Aim: There is little data on whether or not a bilingual upbringing may aggravate specific language problems in children. This study analysed whether there was an interaction of such problems and simultaneous bilingualism.
Methods: Participants were 5- to 7-year-old children with specific language problems (LANG group, N = 56) or who were typically developing (CONTR group, N = 60). Seventy-three children were Swedish–Finnish bilingual and 43 were Swedish-speaking monolingual. Assessments (in Swedish) included tests of expressive language, comprehension, repetition and verbal memory.
Results: Per definition, the LANG group had lower scores than the CONTR group on all language tests. The bilingual group had lower scores than the monolingual group only on a test of body part naming. Importantly, the interaction of group (LANG or CONTR) and bilingualism was not significant on any of the language scores.
Conclusions: Simultaneous bilingualism does not aggravate specific language problems but may result in a slower development of vocabulary both in children with and without specific language problems. Considering also advantages, a bilingual upbringing is an option also for children with specific language problems. In assessment, tests of vocabulary may be sensitive to bilingualism, instead tests assessing comprehension, syntax and nonword repetition may provide less biased methods.
Wechsler Preschool and Primary Scale of Intelligence-Revised
partial eta squared
• Simultaneous bilingualism does not aggravate specific language problems but may result in a slower development of vocabulary both in children with and without specific language problems.
• Considering also advantages, a bilingual upbringing is a valid option also for children with specific language problems.
• In assessment, tests of vocabulary may be sensitive to bilingualism, instead tests assessing comprehension, syntax and nonword repetition may provide less biased methods.
Growing up in a bilingual family may provide a child with the advantage of acquiring two languages simultaneously from the start of language development. This situation could be expected to impose increased demands on the acquisition of vocabulary and linguistic structures, as compared to the demands of a monolingual development. Yet, typically developing children do not suffer any major disadvantage from being simultaneous bilingual, with the exception of a distributed vocabulary and some cross-language influences (1–4). The situation may be more challenging for children with a primary or specific language impairment (SLI). This condition affects particularly language development and is not explained by general developmental delay, neurological disorder, autism spectrum disorder or significant hearing problems. Problems may occur in expressive language, comprehension, syntax and morphology as well as in verbal memory and learning (1,5–8). Various terms have been applied to this condition, but SLI is the most frequent in the literature (9). The specificity of the disorder has, however, been questioned, because subtle nonlinguistic processing weaknesses, for example, in fine motor functions and cognitive processes may accompany language problems (7–9). The term ‘language impairment’ (LI) does not exclude nonlinguistic problems (7,8) or the possibility that a stated LI may partly be explained by bilingualism (10).
Although simultaneous bilingualism is not uncommon few studies have been undertaken concerning a possible risk of a bilingual upbringing for children with SLI or LI. Most studies have focused on or included children who are sequential bilinguals. These children learn a second language (L2) after the first language (L1) is fairly well established although not completely acquired (3,7,8,11). Generally, an increased risk of being classified as LI has been stated in these bilingual children from age 6 onward. Also, performance on language tests has been depressed compared with those of monolingual children (4,12,13).
Canadian bilingual (French–English) and monolingual children have provided populations suitable for studies on simultaneous bilingualism. No difference in grammatical morphology has been found between monolingual and simultaneously bilingual children with SLI (14). However, when focus is beyond morphological or morpho-syntactic aspects of language learning, lower language scores has been found in SLI children exposed at home with two or more languages than in SLI children from monolingual homes (15).
Westman et al. (16) compared children who scored below the 20th percentile on a language screening, considered at risk for LI, with a control group who scored above this threshold. The children were monolingual or simultaneously bilingual from the Swedish-speaking community in Finland. Bilingualism negatively affected two of six verbal tests, that is, body part naming and short-term memory for sentences. However, this effect was comparable for the children at risk for LI and the nonrisk children. Thus, a bilingual upbringing was not more detrimental to children at risk for LI.
The present study was an extension of the study by Westman et al. and partly used the same data. Besides being a larger study, it also differed from the Westman et al. study in that participants with specific language problems were selected to meet criteria defined in terms of cut-off test scores, instead of simply belonging to the 20% with the poorest scores on a screening. The aim was to see whether simultaneous bilingualism would aggravate language problems when these were more strictly defined. The term ‘specific language problems’ was used as SLI was not verified through diagnostic language assessments.
