See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/232524231

Cognitive Benefits and Costs of Bilingualism in Elementary School Students:

The Case of Mathematical Word Problems

Article in Journal of Educational Psychology · August 2011

DOI: 10.1037/a0023619

CITATIONS READS
91 2,733

3 authors:

Sebastian Kempert Henrik Saalbach

Universität Potsdam University of Leipzig
30 PUBLICATIONS 336 CITATIONS 68 PUBLICATIONS 853 CITATIONS

SEE PROFILE SEE PROFILE

Ilonca Hardy
Goethe-Universität Frankfurt am Main
61 PUBLICATIONS 1,065 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Studies on concept formation View project

LIKE: Use of the L1 and L2 in foreign language learning View project

All content following this page was uploaded by Sebastian Kempert on 03 June 2014.

The user has requested enhancement of the downloaded file.

Journal of Educational Psychology © 2011 American Psychological Association
2011, Vol. 103, No. 3, 547–561 0022-0663/11/$12.00 DOI: 10.1037/a0023619

Cognitive Benefits and Costs of Bilingualism in Elementary School Students:

The Case of Mathematical Word Problems

Sebastian Kempert Henrik Saalbach

University of Frankfurt ETH Zurich

Ilonca Hardy
University of Frankfurt

Previous research has emphasized the importance of language for learning mathematics. This is
especially true when mathematical problems have to be extracted from a meaningful context, as in
arithmetic word problems. Bilingual learners with a low command of the instructional language thus
may face challenges when dealing with mathematical concepts. At the same time, speaking two
languages can be associated with cognitive benefits with regard to attentional control processes,
although such benefits have only been found in highly proficient bilinguals. In the present study, we
attempted to disentangle the effects of bilingual proficiency on mathematical problem solving in
Turkish–German bilingual elementary school students. We examined whether the positive cognitive
effects of bilingualism could be found not only in highly proficient bilinguals but also in students
with an immigrant background and a low command of the instructional or native language. Our
findings emphasize the importance of language proficiency for mathematics problem solving, as
shown by the predictive value of students’ proficiency in the language of testing (German/Turkish)
for their performance on mathematical word problems. No additional effect of the language of
instruction (German) was found for problem solving in the bilingual students’ native language
(Turkish). Furthermore, bilinguals gained scores comparable to those of their monolingual peers on
word problems that required attentional control skills although performing significantly below their
monolingual classmates on ordinary word problems, suggesting that bilinguals have an advantage
when it comes to attentional control. Finally, bilingual students with a relatively high command of
the instructional language performed better on word problems presented in German than on those
presented in Turkish, thus facing cognitive costs when transferring knowledge from one language to
the other. Implications of our findings for bilingual education are discussed.

Keywords: bilingualism, executive control, mathematical word problems, instructional language, immi-
grant children

The importance of students’ competence in the instructional
language for school achievement has been an issue of educational
policy ever since international comparative studies revealed that
students with an immigrant background face serious academic
problems both in elementary school and secondary school. For
countries like Germany or Switzerland, studies on school achieve-
ment such as the Program for International Student Assessment
(PISA) or the Progress in International Reading Literacy Study
(PIRLS) have shown that in all three domains of assessment—
reading, mathematics, and science—students with an immigrant
background performed significantly below their native-born peers
This article was published Online First June 20, 2011.
Sebastian Kempert and Ilonca Hardy, Department of Educational Sci-
ences, University of Frankfurt, Frankfurt, Germany; Henrik Saalbach,
Institute of Behavioral Sciences, Eidgenössische Technische Hochschule
(ETH) Zürich, Zürich, Switzerland.
Correspondence concerning this article should be addressed to Se-
bastian Kempert, Department of Educational Sciences, University of
Frankfurt, Frankfurt am Main 60325, Germany. E-mail: kempert@em
.uni-frankfurt.de
(Baumert & Schümer, 2001; Schwippert, Bos, & Lankes, 2004;
Schwippert, Hornberg, & Goy, 2008; Walter & Taskinen, 2007).
Throughout their school careers, bilingual students’ risk of being
retained is two to three times higher than that of their monolingual
classmates (Baumert & Schümer, 2001). Some analyses have
shown the contribution of general measures of socioeconomical
and cultural backgrounds to these striking differences in school
careers (Müller & Stanat, 2006). Most specifically, however, mas-
tering the instructional language is the greatest barrier for students
from immigrant families (Baumert & Schümer, 2001; Organisation
for Economic Co-Operation and Development [OECD], 2006):
Almost 50 % of all 15-year-olds with an immigrant background
failed to reach more than the first reading competence level in
PISA 2000. In other words, these students’ reading abilities barely
go beyond the localization of explicitly stated information and the
interpretation of and the reflection on very simple statements in a
given text. Similar results have been reported from PISA 2003,
2006, and 2009 (Stanat, Rauch, & Segeritz, 2010; Walter, 2008).
On the basis of these findings, it has been suggested that the
deficits of students with an immigrant background in mastering the
language of instruction, as reflected by low reading scores, will

