languages, but
this is not always done.
So we come full circle. Having strongly asserted that a proper definition
of language proficiency is at the core of investigations of language, we
have avoided confronting it in any serious way. At present, at least, there
are no objective tests or accepted standards. Children cannot be subjected
to a simple assessment and classified for their position on some absolute
scale of bilingualism. The insurmountable problem from a methodologi-
cal perspective is simply that bilingualism is not a categorical variable.
What we need to do is to constantly be aware of the issues. Research
with bilingual children must minimally specify how the important vari-
ables distribute in the sample and how they have been controlled in the
selection of research participants. Participants need to be comparable on
the dimensions that matter – monolinguals and bilinguals must represent
similar social backgrounds, and bilinguals must use their two languages in
the same types of contexts. Ideally, proficiency tests that assess compe-
tence in the same type of language use that is being tested in the research
should be routinely administered to participants, and the bilinguals should
be assessed in both languages. Clearly, this is not always feasible.
For the most part, the studies reviewed in this volume consider that the
20 Bilingualism in Development

bilingual children have functional fluency in both languages. Using formal

criteria, especially the type that underlie criterion-referenced tests, this
proficiency would probably not be the same for the two languages, and
probably not the same as monolingual speakers of either. For example, it
is normally the case that bilingual children have a larger productive vocab-
ulary in one of their languages, and that their vocabulary in each language
is less than that of a comparable monolingual speaker (see Chapter 3).
Still, their functional proficiency in the two languages is equivalent – they
can carry on conversations and engage in the same kinds of activities.
Vague though this may be, it is an implicit acknowledgement of the myr-
iad factors that comprise fluency. It may in fact be as close as one can come
to equating underlying proficiency in two languages as a first step to
evaluating the impact of that proficiency on children’s development.
Finally, research on the effect of bilingualism on children’s development
has both a universal and unique dimension. The universal information is
that which we can extrapolate to all bilingual children; the unique infor-
mation follows from the individual circumstances of the children studied
in the investigation. We need both, and we need a means of integrating
them. The purpose of this volume, however, is to focus on the universal:
how do bilingual children learn two languages and what happens to them
in early development?