Materials and Methods
The study included 5- to 7-year-old Finnish kindergarten or preschool students (in Finland, children start formal schooling the year they reach 7 years of age). Assessments were undertaken in Swedish. Candidates to the group with specific language problems (LANG group) were recruited from three sources. First, children with ICD 10 (17) diagnoses indicating SLI were identified from patient files of four clinics with multi-professional services for Swedish-speaking children and invited to assessments.
Owing to scarce diagnostic services for Swedish-speaking children in Finland, many children with SLI are not identified. Therefore, population-based recruitment was also undertaken. The screening just mentioned and described by Westman et al. (16) was undertaken. The children who scored below the 20th percentile on the screening were invited to further assessments.
A third recruitment was undertaken to further increase the number of participants. A questionnaire was distributed to 1511 parents of 5- to 6-year-old children attending Swedish-speaking kindergartens and preschools. Parents of 538 children returned the questionnaire. Children with reported difficulties with language development were invited to assessments. Those with only articulation problems were not considered.
Children with no reported concern regarding development were randomly selected to a control group (CONTR group) from the same kindergartens from which children with language problems were recruited and assessed.
Operational criteria for the group with language problems (LANG group) were defined to correspond to general definitions of SLI (5–9), but medical diagnostic criteria for language disorders (5,17) were not strictly applied. A specific discrepancy between nonverbal intellectual capacity and language level (6,17,18) was not required as this rule has been questioned (7–9). Comprehensive language tests were not used to verify language disorders as no such test had been translated into Swedish and standardized (19). Criteria for the LANG group were the following:
1 A nonverbal reasoning within the average range, that is, a mean score on nonverbal subtests of Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R) (20) ≥ 7 (≥−1 SD);
2 Below-average scores, that is, scores < 7 (<−1 SD) on at least three of six verbal subtests of the A Developmental Neuropsychological Assessment (NEPSY; Table 1). This criterion is based on, but somewhat stricter than, the test manuals’ cut-off for language problems (21);
3 No significant general developmental delay, neurological disorder, autism spectrum disorder or hearing problems.
Table 1. Descriptions of the NEPSY subtests used in the study
Body part naming
This subtest is aimed to assess one aspect of the child’s expressive vocabulary. The child names the body parts of a depicted child or as pointed at on his/her own body (e.g. ‘chin’, ‘heel’)
This subtest aims to assess the child’s ability to access and produce familiar words under time pressure. The child is asked to rapidly name squares and circles by size (small or large), colour (five basic colours) and shape (square or circle) (e.g. ‘..big blue circle..’). Number of correct responses and time were combined to form the total subtest score.
Comprehension of instructions
This subtest aims to assess the child’s ability to process and respond to verbal instructions. The child is asked to follow sequences of circles or crosses in different colours on a sheet according to increasingly complex instructions (e.g. ‘Point to all the crosses and then to a red circle’)
Repetition of nonsense words
This subtest is aimed to assess phonological decoding, encoding and articulation of nonsense words of increasing length and phonological complexity, presented on audiotape. The nonsense words consist of two to five syllables and are wordlike. The child listens to and repeats each word (e.g. ‘bue-lex-tiss’).
This subtest aims to assess the ability to comprehend, recall and narrate a story. The child listens to the story, tells it again (free recall condition) and responds to questions about omitted details (cued recall condition)
Free recall of an item yields 2 points and cued recall 1 point
This subtest aims to assess the child’s ability to repeat verbatim sentences of increasing length and complexity
The criteria for the CONTR group were the following:
1 A mean score on the verbal and nonverbal WPPSI-R subtests within the average range;
2 No signs of specific language problems in recruitment 2 and 3, and not belonging to recruitment 1;
3 No significant general developmental delay, neurological disorder, autism spectrum disorder or hearing problems.
To qualify as monolingual, the children were to come from entirely Swedish-speaking families and to speak only Swedish with parents or other significant adults in the home, and with siblings and peers, as reported by the parents. Bilingual Swedish–Finnish children were to come from families where one adult spoke Swedish with the child and another spoke Finnish with the child. Children who were reported to speak better Finnish than Swedish were not included, but children who, according to parental rapport, spoke Swedish and Finnish equally well were included.