547
548 KEMPERT, SAALBACH, AND HARDY
have a cumulative effect in most academic domains, thus contrib-
uting to an impaired development of competencies in other aca-
demic fields such as mathematics and science (Baumert &
Schümer, 2001; OECD, 2006).
However, these disadvantages of bilingual students with an
immigrant background sharply contrast with research findings that
highlight a number of specific positive cognitive consequences of
speaking two or more languages (for reviews, see Adesope, Lavin,
Thompson, & Ungerleider, 2010; Bialystok, 2009; Bialystok,
Craik, Green, & Gollan, 2009). Numerous studies with highly
proficient bilingual speakers revealed that bilinguals, due to their
daily practice of two different language systems, tend to develop
an increased ability to perform executive control processes (Bia-
lystok, 1988; 1999; 2009; Bialystok, Craik, & Luk, 2008; Bia-
lystok, Craik, & Ryan, 2006; Cromdal, 1999; Martin-Rhee &
Bialystok, 2008; Rodriguez-Fornells, de Diego Balaguer, &
Münte, 2006).
In the present study, we attempted to disentangle the effects of
bilingualism on academic achievement in students with an immi-
grant background. In particular, we examined whether and to what
extent bilingual (Turkish–German) third graders’ mathematical
achievement is affected by aspects of bilingual proficiency.
Turkish–German bilinguals, who constitute one of the largest
groups of primary and secondary students with an immigrant
background in Germany (Ramm, Prenzel, Heidemeier, & Walter,
2005), have been found to be the most strongly disadvantaged
group in school assessments as compared with other immigrant
groups (Stanat, 2008). In the following, we will first discuss the
role of students’ proficiency in the instructional language and
native language for mathematical learning. We will then provide
arguments for why the positive cognitive effects of bilingualism
found in highly proficient bilinguals may be related to mathemat-
ics achievement and show why it is likely that these cognitive
advantages will also extend to bilingual students with an immi-
grant background.
Language and Mathematics Learning
In the present study, we explored the effects of bilingualism on
arithmetic word problem solving. This context was chosen not
only because of the great importance attributed to mathematics in
the school curriculum but also because the learning of mathematics
is more strongly related to language processes than has been
assumed. According to Bialystok (2001), mathematics is a domain
where cognitive effects in bilinguals are likely to occur as lan-
guage and mathematics share common critical features such as
abstract mental representation, conventional notations, and inter-
pretive function. Furthermore, cross-cultural differences in chil-
dren’s mathematical performance have been partly traced back to
differences in the way languages code numerosity (e.g., Gordon,
2004; Miller, 1996; Miller, Smith, Zhu, & Zhang, 1995). Indeed,
both cognitive and educational research has produced ample evi-
dence that language processing and mathematical problem solving
are closely connected.
The importance of language for the acquisition of mathematical
concepts and procedures has also been revealed by experimental
research. For example, in a training study, Spelke and Tsivkin
(2001) had highly proficient adult Russian–English bilinguals
practice different mathematical problems in their respective native
languages in a training course. A posttraining test, carried out in
both languages, revealed that the retrieval of information was
significantly more accurate and faster when the language relied on
for problem solving was the language of instruction rather than
participants’ other language (in which they were proficient as
well). Similar effects were also reported with respect to mathe-
matical word problems (Bernardo, 1998) and in domains other
than mathematics (e. g., Marian & Fausey, 2006). These effects
have been attributed to the specific cognitive costs (so-called
switching costs) that arise when participants are required to switch
languages for information retrieval (Saalbach, Eckstein, & Grab-
ner, 2010; Spelke & Tsivkin, 2001). For example, in a recent
neuropsychological study on Italian–German bilinguals, Saalbach
et al. (2010) found that the cognitive costs associated with lan-
guage switching in the context of mathematical problem solving
were due to additional calculation processes required when trans-
ferring knowledge from the language of instruction to the language
of retrieval.
These findings suggest that the encoding and retrieval of content
taught in school are closely connected to the language of instruc-
tion. Students with a low command of the instructional language
may thus face the risk of insufficient mental representation of
information presented at school, which may eventually result in
poor performance in different subjects. In fact, Marian and Fausey
(2006) found that their dominant Spanish–English bilingual par-
ticipants (i.e., participants whose proficiency was significantly
higher in Spanish than in English) performed significantly below
balanced bilinguals (i.e., participants whose proficiency was the
same in Spanish and in English) in the retrieval of information
when English was the language of both instruction and retrieval. In
school programs like Structured Immersion or Transition, instruc-
tors often cope with this problem by using the students’ native
language as a resource during instruction until students’ profi-
ciency in the second language has developed to an adequate level
(see Hakuta, 1999).
The link between language and mathematics is especially evi-
dent in the case of word problems. According to a prominent
model of arithmetic word problem solving, the first step in learning
to solve such problems is for children to form a situational model
of a given problem by structuring its relevant features. In a next
step, they need to extract a mathematical problem model by
reframing the linguistically coded relations of the situational
model as mathematical relations. Finally, the children need to do
the mathematical calculations, then interpret and validate the re-
sults, and, eventually map them onto the actual problem (c.f.
Reusser, 1997; Verschaffel, Greer, & De Corte, 2002).
Given this model of arithmetic word problem solving, the in-
fluence of language on achievement could be threefold: First,
instructional processes in the classroom are highly language based,
with co-construction processes such as the negotiation of concepts
and norms as well as definitions, explanations, and argumentations
taking place during mathematics instruction (Elbers & de Haan,
2005). Thus, low proficiency in the instructional language may
hamper students’ initial comprehension of the steps needed to
carry out the mathematical modeling process involved in solving a
word problem. Second, there are certain linguistic structures which
imply mathematical structures and operations, thus providing es-
sential information on the mathematical model within the task
statement. Children with a low proficiency in the language in
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM
which the word problem is presented may not recognize and
interpret these verbal structures adequately. For example, the ap-
propriate understanding of relational terms like “more than” or
“less than” is critical for extracting a mathematical model (Stern &
Lehrndorfer, 1992). In fact, simplifying the linguistic structure of
word problems presented in English has been shown to result in an
increase in the performance of learners of English as a second
language while it had no effect on native speakers (Abedi & Lord,
2001). Finally, problems in comprehending the task on a surface
level (i.e., a noun or a verb may be unknown) or the need to
translate certain aspects of a problem may result in a greater load
of working memory resources which, as a result, are not available
for carrying out arithmetic calculations (e.g., Shaftel, Belton-
Kocher, Glasnapp, & Poggio, 2006). While the latter two aspects
refer to effects of language on testing in a narrow sense (compre-
hension of the problems to be solved), the first aspect involves the
social co-construction of meaning during instruction that exerts a
cumulative influence on individuals’ academic achievement.
Cognitive Consequences of Bilingualism
The so-called threshold hypothesis proposed by Cummins
(1979) particularly emphasized cognitive disadvantages due to low
language proficiency. Here, negative effects of bilingualism may
be particularly prevalent in bilingual speakers who have deficits in
both languages (semilinguals or weak bilinguals). Given the find-
ings reported earlier, two types of cognitive costs associated with
bilingualism in the context of academic learning are possible: One
type of cost results from low proficiency in the instructional
language, the other one from the need to switch between languages
of instruction and retrieval. In both cases, cognitive costs refer to
the additional mental effort that leads to disadvantages in academic
learning or performance. In the case of language switching, as
suggested previously, additional information processing is re-
quired in the language of retrieval. In the case of low language
proficiency, more cognitive resources are taken up by, for exam-
ple, language comprehension or repeated modeling processes due
to incorrect mappings. Thus, fewer cognitive resources will be
available for the actual learning process.
Beyond postulating negative consequences for students with
low language proficiency, Cummins also included in the threshold
hypothesis the assumption that there are positive cognitive conse-
quences of speaking two (or more) languages, which are expected
to occur especially in bilingual speakers who are highly proficient
(i.e., native-like) in both of their languages (so-called balanced
bilinguals). According to Cummins, at a lower level of bilingual
proficiency with only one of the two languages at the level of
native speakers, the negative cognitive consequences of bilingual-
ism may be avoided (in so-called dominant bilinguals).
Several experimental studies in which both linguistic and non-
linguistic tasks were used have indeed provided evidence that
balanced bilinguals will show specific cognitive benefits. In par-
ticular, these bilinguals are assumed to have developed an im-
proved ability of executive control due to their daily practice of
keeping their two language systems apart in everyday situations
(Bialystok, 1988; 1999; 2009; Bialystok et al., 2006, 2008; Carlson
& Meltzoff, 2008; Cromdal, 1999; Martin-Rhee & Bialystok,
2008; Rodriguez-Fornells et al., 2006). According to Bialystok,
this practice helps them to “selectively attend to specific aspects of
549
representation, particularly in misleading situations” (Bialystok,
2001, p. 131). One commonality is underlying all the tasks having
revealed a bilingual advantage: They require subjects to respond to
cues that are embedded in a misleading context and thus create a
cognitive conflict. A cognitive conflict can arise because two
competing cues have to be monitored (e. g., ignoring the position
of the cue on a screen when responding to the color of the cue) or
because the habitual response has to be suppressed in favor of a
less familiar one (e. g., labeling pictures of the sun as “night”). In
a series of experiments, Martin-Rhee and Bialystok (2008) found
that bilinguals’ advantage is due to their improved control of
attention rather than to the inhibition of a habitual response (see
also Carlson & Meltzoff, 2008; Colzato et al., 2008). They argued,
“For bilinguals, their two linguistic systems function as bivalent
representations, offering different, potentially competing response
options to the same intention or goal. To manage this conflict,
bilinguals must attend to the relevant language system and ignore
the unwanted system to assure fluency in speech production”
(Martin-Rhee & Bialystok, 2008, p. 85).
Given the findings on balanced bilinguals’ improved ability to
monitor competing cues, it is extremely interesting to ask whether
this cognitive advantage of bilingualism also applies to learners
with high proficiency in one language only (dominant bilinguals).
According to Bialystok and colleagues, the routine of switching
between two (or more) symbol systems and the ability to do so in
a controlled and adequate manner (in terms of keeping the symbol
systems apart) may be the mechanism that underlies the gain in
attentional control skills in the bilingual brain. Therefore, we posit
that as soon as a basic communicative level is established in both
languages, switching becomes a routine, and attentional control
skills are trained in the process. Turkish–German bilingual stu-
dents, for example, typically switch languages between home and
school (Stanat, 2008). We may thus expect the cognitive benefits
of bilingualism for executive control to show not only in highly
proficient bilinguals but also in bilinguals with different degrees of
proficiency in their languages. To our knowledge, there have been
no published studies so far focusing on whether dominant bilingual
children in general and students with an immigrant background in
particular (i.e., a group that is usually associated with academic
difficulties) show any beneficial cognitive consequences due to
their bilingual experience.
Background of the Present Study
In the present research, we attempted to disentangle the possible
costs and benefits of bilingualism with respect to the mathematical
achievement of students with an immigrant background. We were
particularly interested in the nature of the cognitive consequences
of bilingualism for immigrant children with a low socioeconomic
status, as recent discussions in educational science and policy have
typically focused on their academic problems. We believe that
detecting cognitive advantages in this bilingual group would pro-
vide an additional perspective on their academic situation as well
as possible means of support.
We used a series of mathematical word problems as a measure
of mathematics achievement. Children’s performance on arithme-
tic word problems is considered a reliable predictor of their later
mathematical competence. In fact, the Munich Longitudinal Study
on the Genesis of Individual Competencies (LOGIK) revealed that
550 KEMPERT, SAALBACH, AND HARDY
individual differences in math achievement during primary school,
as measured with word problems, predict differences in math
achievement at the end of secondary school significantly better
than differences in IQ do (correlation coefficients: .64 vs. .04 at
age 8 years; Stern, 2009). Furthermore, solving mathematical word
problems may not only be an ecologically valid but also a sensitive
measure of the cognitive benefits of bilingualism for attentional
control processes (Lee, Ng, & Ng, 2009; Pasolunghi, Cornoldi, &
De Liberto, 1999). As illustrated earlier, problem solving in math-
ematics is characterized by a process of cognitive modeling. In
order to construct an adequate situational and mathematical model,
students need to identify and extract the relevant pieces of infor-
mation from a given problem context and, at the same time,
suppress any misleading or irrelevant linguistic or numerical in-
formation. Research on word problems has shown that solution
rates of different problems with a mathematically isomorphic
structure differed strongly according to the level of the ambiguity
of the context. Staub and Reusser (1995), for example, found that
first and third graders’ solution rates decreased dramatically when
the problem “Joe had 3 marbles. Then Tom gave him 5 more
marbles. How many marbles does Joe have now?” was reworded
into “Today Dane got 11 marbles from Susan. Yesterday Dane
found 5 marbles. How many marbles does Dane have now?” In the
present research, we thus used two types of mathematical word
problems: common problems and word problems with distractors.
The former type was derived from a set of empirically tested word
problems with established solution rates for primary school stu-
dents of various ages (Stern, 1998). For the latter type, we artifi-
cially enriched common problems by including numerical infor-
mation that was not needed for solving the problem (distractors):
for example, “Maria had 3 marbles. Altogether, the marbles cost
90 cents. Then Hans gave her 5 marbles. They cost 1.50 euros.
How many marbles does Maria have now?” In this case, atten-
tional control processes are particularly called for as children need
to suppress the misleading numerical information in order to
extract the relevant data.
To assess language skills, we used a direct measure of oral
proficiency instead of the reading comprehension measure that has
typically been used in previous studies. As reported previously,
findings from international comparative studies on school achieve-
ment showed a close relation between bilingual students’ skills in
the instructional language and their achievement in different do-
mains including mathematics. In these studies, however, language
proficiency was indicated only by reading comprehension rather
than being assessed with language tests. Reading comprehension is
a global indicator of language proficiency as its acquisition also
depends on cognitive skills unrelated to language but critical to
school-related learning. In fact, it has been shown that perfor-
mance of poor readers on math tests increased significantly when
the test items were presented orally (e. g., Helwig, Rozek-Tedesco,
Tindal, Heath, & Almond, 1999). Furthermore, language profi-
ciency was usually only measured for the language of instruction
but not for the native language. Therefore, it remains unclear what
types of bilinguals in terms of Cummins’ framework (i.e., bal-
anced, dominant, or weak) were tested and how this might relate to
students’ performance. In the present study, we thus used oral
proficiency measures in both the instructional and the native
language, and we presented the mathematical word problems
orally.
Research Questions and Hypotheses
Drawing on the theoretical insights and empirical evidence
outlined, we examined whether in a group of bilingual students the
level of proficiency in the instructional language predicted
achievement on mathematics word problems in the same language.
We expected language proficiency to be closely related to the
solution rate of word problems even when the influence of arith-
metic skills, general cognitive ability, and socioeconomic status
was controlled. We also expected that our measure of (oral)
proficiency in the language of instruction would be a stronger
predictor of mathematics problem solving in that language than
reading comprehension. For word problems in students’ native
language, we expected an effect of proficiency in the language of
instruction beyond the impact of students’ proficiency in their
native language. Furthermore, we tested whether bilingual children
with an immigrant background would show a cognitive benefit, as
compared with monolinguals, concerning attentional control pro-
cesses. This effect of bilingualism was expected to be evident only
in those tasks which especially require attentional control skills,
after the effects of general cognitive ability, socioeconomic status,
and arithmetic skills were controlled.
Method
Participants
Participants were 78 third graders from five primary schools in
urban districts of Berlin and Frankfurt with a high percentage of
Turkish immigrants. Of the participants, 44 were bilingual students
with a Turkish–German background and 34 were German mono-
linguals. The bilingual sample consisted of 20 female and 24 male
students with a mean age of 8.5 years. In the group of monolingual
students, there were 14 female and 20 male students with a mean
age of 8.6 years. The Turkish–German bilingual students and half
of the German monolingual students were recruited from the same
classes within three schools in Berlin. The other half of the
monolingual sample was recruited from two schools in Frankfurt.
The sample consisted of three subsamples. Two samples were
tested in two consecutive school years (2007 and 2008, with the
testing done between January and March in both years) in the same
schools (Berlin). The third sample was tested in January 2011
(Frankfurt). Subsample 1 consisted of 43 students (32 bilingual, 11
monolingual); Subsample 2, tested in the consecutive year, con-
sisted of 16 additional students (12 bilingual, four monolingual).
Subsample 3 contained only monolinguals (19 students). Compar-
isons with regard to the control measures used (socioeconomic
status, cognitive ability, mathematics achievement; see the instru-
ments listed in the Materials section) revealed no statistically
detectable differences between the monolingual and bilingual stu-
dents of Subsamples 1 and 2 as well as between the monolingual
students of Subsamples 1, 2, and 3. Therefore, we will refer to the
entire sample in the following sections. As we used the sample of
monolingual students as a comparison group when estimating the
effects of bilingualisms in immigrant students, it was pivotal to
establish the comparability of the monolingual and bilingual sam-
ples on the various control measures.
In the three schools in Berlin selected for this study, the lan-
guage of instruction was German; additionally, beginning in first
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM
grade, Turkish language classes were provided for students with
Turkish background. Although this was not a bilingual school
setting, we assumed that the additional instruction in the Turkish
language would contribute to the development of a reasonable
level of Turkish proficiency in the participating bilingual students.
Materials
The instruments employed can be divided into language measures
(language proficiency, reading comprehension), measures of the de-
pendent variable (mathematical word problems), and control mea-
sures (cognitive ability, arithmetic skills, socioeconomical status).
Language measures.
Test of language proficiency. To assess participants’ lan-
guage proficiency we applied the Bilingual Verbal Ability Test
(BVAT; Muñoz-Sandoval, Cummins, Alvarado, & Ruef, 1998)
which is a standardized test originally developed for measuring
bilinguals’ oral language proficiency. Its three subtests are derived
from the Woodcock Language Proficiency Battery–Revised
(Woodcock, 1991). It is suited for persons from age five to
adulthood. The three subtests of picture vocabulary, oral vocabu-
lary–synonyms/antonyms, and verbal analogies were administered
to the bilingual group in both the Turkish and the German versions,
and to the monolingual group in the German version only. The
German version and the Turkish version are comparable with
regard to the range of vocabulary tested; yet, specific vocabulary
items may differ in difficulty according to their frequency of use.
In the first subtest, picture vocabulary, the children had to point at
one of four pictures that correctly showed the word the experi-
menter had presented orally. In the second subtest, oral vocabu-
lary, the children had to name synonyms to (orally) given words or
antonyms to (orally) given words. In the subtest of verbal analo-
gies, the children had to finish sentences on the basis of analogies
such as “hungry is to eat like tired to . . . (sleep).” As required in
the standardized testing procedure, individual testing in each of the
subtests continued until a child made four consecutive errors. The
German and the Turkish versions of the BVAT used in this study
were published translations of the original English version, based
on the so-called consensus translation and standardization process
(see Muñoz-Sandoval et al., 1998). Reliabilities for each subtest
are only reported for the English standardized version: rtt ⫽ .85
(picture vocabulary), rtt ⫽ .87 (oral vocabulary), and rtt ⫽ .86
(verbal analogies).
Test of reading comprehension. The Ein Leseverständnistest
für Erst- bis Sechstklässler (ELFE 1– 6; Lenhard & Schneider,
2006) is a German reading comprehension test for children in
Grades 1– 6. It consists of three subtests: (a) word comprehension,
(b) sentence comprehension, and (c) text comprehension. For the
following analyses, the subscale text comprehension was used as
an indicator of students’ reading comprehension in the context of
word problems (reported reliability of .86; Lenhard & Schneider,
2006). In this test, students have to read a short paragraph on a
topic of everyday life and then have to answer several compre-
hension questions on the text in multiple-choice format. This test
was administered in German only.
Measures of the dependent variable: Mathematical word
problems. To assess mathematical problem solving abilities, we
used two types of mathematical word problems: problems without
distractors and problems with distractors.
551
Word problems without distractors. A set of nine word
problems, consisting of three exchange problems, three compari-
son problems, two complex comparison problems, and one com-
plex combination problem, was employed. All problems were
derived from a set of empirically tested word problems with
established solution rates for primary school students of various
ages (Stern, 1998). The categorization of word problems (addition
and subtraction) with respect to the situational model (exchange,
comparison, and combination) was originally provided by Riley,
Greeno, and Heller (1983). Solution rates for the different kinds of
problems were obtained in several large-scale studies with children
from kindergarten to third grade (e.g., Riley & Greeno 1988; Stern,
1998). For the present study, it was thus possible to choose a set of
word problems which was well-balanced with respect to the ex-
pected solution rates of the sample of third-graders: According to
previous studies, the exchange problems ranged in item difficulty
from .8 to 1.0; the comparison problems ranged from .75 to 1.0,
and the complex combination problem was reported with an item
difficulty of .58. All problems had been presented in written
format in previous studies.
Combination problems require students to find the combined
amount of two quantities or the partial amount of a sum. The
complex version requires an additional step of calculation as more
than two quantities need to be considered: for example, “Paul and
Tina have 6 euros altogether. Tina has 4 euros. Kai and Nina have
9 euros altogether. Nina has 3 euros. How many euros do Paul and
Kai have altogether?” Exchange problems are characterized by the
fact that a certain amount of objects changes hands: for example,
“Sabine had 3 euros. Then Philip gave her 5 euros. How many
euros does Sabine have now?” There are three versions of ex-
change problems—final quantity unknown, exchange quantity un-
known, and initial quantity unknown—all of which were used in
this study. Finally, comparison problems require students to con-
sider the difference between two quantities. These problems are
the most difficult ones for students to solve because besides a
concept of cardinal numbers, the problems additionally require an
understanding of relational numbers: for example, “Beate has 5
stickers. Max has 8 stickers. How many more stickers does Max
have than Beate?” There are three kinds of comparison prob-
lems— quantity of difference unknown, quantity of comparison
unknown, and quantity of reference unknown—all of which were
presented in this study. The complex versions of the comparison
problems again require an additional step of calculation (see the
Appendix for an overview).
Word problems with distractors. In addition to the word
problems without distractors, we constructed a parallel set of nine
word problems that were artificially enriched by numerical infor-
mation that is not necessary for solving the problem (distractors).
As discussed earlier, these problems were specifically designed to
allow us to investigate the hypothesized differences in executive
control between monolingual and bilingual children; for example,
“Maria had 3 marbles. The marbles cost 90 cents. Then Hans gave
her 5 marbles. They cost 1.50 euros. How many marbles does
Maria have now?”
The two sets of nine problems each were then translated into
Turkish by a fluent speaker of Turkish and German. The resulting
set of nine word problems without distractors and nine word
problems with distractors showed a satisfactory internal consis-
tency (e.g., Field, 2009), with Cronbach’s ␣ ⫽ .83 for the German
552 KEMPERT, SAALBACH, AND HARDY

versions and Cronbach’s ␣ ⫽ .76 for the Turkish versions. All
word problems were administered orally in order to avoid any
direct influence of students’ reading comprehension skills (see
Helwig et al., 1999). For each word problem, a score of 2 was
assigned if a child provided a correct solution after the first
presentation of the word problem. If the problem had to be pre-
sented a second time (due to nonresponse, incorrect response, or
misunderstanding after the first presentation), a score of 1 was
assigned for a correct solution. After two incorrect answers or
nonresponses, a score of 0 was assigned. Each child could thus
attain a maximum of 18 points per set of word problems (with or
without distractors).
Control measures.
Test of cognitive ability. To control for students’ cognitive
ability, we administered a subtest of a widely used measure of
intelligence (Grundintelligenz Skala 2/Culture Fair Intelligence
Test 20 (CFT20); Wei␤, 1998) which is standardized for children
from Grades 3–10. The CFT20 is a nonverbal test and is consid-
ered to be culture-fair. The four employed subscales of (a) series
continuation, (b) classification, (c) matrices, and (d) topologic
reasoning were presented in a graphical format. Its reported reli-
ability for children in Grade 3 is .90 (Weiß, 1998).
Test of arithmetic skills. We administered the German Math-
ematics Test for Second Grade (DEMAT 2⫹; Krajewski, Liehm,
& Schneider, 2004) as an indicator of students’ ability to perform
arithmetic operations. Only subtests that test the basic arithmetic
operations of addition and subtraction were used. The children
were required to work on eight arithmetic problems that dealt with
two-digit numbers up to 99. The reported reliability (internal
consistency) for the chosen subtest is .69 (Krajewski et al., 2004).
Indicator of socioeconomic-status (SES). Finally, to control
for children’s SES, we used the “books at home” index, where
children are asked to estimate the number of books their family
owns. This measure is considered to be a quite valid estimation
of family SES (Lehmann & Nikolova, 2005; Verhoeven &
Aarts, 1998). Children in Grade 3 can answer this question quite
reliably (Schnepf, 2004). In the item that was presented, the
children could choose between five pictures showing book
shelves with about 10 books, 25 books, 100 books, 200 books,
and more than 200 books.
Results
In order to estimate the differences between the groups of
monolingual students (N ⫽ 34) and bilingual students (N ⫽ 44) in
the language measures and the control measures, we conducted
two separate multivariate analyses of variance (MANOVAs), with
univariate post hoc tests for the language measures and the control
measures. There was no significant difference between the mono-
lingual and the bilingual groups with respect to the control mea-
sures of books at home, cognitive ability, and arithmetical skills,
⌳ ⫽ .903, F(3, 74) ⫽ 2.66, p ⬎ .05. As expected, we found a
significant difference in the MANOVA on language proficiency
(reading comprehension and oral proficiency) between the two
groups, ⌳ ⫽ .688, F(2, 75) ⫽ 16.98, p ⬍ .001; ␩2p ⫽ .31. The
univariate follow-up ANOVAs revealed significant differences in
oral proficiency—BVAT in German, F(1, 76) ⫽ 34.10, p ⬍ .001,
␩2p ⫽ .31; 95% CI [11.01, 22.42]—as well as in reading compre-
hension—ELFE 1– 6, F(1, 76) ⫽ 8.63, p ⬍ .01, ␩2p ⫽ .10; 95% CI
[.85, 4.44]— both in favor of the monolinguals. For the group of
bilingual students, the difference between scores of the German
and Turkish BVAT versions suggests that these students were
more proficient in German than in Turkish. Table 1 provides a
summary of the means and standard deviations for the two groups.
Tables 2 and 3 show the intercorrelations between the various
measures for bilinguals and monolinguals separately.
Research Question 1: Does the Level of Proficiency in
the Instructional Language Predict Achievement on
Mathematical Word Problems Presented in That
Language?
In order to estimate the contribution of students’ German language
proficiency to their solution rate of ordinary mathematical word
problems (i.e., problems without distractors), we conducted a series of
multiple regression analyses with the sample of bilingual students
Table 1
Means and Standard Deviations of the Groups of Monolingual
and Bilingual Students
Monolinguals Bilinguals