After application of the criteria, the LANG group included 56 children. Eighteen of the children came from clinics (source one), 23 from the screening (source two) and 15 children had been recruited through questionnaires (source three). After the selection with respect to background variables and application of criteria, the CONTR group included 60 children. Seventy-three children were Swedish–Finnish bilingual and 43 were Swedish-speaking monolingual.
Tests and questionnaire
An abbreviated form of the WPPSI-R was used to estimate the children’s cognitive ability. The subtests Information, Arithmetic, Block Design and Picture Completion were administered. The subtests had high correlations with total IQ in the Swedish WPPSI-R standardization sample (r = 0.65–0.77) (20). Mean subtest scores were calculated for verbal and nonverbal subtests (range 1–19, M = 10, SD = 3).
Verbal capacities were assessed with the Swedish standardized version of the NEPSY, as its successor NEPSY II (22) was not yet available in Swedish. The NEPSY and NEPSY II are neuropsychological tests and widely used by psychologists in Northern Europe and are standardized in many other countries, including the US, France, the Netherlands and Italy. They assess several domains of development, including language and memory. Subtests may be selected according to the child’s problems. The subtest body part naming was originally designed for children aged 3–4 years but is also suitable for somewhat older children. The subtests are described in Table 1.
The language background of the child was assessed using questionnaires encompassing the following topics: Which is the child’s stronger language; languages used by mother and by father, including other significant adults at home, to communicate with each other and with the child; and languages used by the child to communicate with parents, siblings and peers. For these questions, parents checked the appropriate alternative: Finnish, Swedish, Equal/Both (as appropriate) and Other. Parents were also asked to report the child’s possible medical diagnoses and their own education. The latter was classified into Finnish equivalents to the following categories: primary; lower secondary; upper secondary or baccalaureate; academic (tertiary) degree (23). Owing to missing data for fathers only maternal education was used. Finally, parents to children from source three were asked if the child had presented problems with language development, with space provided for description.
Assessments took place in kindergartens, preschools or clinics in one session of about 1.5 h, with a break if necessary. Examiners were psychologists with experience of child neuropsychological assessment or trained master’s students of psychology. Parents obtained a written summary of their child’s result.
The study plan was approved by the Ethical Committee of the Helsinki University Central Hospitals, the Hospital for Children and Adolescents and by the Department of Psychology at Åbo Akademi University and was followed accordingly. At least one parent of each participating child gave their informed consent in writing prior to assessments.
Data preparation and statistical analyses
For statistical calculations, rather than using the Swedish NEPSY norm scores, which are based on percentile ranges, the raw scores were converted to a standard scale (range = 0–20, M = 10, SD = 3) using the distributions of the CONTR group [(3 × child’s subtest raw score − 3 × CONTR group mean raw score + SD × 10)/SD]. A correction for age was entered into the formula based on regression coefficients expressing the expected effect of age on each subtest except the body part naming, which did not correlate with age. Data were checked for distributions, ceiling effects and outliers. Missing NEPSY subtest scores (n = 5) were replaced with the individual’s mean score. Negative values were replaced with 0.00.
Cross-tabulation using the chi-squared test and Student’s t-tests for independent samples were undertaken to compare groups with respect to background variables. Repeated-measures MANOVA (group × bilingualism × six NEPSY language scores) was undertaken to compare the groups’ test profiles, followed by two-way ANOVA’s (group × bilingualism) with each NEPSY language scores at a time as the dependent variable. To control for background influences, corresponding MANCOVA and ANCOVA’s were then undertaken with background variables as covariates. Partial Eta Squared (η2p) was used to indicate effect size. The α-level of significance was set at 0.05. Two-tailed probability values were used for significance testing.
Distributions of independent and background variables
In the LANG group, 39 children were bilingual and 17 were monolingual. In the CONTR group, 34 children were bilingual and 26 children were monolingual. The distributions of bilingual and monolingual children did not differ significantly between groups.