Variable M SD M SD
Procedure
Socioeconomic status (no. of books at
home) 3.03 1.00 2.57 0.97
All tests were administered in two individual testing sessions of Cognitive ability (CFT20) 25.00 5.03 23.68 6.02
approximately 55 min each for bilingual children and two testing Arithmetic ability (DEMAT 2⫹) 4.82 2.51 3.80 2.00
sessions of approximately 40 min each for monolingual children. Reading comprehension (ELFE 1–6) 9.85 4.03 7.20 3.87
German proficiency (BVAT German) 71.76 13.16 55.05 12.04
Between Session 1 and Session 2, there was at least 1 week. The
Turkish proficiency (BVAT Turkish) — — 40.91 9.92
sessions took place in a separate testing room within the school Word problems without distractors
setting. (German) 11.82 2.51 9.34 3.68
In Session 1, the tests were administered in the following order: Word problems with distractors
(German) 7.24 2.34 6.48 3.49
collection of background data (name, age, gender, SES), word
Word problems without distractors
problems in German, language proficiency in German, and reading (Turkish) — — 9.39 3.27
comprehension in German. In Session 2, we administered a test of
cognitive ability, arithmetic skills, and, for the bilingual students, Note. CFT20 ⫽ Grundintelligenz Skala 2/Culture Fair Intelligence Test
20; DEMAT 2⫹ ⫽ German Test of Mathematics for Grade 2; ELFE
word problems in Turkish and language proficiency in Turkish.
1– 6 ⫽ Ein Leseverständnistest für Erst- bis Sechstklässler [A reading
The sequence of Turkish and German sessions was alternated comprehension test for first to sixth graders]; BVAT ⫽ Bilingual Verbal
randomly for bilingual students. Ability Test.
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM 553

Table 2
Intercorrelations Among Measures for the Group of Monolinguals

Variable 1 2 3 4 5 6 7

1. Books at home —
2. Cognitive ability .289 —
3. Arithmetic ability ⫺.118 .297 —
4. Oral proficiency (German) .286 .307 .346ⴱ —
5. Reading comprehension (German) ⫺.036 .326 .561ⴱⴱ .382ⴱ —
6. Word problems without distractors .050 .481ⴱⴱ .608ⴱⴱ .258 .483ⴱⴱ —
7. Word problems with distractors .113 .179 .310 .179 .218 .376ⴱ —
ⴱ ⴱⴱ
p ⬍ .05. p ⬍ .01.

only. In the main analyses, we employed only problems without
distractors as the dependent variable because these are the ones that
have been empirically validated in other studies. However, we ex-
pected language proficiency to be the crucial factor for both types of
problems. In Model 1, we included the control variables of SES,
cognitive ability, and arithmetic skills as they have been established to
be influences on mathematical reasoning processes in previous studies
(e.g., OECD, 2006; Stern, 1998). In Model 2, we additionally in-
cluded the indicator of students’ reading comprehension scores in
German because reading comprehension has typically been used as an
estimate for language proficiency (e.g., OECD, 2006). In Model 3, we
additionally included students’ scores of German language profi-
ciency in order to estimate the additional contribution of students’
German language proficiency. We used “forced entry” as the method
of entering predictors into the regression models. Table 4 lists the
results of the regression analyses for the three models. Altogether,
there was a marginally significant change of R2 in Model 2 compared
with in Model 1 (p ⫽ .058), and a significant change of R2 in Model
3 compared with in Model 2 (p ⫽ .002). Thus, language proficiency
explained a significantly larger amount of variance in the solution
rates of word problems than did reading comprehension alone. In
Model 3, arithmetic skills and German language proficiency were the
strongest predictors for students’ achievement in mathematical word
problems, with ␤ ⫽ .384 and ␤ ⫽ .430, respectively. All other
predictors, including reading comprehension, remained insignificant
in Model 3. The results of this regression analysis thus indicate that
the level of proficiency in the instructional language, together with
students’ arithmetic skills, is strongly predictive of students’ ability to
solve mathematical word problems. Additional analyses in which
problems with distractors were used as the dependent variable re-
vealed the same pattern as the analyses of problems without distrac-
tors: In Model 3, arithmetic ability and German language proficiency
remained the strongest predictors (␤ ⫽ .378 and ␤ ⫽ .403, respec-
tively; both p ⬍ .01; total R2 ⫽ .437).
In a second step, we examined whether the effect of the instruc-
tional language also predicted problem solving in students’ native
language (Turkish). To do this, we tested in a series of regression
analyses (method: forced entry) whether there was an effect of the
instructional language of German beyond the effect of bilingual
students’ Turkish language proficiency on their solving of word
problems presented in Turkish. In addition to Model 1 (which was
the same as in the first regression analysis), we included language
proficiency in the language of testing (Turkish) in Model 2 and, as
a further predictor, language proficiency in German in Model 3.
Language proficiency in German was entered last in order to allow
us to investigate the additional impact of instructional language
proficiency. Table 5 lists the results of Models 1–3. There was a
significant change in R2 from Model 1 to Model 2 (p ⫽ .004),
indicating the impact of the language of testing. The inclusion of
German language proficiency in Model 3, however, did not sig-
nificantly change R2, although it represents the second best pre-
dictor after Turkish language proficiency in Model 3 (Turkish:
␤ ⫽ .381; German: ␤ ⫽ .214). Altogether, the total R2 of Model
3 was moderate, with a total amount of 32% of variance explained.
In addition, in comparison to the regression analyses with German
word problems, the contribution of arithmetic skills seemed to be
diminished, suggesting that factors not represented in the model
contributed to problem solving in Turkish to a greater degree than

Table 3
Intercorrelations Among Measures for the Group of Bilinguals

Variable 1 2 3 4 5 6 7 8 9 10

1. Books at home —
2. Cognitive ability ⫺.080 —
3. Arithmetic ability .133 .477ⴱⴱ —
4. Oral proficiency (German) .242 .122 .341ⴱ —
5. Reading comprehension (German) ⫺.007 .183 .472ⴱⴱ .449ⴱⴱ —
6. Oral proficiency (Turkish) .095 ⫺.030 .155 .281 .321ⴱ —
7. Word problems without distractors (German) .178 .241 .579ⴱⴱ .613ⴱⴱ .479ⴱⴱ .225 —
8. Word problems with distractors (German) .048 .012 .445ⴱⴱ .535ⴱⴱ .451ⴱⴱ .147 .760ⴱⴱ —
9. Word problems without distractors (Turkish) ⫺.034 .192 .315ⴱ .354ⴱ .364ⴱ .450ⴱⴱ .348ⴱ .468ⴱ —
10. Word problems with distractors (Turkish) .070 .215 .316ⴱ .290 .169 .427ⴱⴱ .282 .339ⴱ .779ⴱⴱ —
ⴱ ⴱⴱ
p ⬍ .05. p ⬍ .01.
554 KEMPERT, SAALBACH, AND HARDY

Table 4
Regression Analyses (Models 1–3) With Bilingual Sample on Predictors of Word Problems Without Distractors (German)

Variable ␤ R2 R2 change p Tolerance VIF

Model 1 .346 .346 .001

SES .099 .956 1.046
Cognitive ability ⫺.027 .752 1.330
Arithmetic skills .579ⴱⴱⴱ .743 1.345
Model 2 .404 .058 .058
SES .121 .948 1.055
Cognitive ability ⫺.008 .748 1.336
Arithmetic skills .437ⴱⴱ .791 1.691
Reading comprehension .275 .769 1.301
Model 3 .539 .134 .002
SES .023 .887 1.127
Cognitive ability ⫺.013 .748 1.336
Arithmetic skills .385ⴱ .584 1.711
Reading comprehension .107 .662 1.511
Oral proficiency (German) .430ⴱⴱ .727 1.376

Note. SES ⫽ socioeconomic status; VIF ⫽ variance inflation factor.

ⴱ
p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.

in German. Similar to the first regression, an additional analysis
with the problems with distractors as the dependent variable re-
sulted in a comparable pattern, presenting Turkish proficiency as
the strongest predictor (␤ ⫽ .380, p ⬍ .05, total R2 ⫽ .272) while
the contribution of German language proficiency decreased further
(␤ ⫽ .116, ns). Thus, the findings of Model 3 suggest that the
importance of the instructional language cannot be simply gener-
alized to problem solving in another language (i.e., the students’
native language). Rather, it was proficiency in the language of
testing that exerted the strongest influence on students’ achieve-
ment in both German and Turkish.
Research Question 2: Are There Any Cognitive
Benefits of Bilingual Students as Compared With
Monolinguals?
To investigate whether benefits of bilingualism with regard to
mathematics problem solving could be found in our sample, we
compared the groups of bilingual and monolingual students on
word problems with and without distractors by performing a
repeated measures ANOVA with Type of Word Problem as
within-subjects factor (problems without distractors vs. problems
with distractors) and Language Group as between-subjects factor
(bilinguals vs. monolinguals). As covariates we entered SES, cog-
nitive ability, and arithmetic skills. As expected, the results
showed a significant interaction of Type of Word Problem ⫻
Language Group, ⌳ ⫽ .939, F(1, 73) ⫽ 4.76, p ⬍ .05, ␩2p ⫽ .06,
while the main effects of Type of Problem, ⌳ ⫽ .999, F(1, 73) ⫽
0.06, p ⬎ .05, and Language Group, F(1, 73) ⫽ 1.79, p ⬎ .05,
were not significant. Considering the influence of the control
variables, there was an interaction of Type of Problem ⫻ Cogni-
tive Ability, ⌳ ⫽ .944, F(1, 73) ⫽ 4.32, p ⬍ .05, ␩2p ⫽ .05, which
can be traced back to the close relation between solutions of
problems without distractors and cognitive ability (see Tables 2
and 3). Looking at the mean scores for both groups on word
problems with and without distractors (see Table 1 and Figure 1),

Table 5
Regression Analyses (Models 1–3) With Bilingual Sample on Predictors of Word Problems Without Distractors (Turkish)

Variable ␤ R2 R2 change p Tolerance VIF

Model 1 .106 .106 .207

SES ⫺.071 .956 1.046
Cognitive ability .041 .752 1.330
Arithmetic skills .305† .743 1.345
Model 2 .284 .176 .004
SES ⫺.096 .953 1.050
Cognitive ability .094 .743 1.346
Arithmetic skills .217 .719 1.390
Oral proficiency (Turkish) .429ⴱⴱ .959 1.043
Model 3 .319 .037 .161
SES ⫺.135 .915 1.092
Cognitive ability .091 .743 1.346
Arithmetic skills .157 .673 1.486
Oral proficiency (Turkish) .381ⴱⴱ .905 1.104
Oral proficiency (German) .214 .797 1.254

Note. SES ⫽ socioeconomic status; VIF ⫽ variance inflation factor.

†
p ⬍ .10. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01.
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM
Monolingulas
Bilinguals
16
14
12
Scores on Word Problems
10
8
6 trast, comprised students of varying proficiencies in native lan-
4 proficient bilinguals performing on a level comparable to that of
2
0
Without Distractors With Distractors
Figure 1. Means and standard deviations on German word problems with
distractors and without distractors for monolinguals and bilinguals. Error
bars indicate standard deviation.
the interaction Type of Problem ⫻ Language Group can be inter-
preted in terms of a relative benefit for the bilingual students with
respect to word problems with distractors. In other words, the
substantial disadvantage of bilinguals for problems without dis-
tractors relative to monolinguals diminished for problems with
distractors. Two subsequent post hoc ANOVAs (same covariates
as in the former analyses) confirmed that the monolingual students
outperformed the bilingual students on word problems without
distractors, F(1, 73) ⫽ 5.07, p ⬍ .05, ␩2p ⫽ .06, 95% CI [0.16,
2.72], while no significant group differences were obtained for
word problems with distractors, F(1, 73) ⫽ 0.04, p ⬎ .05, ␩2p ⫽
.00, 95% CI [⫺1.22, 1.52]. The stronger language proficiency in
the language of testing apparently contributed to the monolinguals’
advantage on problems without distractors. Yet for problems with
distractors, the bilingual students seem to partly compensate for
their language deficits by displaying an enhanced ability in atten-
tional control.
The bivariate correlations of German language proficiency with
performance on problems with distractors varied marginally be-
tween the groups of monolingual and bilingual students:
rmonolingual ⫽ .179; rbilingual ⫽ .535, p ⬍ .10. This finding points
to the importance of language proficiency mediating the effect of
achievement on problems with distractors, suggesting that the
cognitive benefit of bilingualism is greater in bilinguals with high
language proficiency than in those with low proficiency.
Exploratory Analysis: Are There Cognitive Costs
When the Language of Instruction Differs from the
Language of Testing?
Finally, we conducted an exploratory data analysis to test
whether bilingual students face specific cognitive costs (switching
costs) when the language of retrieval differs from the language of
instruction. This analysis picks up on Research Question 1 (second
regression analysis) regarding the influence of proficiency in the
instructional language for mathematical achievement in students’
555
native language. In the regression analysis, the addition of the
variable of instructional language in Model 3 did not contribute to
a significant change of R2 in Model 2. This seems to suggest that
there are no switching costs in mathematical word problem solving
in bilingual students of this age since achievement in word prob-
lems is apparently influenced solely by language proficiency in the
language of testing. However, in previous research on this issue
(e.g. Marian & Fausey, 2006; Saalbach et al., 2010; Spelke &
Tsivkin, 2001), highly proficient bilinguals have been used to
reveal switching costs. Our sample of bilingual students, in con-
guage and instructional language. Thus, having a group of rather
monolinguals may be needed to test the existence of switching
costs. In contrast, students with a weak command in the instruc-
tional language may actually use their native language in order to
compensate for deficits in the instructional language when solving
word problems in the instructional language. In order to test these
hypotheses, we divided our bilingual sample into three equal parts
(N ⫽ 14) by rank ordering their scores on German BVAT. The
upper third was labeled dominant bilinguals, and the lowest third
was labeled weak bilinguals. Leaving the middle third aside, we
now had two groups that we analyzed in detail. While the two
groups varied in their Turkish proficiency, we did not use this
variable as a criterion for group assignment as the distribution was
restricted in variance in the entire sample (see Table 6 for descrip-
tives of the two resulting groups). A multivariate ANOVA re-
vealed no significant differences between the groups regarding the
measures of SES, cognitive ability, and arithmetic skills. A re-
peated measures ANOVA with Language of Testing (problems
without distractors in German vs. problems without distractors in
Turkish) as a within-subjects factor, Bilingual Proficiency (dom-
inant vs. weak) as a between-subjects factor, and Turkish profi-
ciency as a covariate revealed a multivariate main effect for
Language of Testing, ⌳ ⫽ .779, F(1, 25) ⫽ 7.09, p ⬍ .05, ␩2p ⫽
.22, a significant main effect for Bilingual Proficiency F(1, 25) ⫽
21.47, p ⬍ .001, ␩2p ⫽ .46, and a significant interaction of Lan-
guage of Testing ⫻ Bilingual Proficiency, ⌳ ⫽ .577, F(1, 25) ⫽
18.35, p ⬍ .001, ␩2p ⫽ .42. In addition, there was a significant
Table 6
Means and Standard Deviations of Bilingual Subgroups
Dominant Weak
(n ⫽ 14) (n ⫽ 14)
Variable M SD M SD
German language proficiency 68.07 5.82 40.03 5.45
Turkish language proficiency 43.00 9.80 36.00 8.05
SES 2.86 1.02 2.29 1.07
Cognitive ability 23.79 6.81 22.64 6.91
Arithmetic skills 4.71 1.89 3.14 1.87
Word problems without distractors (German) 12.43 2.31 6.57 1.78
Word problems without distractors (Turkish) 10.36 2.80 7.93 2.76
Word problems with distractors (German) 8.64 3.45 4.00 2.11
Word problems with distractors (Turkish) 6.71 3.51 4.86 3.18
Note. Dominant bilinguals ⫽ bilingual speakers who have high profi-
ciency in one language only; weak bilinguals ⫽ bilingual speakers who
have deficits in both languages; SES ⫽ socioeconomic status.
556 KEMPERT, SAALBACH, AND HARDY
interaction between Language of Testing and the covariate of
Turkish proficiency, ⌳ ⫽ .792, F(1, 25) ⫽ 6.56, p ⬍ .05, ␩2p ⫽ .20.
Univariate follow-up tests revealed that the dominant bilinguals
outperformed the weak bilinguals on German word problems, F(1,
25) ⫽ 49.24, p ⬍ .001, ␩2p ⫽ .66, 95% CI [⫺7.76, ⫺4.24], which
is consistent with the regression analyses underlining the impor-
tance of language proficiency for problem solving in the respective
language. Moreover, there were no significant differences for the
two subgroups on Turkish word problems, F(1, 25) ⫽ 15.04, p ⬎
.05, 95% CI [⫺3.78, 0.61], which again suggests that the role of
the instructional language does not generally extend to testing in
the native language. More important for the explorative question,
however, is the interaction of Language of Testing ⫻ Bilingual
Proficiency, which is depicted in Figure 2. Post hoc tests for the
differences between achievement scores in Turkish and German
for each group revealed that the group of dominant bilinguals
scored significantly higher when tested in German than when
tested in Turkish, t(13) ⫽ 3.31, p ⬍ .01, d ⫽ 0.80, 95% CI [0.72,
3.42], while there were no statistically significant differences for
the group of weak bilinguals, d ⫽ 0.58; 95% CI [⫺3.19, 0.47].
However, the latter group showed a reversed trend of switching
costs, with higher means when tested in Turkish than when tested
in German.
For problems with distractors, we found the same pattern of
results as for problems without distractors: Language of Testing,
⌳ ⫽ .781, F(1, 25) ⫽ 7.00, p ⬍ .05, ␩2p ⫽ .22; Language of
Testing ⫻ Bilingual Proficiency, ⌳ ⫽ .749, F(1, 25) ⫽ 8.39, p ⬍
.01, ␩2p ⫽ .25; Bilingual Proficiency, F(1, 25) ⫽ 8.52, p ⬍ .01,
␩2p ⫽ .25. Again, dominant bilinguals scored higher when tested in
German than when tested in Turkish, while the reverse trend was
observable with weak bilinguals (see Table 6 for means and
standard deviations).
German Word problems
Turkish Word problems
16
14
12
Scores on Word Problems
10
8
6
4
2 language of retrieval emerged in our analyses.
0 gualism has beneficial effects even for students who are not highly
Dominant Bilinguals Weak Bilinguals
Figure 2. Interaction of Language of Testing ⫻ Bilingual Proficiency for
word problems without distractors. Dominant bilinguals ⫽ bilingual speak-
ers who have high proficiency in only one language; weak bilinguals ⫽
bilingual speakers who have deficits in both languages. Error bars indicate
standard deviation.
This pattern of results is partly consistent with the findings of
previous research (e.g. Marian & Fausey, 2006; Saalbach et al.,
2010; Spelke & Tsivkin, 2001). Only students from the dominant
bilingual group performed better when the instructional language
and the testing language matched. It is interesting that students in
the low-proficiency group tended to show higher scores when the
problems were presented in Turkish rather than German. In other
words, our results suggest that the effect of instructional language
on bilinguals’ performance in mathematical problem solving is
mediated by the proficiency in that language: Bilinguals with a
high command of German did significantly better on problems
presented in German than on problems presented in Turkish, while
students with a weak command of German tended to be better in
Turkish. The latter result might reflect bilingual students’ tendency
to use their native language in order to compensate for deficits in
the instructional language when solving word problems.
Discussion
In the present research, we examined the relation between
bilingualism and school-related learning. In particular, we aimed
to disentangle the possible costs and benefits of bilingualism with
respect to learning mathematics for students with an immigrant
background. To this end, we compared their achievement on
mathematical word problems with that of their monolingual peers.
First of all, our study replicated previous findings on the im-
portant role of language proficiency for children’s learning. In
contrast to previous studies, however, where language assessment
was mostly realized by self-reports or via the assessment of read-
ing comprehension, we used a measure that directly tapped (oral)
language proficiency. Regression analyses with the sample of
bilingual students revealed a close relation between proficiency in
German and performance on mathematical word problems in that
language. In fact, language proficiency explained as much vari-
ance as arithmetic skills did and far more variance than was
explained by cognitive ability, SES, and reading comprehension.
The contribution of reading comprehension was rendered insignif-
icant when our measure of language proficiency was entered in
Model 3 of the regression analysis. In sum, students who were
rather proficient in German were much more likely to solve
mathematical word problems in German correctly than students
with poor German language skills. A similar pattern appeared for
Turkish word problems as a criterion. Again, proficiency in this
language (students’ native language) was the crucial predictor. An
additional effect of the instructional language (German) was not
found for mathematics problem solving in students’ native lan-
guage. Accordingly, we cannot presume a generalized effect of
learning processes in German to solution processes in Turkish.
Rather, as we will describe, a differentiated picture of cognitive
costs when switching between the language of instruction and the
Second, our data indirectly support the assumption that bilin-
proficient in both languages and who are from a lower socioeco-
nomic background (as indicated by the district where their schools
were located and our measure of SES). Although bilinguals
showed substantially lower German language proficiencies than
monolinguals, they performed as well as their monolingual peers
on problems with distractors. In contrast, as expected from the
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM
regression analyses, bilinguals’ achievement on ordinary word
problems was significantly lower than that of monolinguals. This
suggests that the costs associated with word problems that required
executive control skills (problems with distractors) were less
prominent for bilinguals than for monolinguals.
The Role of Language Proficiency in Learning
Mathematics
Our data confirm the important role of language proficiency in
school-related learning established by prior research in general and
in solving mathematical word problems, in particular. As de-
scribed in the introduction, the influence of language proficiency
on problem solving in mathematics may be threefold: language-
based knowledge transfer for mathematical procedures acquired in
the classroom, knowledge of relational terms and logical operators
during the modeling process, and linguistic comprehension of the
information in the tasks on a surface level. In the present study, the
role of language cannot be traced back to comprehension difficul-
ties on a surface level as we used only very frequently used words
and simple grammatical structure. Rather, our findings suggest that
the difficulties shown by children with poor skills in the instruc-
tional language may be due to their difficulties in understanding
the content of prior classroom instruction and in constructing a
mathematical model of the problem at hand. Previous research has
usually focused on bilingual students’ difficulties following class-
room instructions (e.g., Elbers & de Haan, 2005). Our data confirm
the importance of the mastery of instructional language for math-
ematical problem solving. However, our results also suggest that
this importance may be, at least partly, attributed to the role of
language during the modeling of the problem at hand reflected by
a significant influence of the language of testing. In other words,
language proficiency affects mathematical achievement on both
the level of instruction and the level of testing. In the future,
researchers should therefore be aware that testing low-proficiency
bilingual students in their instructional language may lead to
underestimation of their competencies and thus testing in their
native language may be more appropriate. This is also supported
by our finding that students with a weak command of the instruc-
tional language have the tendency to show better performance
when tested in their native language.
Overall, our findings make a strong case for improving bilingual
children’s language skills in the language of instruction. Thus,
efforts to make language assessment obligatory before schooling
seem to be a step in the right direction, provided these assessments
are followed by appropriate training programs (see Fried, 2008).
Cognitive Costs of Bilinguals
An exploratory analysis revealed that bilinguals with a high
proficiency in one of the languages (i.e. dominant bilinguals)
scored higher when the language of instruction and the language of
testing were the same (i.e., German) than when they differed (i.e.,
Turkish word problems), which partly confirms previous findings
on the cognitive costs that accompany the switching between
languages of instruction and language of retrieval (e.g., Spelke &
Tsivkin, 2001). However, bilingual students with a weak command
of German (and Turkish) tended to show better performance when
tested in Turkish than when tested in German. Our results thus
557
indicate that, given a low proficiency in German, these students
may find that the Turkish language functions as a resource in the
mathematical modeling process, where the students’ native lan-
guage may support more adequate mathematical modeling.
We should mention that in the group of weak bilinguals, stu-
dents’ language proficiency in both the native and the instructional
language may be considered deficient. Since these students nev-
ertheless performed better when tested in Turkish, we can presume
that further factors such as motivation or the frequency of aca-
demic language use in Turkish may have influenced their perfor-
mance on Turkish word problems. This interpretation would be in
line with results of our regression analyses on Turkish word
problem in Research Question 1. Overall, the effects of language
switching in general and the role of the native language for
students of low language proficiency in particular should be in-
vestigated in more detail.
Cognitive Benefits of Bilinguals
Bilingualism in immigrant children does not come with costs
only. In our study, we indirectly replicated findings that have
shown the cognitive benefits of bilingualism in terms of bilinguals’
executive control skills (see Adesope et al., 2010; Bialystok, 2009;
Bialystok et al., 2009), even for a sample of rather weak bilinguals.
As analyses of students’ socioeconomic background, their cogni-
tive ability, and their arithmetic skills showed, the group of bilin-
guals was comparable to the group of monolingual children on
these measures while their language proficiency was far below
(⬎1 SD). Our results showed that immigrant children with a
substantially lower proficiency in German than that shown by their
monolingual peers performed equally well as their monolingual
peers on tasks designed to test executive control while attaining
significantly lower scores than their monolingual peers on ordinary
word problems. These findings are quite in line with the conjec-
tures put forward by Bialystok (1988; 1999). Thus, even nonbal-
anced forms of bilingualism have the potential to promote cogni-
tive benefits. As has been assumed in previous research, it is not
the level of language proficiency as such but most likely the
routine of keeping apart two language systems by frequent usage
and exposure that promote improved executive control skills. This
is supported by a recent finding of cognitive gains in 7-month-old
infants who were exposed to a bilingual home environment (Kovác
& Mehler, 2009). It would thus be extremely interesting to study
whether an index of frequency of language use would be a better
predictor for cognitive benefits of bilingualism in children than a
proficiency measure. Unfortunately, this type of data was not
available in the present study. Furthermore, the strong relation
between bilinguals’ language proficiency and performance on
problems with distractors suggests that cognitive benefits of bilin-
gualism are greater in bilinguals with high language proficiency
than in those with low proficiency. It is a question for further
investigations whether and to what extent proficient bilinguals
from this student population, comparable to the balanced bilin-
guals commonly investigated in studies on cognitive benefits, can
even outperform their monolingual peers on complex mathemati-
cal problem with distracting information.
Taken together, the results of this study are promising. Our
findings emphasize the importance of language proficiency for
academic success. We are aware that due to the small sample size
558 KEMPERT, SAALBACH, AND HARDY
of the present study, broad generalizations need to be drawn with
caution. However, since the results are generally in line with
previous findings, we are certain that our findings are noteworthy
with respect to the understanding of the relation among bilingual-
ism, cognitive development, and academic performance. Further,
they may offer some practical implications. For example, our
results can be seen as a support for many Western European
countries’ current efforts to promote language training programs
for children with an immigrant background prior to school entry or
in their first years at school. Furthermore, our study confirmed the
intricate relationship between language and mathematics. Those
who design programs to promote language skills in students with
an immigrant background should therefore consider including ex-
plicit training in solving mathematical word problems, that is, in
extracting a mathematical model from a situational context (Kroes-
bergen, Leseman, Scheele, & Mayo, 2009). Finally, our findings
suggest that bilingualism in children with an immigrant back-
ground should be considered a resource rather than a burden. The
cognitive advantage of speaking two languages may compensate
for some of the disadvantages that occur with lower skills in the
instructional language as well as with low SES. Therefore, pro-
moting native language skills in addition to instructional language
proficiency is a reasonable step to take not only because bilingual-
ism can play important role in the development of a binational
identity (see also Cummins, 2004) but also because it offers the
advantage of cognitive benefits.
References
Abedi, J., & Lord, C. (2001). The language factor in mathematics tests.
Applied Measurement in Education, 14, 219 –234. doi:10.1207/
S15324818AME1403_2
Adesope, O. O., Lavin, T., Thompson, T., & Ungerleider, C. (2010). A
systematic review and meta-analysis of the cognitive correlates of bi-
lingualism. Review of Educational Research, 80, 207–245. doi:10.3102/
0034654310368803
Baumert, J., & Schümer, G. (2001). Familiäre Lebensverhältnisse, Bil-
dungsbeteiligung und Kompetenzerwerb [Family circumstances, partic-
ipation in education, and acquisition of competencies]. In J. Baumert, E.
Klieme, M. Neubrand, M. Prenzel, U. Schiefele, W. Schneider, . . . M.
Wei␤ (Eds.), PISA 2000: Basiskompetenzen von Schülerinnen und
Schülern im internationalen Vergleich [PISA (Program for International
Student Assessment) 2000: Students’ basic competencies in interna-
tional comparison]. Opladen, Germany: Leske ⫹ Budrich.
Bernardo, A. (1998). Language format and analogical transfer among
bilingual problem solvers in the Philippines. International Journal of
Psychology, 33, 33– 44. doi:10.1080/002075998400592
Bialystok, E. (1988). Levels of bilingualism and levels of linguistic aware-
ness. Developmental Psychology, 24, 560 –567. doi:10.1037/0012-
1649.24.4.560
Bialystok, E. (1999). Cognitive complexity and attentional control in the
bilingual mind. Child Development, 70, 636 – 644. doi:10.1111/1467-
8624.00046
Bialystok, E. (2001). Bilingualism in development: Language, literacy, and
cognition. Cambridge, England: Cambridge University Press. doi:
10.1017/CBO9780511605963
Bialystok, E. (2009). Bilingualism: The good, the bad, and the indifferent.
Bilingualism: Language and Cognition, 12, 3–11. doi:10.1017/
S1366728908003477
Bialystok, E., Craik, F. I. M., Green, D. W., & Gollan, T. H. (2009).
Billingual minds. Psychological Science in the Public Interest, 10,
89 –129. doi:10.1177/1529100610387084
Bialystok, E., Craik, F. I. M., & Luk, G. (2008). Cognitive control and
lexical access in younger and older bilinguals. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 34, 859 – 873. doi:
10.1037/0278-7393.34.4.859
Bialystok, E., Craik, F. I. M., & Ryan, J. (2006). Executive control in a
modified anti-saccade task: Effects on aging and bilingualism. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 32,
1341–1354. doi:10.1037/0278-7393.32.6.1341
Carlson, S. M., & Meltzoff, A. N. (2008). Bilingual experience and
executive functioning in young children. Developmental Science, 11,
282–298. doi:10.1111/j.1467-7687.2008.00675.x
Colzato, L. S., Bajo, M. T., van den Wildenberg, W., Paolieri, D., Nieu-
wenhuis, S., La Heij, W., & Hommel, B. (2008). How does bilingualism
improve executive control? A comparison of active and reactive inhibi-
tion mechanisms. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 34, 302–312. doi:10.1037/0278-7393.34.2.302
Cromdal, J. (1999). Childhood bilingualism and metalinguistic skills:
Analysis and control in young Swedish–English bilinguals. Applied
Psycholinguistics, 20, 1–20. doi:10.1017/S0142716499001010
Cummins, J. (1979). Linguistic interdependence and educational develop-
ment of bilingual children. Review of Educational Research, 49, 222–
251.
Cummins, J. (2004). Language, power and pedagogy: Bilingual children in
the crossfire. Buffalo, NY: Multilingual Matters.
Elbers, E., & de Haan, M. (2005). The construction of word meaning in a
multicultural classroom. Mediational tools in peer collaboration during
mathematics lessons. European Journal of Psychology of Education, 20,
45–59. doi:10.1007/BF03173210
Field, A. P. (2009). Discovering statistics using SPSS (3rd ed.). London,
England: Sage.
Fried, L. (2008). Pädagogische Sprachstandsdiagnostik für Vorschul-
kinder: Dynamik, Stand und Ausblick [Educational language assessment
for preschoolers: Dynamics, state, and outlook]. Zeitschrift für Erzie-
hungswissenschaft, 10 (Sonderheft 11), 63–78.
Gordon, P. (2004, October 15). Numerical cognition without words: Evi-
dence from Amazonia. Science, 306, 496 – 499.
Hakuta, K. (1999). The debate on bilingual education. Developmental and
Behavioral Pediatrics, 20, 36 –37.
Helwig, R., Rozek-Tedesco, M. A., Tindal, G., Heath, B., & Almond, P. J.
(1999). Reading as an access to mathematics problem solving on
multiple-choice tests for sixth-grade students. Journal of Educational
Research, 93(2), 113–125. doi:10.1080/00220679909597635
Kovács, A. M., & Mehler, J. (2009). Cognitive gains in 7-month-old
bilingual infants. Proceedings of the National Academy of Sciences of
the United States of America, 106, 6556 – 6560. doi:10.1073/
pnas.0811323106
Krajewski, K., Liehm, S., & Schneider, W. (2004). DEMAT 2⫹: Deutscher
Mathematiktest für zweite Klassen [DEMAT 2⫹: German Test of Math-
ematics for Grade 2]. Göttingen, Germany: Beltz Test.
Kroesbergen, E. H., Leseman, P., Scheele, A. F., & Mayo, A. Y. (2009,
August). Math talk at home and the development of pre-mathematical
skills of monolingual native and bilingual immigrant Dutch four- to
six-year-olds. Paper presented at 13th biennial conference of the Euro-
pean Association for Research on Learning and Instruction (EARLI),
Amsterdam, the Netherlands.
Lee, K., Ng, E. L., & Ng, S. F. (2009). The contributions of working
memory and executive functioning to problem solving representation
and solution generation in algebraic word problems. Journal of Educa-
tional Psychology, 101, 373–387. doi:10.1037/a0013843
Lehmann, R., & Nikolova, R. (2005). ELEMENT: Erhebung zum Lese- und
Mathematikverständnis. Entwicklung in den Jahrgangsstufen 4 bis 6 in
Berlin [ELEMENT: Investigation of reading and mathematics skills.
Development from Grades 4 to 6 in Berlin]. Berlin, Germany: Senats-
verwaltung Bildung Jugend und Sport.
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM
Lenhard, W., & Schneider, W. (2006). ELFE 1– 6: Ein Leseverständnistest
für Erst- bis Sechstklässler [ELFE 1– 6: A reading comprehension test
for first to sixth graders]. Göttingen, Germany: Hogrefe.
Marian, V., & Fausey, C. (2006). Language-dependent memory in bilin-
gual learning. Applied Cognitive Psychology, 20, 1025–1047. doi:
10.1002/acp.1242
Martin-Rhee, M., & Bialystok, E. (2008). The development of two types of
inhibitory control in monolingual and bilingual children. Bilingualism:
Language and Cognition, 11, 81–93. doi:10.1017/S1366728907003227
Miller, K. F. (1996). Linguistic structure and cognitive development:
Chinese/English comparisons. Studies in the Linguistic Science, 26,
237–254.
Miller, K. F., Smith, C. M., Zhu, J., & Zhang, H. (1996). Preschool origins
of cross-national differences in mathematical competence: The role of
number-naming systems. Psychological Science, 6, 56 – 60.
Müller, A. G., & Stanat, P. (2006). Schulischer Erfolg von Schülerinnen
und Schülern mit Migrationshintergrund: Analysen zur Situation von
Zuwanderern aus der ehemaligen Sowjetunion und aus der Türkei [Ac-
ademic achievement of students with migrant background: Analyses of
the situation of immigrants from the former Soviet Union and from
Turkey]. In J. Baumert, P. Stanat, & R. Watermann (Eds.), Herkunfts-
bedingte Disparitäten im Bildungswesen Differenzielle Bildungsproz-
esse und Probleme der Verteilungsgerechtigkeit. Vertiefende Analysen
im Rahmen von PISA 2000 [Background-associated disparities in the
educational system: Extended analyses of PISA (Program for Interna-
tional Student Assessment) 2000] (pp. 223–235). Wiesbaden, Germany:
VS Verlag für Sozialwissenschaften.
Muñoz-Sandoval, A., Cummins, J., Alvarado, C. G., & Ruef, M. R. (1998).
Bilingual Verbal Ability Tests. Itasca, IL: Riverside.
Organisation for Economic Cooperation and Development. (2006). Where
immigrant students succeed: A comparative review of performance and
engagement in PISA 2003. Paris, France: Author.
Pasolunghi, M. C., Cornoldi, C., & De Liberto, S. (1999). Working
memory and intrusions of irrelevant information in a group of specific
poor problem solvers. Memory & Cognition, 27, 779 –790. doi:10.3758/
BF03198531
Ramm, G., Prenzel, M., Heidemeier, H., & Walter, O. (2005). Soziokul-
turelle Herkunft und Migration im Ländervergleich [Socio-cultural and
migration background in the Federal States of Germany]. In M. Prenzel
(Ed.), PISA 2003: Der zweite Vergleich der Länder in Deutschland. Was
wissen und können Jugendliche? [PISA (Program for International Stu-
dent Assessment) 2003: Second comparison between the federal states
of Germany. Which knowledge and competencies do teenagers pos-
sess?] (pp. 269 –298). Münster, Germany: Waxmann.
Reusser, K. (1997). Erwerb mathematischer Kompetenzen: Literaturüber-
blick [The acquisition of mathematical skills: Overview of the litera-
ture]. In F. E. Weinert, & A. Helmke (Eds.), Entwicklung im Grund-
schulalter [Development in elementary school age] (pp. 141–155).
Weinheim, Germany: Beltz.
Riley, M. S., & Greeno, J. G. (1988). Developmental analysis of under-
standing language about quantities and solving problems. Cognition and
Instruction, 5, 49 –101. doi:10.1207/s1532690xci0501_2
Riley, M. S., Greeno, J. G., & Heller, J. H. (1983). Development of
children’s problem solving ability in arithmetic. In H. P. Ginsburg (Ed.),
The development of mathematical thinking (pp. 153–200). New York,
NY: Academic Press.
Rodriguez-Fornells, A., de Diego Balaguer, R., & Münte, T. F. (2006).
Executive control in bilingual language processing. Language Learning,
56, 133–190. doi:10.1111/j.1467-9922.2006.00359.x
Saalbach, H., Eckstein, D., & Grabner, R. (2010, June). Bilingual learning
and mathematical knowledge representation in the brain. Poster session
presented at 1st symposium of the European Association for Research on
Learning and Instruction (EARLI) Special Interest Group on Neurosci-
ence and Education. Zurich, Switzerland.
559
Schnepf, S. V. (2004). How different are immigrants? A cross-country and
cross-survey analysis of educational achievement (IZA Discussion Pa-
per No. 1398). Bonn, Germany: IZA [Institute for the Study of Labor].
Schwippert, K., Bos, W., & Lankes, E.-M. (2004). Heterogenität und
Chancengleichheit am Ende der vierten Jahrgangsstufe im interna-
tionalen Vergleich [International comparison of heterogeneity and
equality of opportunities at the end of grade four]. In W. Bos, E.-M.
Lankes, M. Prenzel, K. Schwippert, G. Walther, & R. Valtin (Eds.),
Erste Ergebnisse aus IGLU: Schülerleistungen am Ende der vierten
Jahrgangsstufe im internationalen Vergleich [First results from the
Progress in International Reading Literacy Study (PIRLS): Students’
school achievement at the end of grade four in international compar-
ison] (pp. 265–302). Münster, Germany: Waxmann.
Schwippert, K., Hornberg, S., & Goy, M. (2008). Lesekompetenzen von
Kindern mit Migrationshintergrund im nationalen Vergleich [National
comparisons of reading skills in children with a migrant background]. In
W. Bos, E.-M. Lankes, & K. Schwippert (Eds.), IGLU-E 2006: Die
Länder der Bundesrepublik Deutschland im nationalen und internation-
alen Vergleich [International Primary School Reading Literacy Study
National Extension (IGLU-E) 2006: The federal states of Germany by
national and international comparison] (pp. 111–125). Münster, Ger-
many: Waxmann.
Shaftel, J., Belton-Kocher, E., Glasnapp, D. R., & Poggio, J. P. (2006). The
impact of test item characteristics on the performance of English lan-
guage learners and students with disabilities on large-scale mathematics
assessments. Educational Assessment, 11, 105–126. doi:10.1207/
s15326977ea1102_2
Spelke, E. S., & Tsivkin, S. (2001). Language and number: A bilingual
training study. Cognition, 78, 45– 88. doi:10.1016/S0010-0277
(00)00108-6
Stanat, P. (2008). Heranwachsende mit Migrationshintergrund im
deutschen Bildungswesen [Adolescents with a migrant background in
the German educational system]. In K. S. Cortina (Ed.), Das Bildung-
swesen in der Bundesrepublik Deutschland: Strukturen und Entwicklun-
gen im Überblick [The German educational system: Overview of struc-
tures and development] (pp. 685– 673). Reinbek bei Hamburg,
Germany: Rowohlt-Taschenbuch-Verl.
Stanat, P., Rauch, D., & Segeritz, M. (2010). Schülerinnen und Schüler mit
Migrationshintergrund [Students with migrant background]. In E. Kli-
eme, C. Artelt, J. Hartig, N. Jude, O. Köller, M. Prenzel, . . . P. Stanat
(Eds.), PISA 2009: Bilanz nach einem Jahrzehnt [Balance after one
decade] (pp. 200 –228). Münster, Germany: Waxmann.
Staub, F. C., & Reusser, K. (1995). The role of presentational structures in
understanding and solving mathematical word problems. In C. A.
Weaver, S. Mannes, & C. R. Fletcher (Eds.), Discourse comprehension.
Essays in honor of Walter Kintsch (pp. 285–305). Hillsdale, NJ: Erl-
baum.
Stern, E. (1998). Die Entwicklung des mathematischen Verständnisses im
Kindesalter [The development of mathematical competencies in early
childhood]. Berlin, Germany: Pabst Science.
Stern, E. (2009). The development of mathematical competencies: Sources
of individual differences and their developmental trajectories. In W.
Schneider, & M. Bullock (Eds.), Human development from early child-
hood to early adulthood: Evidence from the Munich Longitudinal Study
on the Genesis of Individual Competencies (LOGIC) (pp. 221–237).
Mahwah, NJ: Erlbaum.
Stern, E., & Lehrndorfer, A. (1992). The role of situational context in
solving word problems. Cognitive Development, 7, 259 –268. doi:
10.1016/0885-2014(92)90014-I
Verhoeven, L., & Aarts, R. (1998). Attaining functional biliteracy in the
Netherlands. In A. Durgunoglu, & L. Verhoeven (Eds.), Literacy devel-
opment in a multilingual context (pp. 111–133). Mahwah, NJ: Erlbaum.
Verschaffel, L., Greer, B., & De Corte, E. (2002). Everyday knowledge and
mathematical modeling of school word problems. In K. Gravemeijer
560 KEMPERT, SAALBACH, AND HARDY

(Ed.), Symbolizing, modeling and tool use in mathematics education (pp.
257–276). Dordrecht, the Netherlands: Kluwer Academic.
Walter, O. (2008). Herkunftsassoziierte Disparitäten im Lesen, der Math-
ematik, und den Naturwissenschaften: Ein Vergleich zwischen PISA
2000, PISA 2003, und PISA 2006 [Background-associated disparities in
reading, mathematics, and the natural sciences: A comparison of PISA
(Program for International Student Assessment) 2000, PISA 2003, and
PISA 2006]. Zeitschrift für Erziehungswissenschaft, 10, 149 –168.
Walter, O., & Taskinen, P. (2007). Kompetenzen und bildungsrelevante
Einstellungen von Jugendlichen mit Migrationshintergrund in
Deutschland: Ein Vergleich mit ausgewählten OECD-Staaten [Compe-
tencies and educationally relevant attitudes of adolescents with migrant
background in Germany: A comparison of selected OECD (Organisation
for Economic Co-operation and Development) nations]. In M. Prenzel
(Ed.), PISA 2006: Die Ergebnisse der dritten internationalen Ver-
gleichsstudie [PISA (Program for International Student Assessment)
2006: Results of the third international comparison] (pp. 337–366).
Münster, Germany: Waxmann.
Weiß, R. H. (1998). Grundintelligenztest Skala 2: CFT 20 [Basic Intelligence
Test 2: Culture Fair Intelligence Test 20]. Göttingen, Germany: Hogrefe.
Woodcock, R. W. (1991). Woodcock Language Proficiency Battery—
Revised. Rolling Meadows, IL: Riverside.

Appendix

Mathematical Word Problems With and Without Distractors (English Translations)

Type of word problem Without distractors With distractors

Exchange problems

Final quantity unknown
Exchange quantity unknown
Initial quantity unknown
Comparison problems
Sabine had 3 euros. Then Philipp gave
her 5 euros. How many euros does
Sabine have now?
Julia had 2 crayons. Then Tobi gave her
some crayons. Now Julia has 9
crayons. How many crayons did Tobi
give her?
At first Maria had some marbles. Then
she gave 2 marbles to Hans. Now
Maria has 6 marbles. How many
marbles did Maria have at first?
Maria had 3 marbles. Together these marbles cost
90 cents. Then Hans gave here 5 marbles.
These cost 1.50 euros. How many marbles
does Maria have now?
Lisa had 2 euros. Then Kai gave her some euros.
Lisa wants to buy herself 8 crayons. Now Lisa
has 9 euros. The crayons cost 10 euros
altogether. How many euros did Kai give her?
Tina had some pieces of chewing gum. Then she
gave 2 pieces of chewing gum to Karl. One
pack of chewing gums costs 80 cents. Now
Tina has 6 pieces of chewing gum. She wants
to buy herself 4 new packs. How many pieces
of chewing gum did she have at first?

Quantity of difference unknown
Quantity of comparison unknown
Quantity of reference unknown
Complex comparison problem
Beate has 5 stickers. Max has 8 stickers.
How many stickers does Max have
more than Beate?
Maria has 3 crayons. Hans has 4
crayons more than Maria. How many
crayons does Hans have?
Lisa has 9 balloons. She has 4 balloons
more than Lennart. How many
balloons does Lennart have?
Meike has 5 marbles. She is 11 years old. Tom
has 8 marbles. He is 4 years older than Meike.
How many marbles does Tom have more than
Meike?
Julia is 5 years old. She weights 25 kilo. Tobi is
4 years older than Julia. He weights 45 kilo.
How old is Tobi?
Maria has 9 marbles. She gets 12 euros pocket
money. Maria had 4 marbles more than Hans.
Hans gets 10 euros pocket money. How many
marbles does Hans have?

Three-person comparison
Meike has 2 euros. Tim has 5 euros
more than Meike. How many euros
do Meike and Tim have together?
Claudia has 5 playing balls. She has 5
playing balls less than Thomas.
Thomas has 2 playing balls more than
Oliver. How many playing balls does
Oliver have?
Maria has 2 playing balls. They cost 10 euros
altogether. Hans has 5 playing balls more than
Maria. They cost 15 euros altogether. How
many playing balls do Hans and Maria have
together?
Lena has 3 euros. She is 8 years old. Karl is 2
years older than Lena. Lena has 4 euros less
than Karl. Karl has 3 euros more than Timo.
How many euros does Timo have?

(Appendix continues)
 COGNITIVE BENFITS AND COSTS OF BILINGUALISM 561

Appendix (continued)

Type of word problem Without distractors With distractors

Complex combination problem

Total quantity unknown and Paul and Tina have 6 euros together.
subset unknown Tina has 4 euros. Kai and Nina have
9 euros altogether. Nina has 3 euros.
How many euros do Paul and Kai
have together?
Kai and Lena have 6 drops together. Lena has 4
drops. Both children are 12 years old. Karl and
Sabine have 9 drops together. Sabine has 3
drops. Karl and Sabine are 2 years younger
than the others. How many drops have Kai and
Karl together?

Received April 8, 2010

Revision received March 8, 2011
Accepted March 18, 2011 䡲

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at http://notify.apa.org/ and you will be
notified by e-mail when issues of interest to you become available!

View publication stats