The LANG group and the CONTR group did not differ with respect to gender. Maternal education was significantly lower in the LANG group, t(103) = 2.9, p = 0.004. Mean age was on average 3.2 months lower in the LANG group compared with the CONTR group, t(88.14) = 3.19, p = 0.002. As expected, the WPPSI-R verbal mean score of the LANG group was significantly lower than that of the CONTR group, t(101.35) = 9.01, p < 0.001. The LANG and CONTR groups also differed significantly with respect to mean nonverbal WPPSI-R scores, t(114) = 3.14, p = 0.002, to the advantage of the CONTR group (Table 2).
Table 2. Gender, age, maternal education and Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R) mean score distributions of participants when subdivided according to two classifying variables: group (LANG and CONTR) and bilingualism (bilingual and monolingual)
LANG group (N = 56)
CONTR group (N = 60)
n.s., not significant; y, year; m, month.
Age (range 5 y 1m–7y 4m)
Maternal education (classes 1–4)
Bilingual (N = 73)
Monolingual (N = 43)
Age: years, months
When dividing participants into monolingual and bilingual groups, the groups did not differ with respect to gender or with respect to maternal education. Mean age was on average 2.3 months lower in the bilingual group compared with the monolingual group, t(114) = 2.15, p = 0.034. The groups did not differ significantly with respect to WPPSI-R verbal or performance scores (Table 2).
Language profiles of groups
The test profiles according to group and bilingualism are illustrated in Figure 1. On the repeated-measures MANOVA, the between-groups effect of group on the NEPSY subtests was, by definition, very significant, F1,112 = 215,21, p < 0.001, η2p = 0.66. The LANG group scored significantly lower than the CONTR group on all subtests: body part naming, t(114) = 7.97, p < 0.001; speeded naming, t(114) = 9.74, p < 0.001; comprehension of instructions, t(114) = 7.82, p < 0.001; repetition of nonsense words, t(101.46) = 5.98, p < 0.001; narrative memory, t(114) = 8.89, p < 0.001; and sentence repetition, t(114) = 9.89, p < 0.001. The between-groups effect of bilingualism was not significant. On ANOVA, the bilingual group scored significantly lower than the monolingual group only on body part naming, t(114) = 4.44, p < 0.001.
Importantly, the interaction of group and bilingualism on the six NEPSY scores, taken together (repeated-measures MANOVA) was not significant (η2p = 0.011), nor were interaction effects on the separate subtest scores significant on subsequent ANOVA’s. Thus, simultaneous bilingualism did not affect the scores of the LANG group proportionally more than those of the CONTR group. When entered as a covariate in the main profile analysis, age did not affect the scores significantly. Nor was the effect of maternal education significant when entered as a covariate.
To verify that nondiagnosed children (n = 38) were comparable to children with ICD 10 diagnoses suggesting SLI (n = 18) a one-way repeated-measures MANOVA (subgroup according to diagnosis vs. no diagnosis × six NEPSY subtest scores) was undertaken with bilingualism as a covariate. The groups did not differ significantly (η2p = 0.041). This indicated that the total group was comparable to children who had diagnoses suggesting SLI.
The question of whether or not a bilingual upbringing places a particular burden on the language development of children with primary language problems has remained under-studied. We assessed monolingual and simultaneously bilingual children, who either had specific language problems or were typically developing, in their stronger language, Swedish. A comprehensive assessment was used that tapped expressive vocabulary, speeded naming, comprehension of instructions, repetition of nonsense words, memory for and producing a narration and memory for sentences. The interaction of language skills (LANG group or CONTR group) and bilingualism was not significant. Bilingual children displayed poorer scores on body part naming, but this effect was found in both typically developing bilinguals and in bilingual children with language problems. Bilingualism did not affect any other language subtest significantly. This finding indicates that simultaneous bilingualism does not affect children with language problems proportionately more than it does typically developing children.
The Swedish-speaking minority in Finland is optimal for studies on simultaneous bilingualism. It has its own day-care centres, kindergartens and schools, television and radio channels. Thus, some children may be entirely Swedish-speaking, without knowledge of Finnish, until they start mandatory studies of Finnish in school at 9 years of age. Yet, linguistically mixed families are more common. Many such families choose to offer their children the opportunity to become bilingual from the start. A widely adopted practice is that each parent consistently speaks only his/her own maternal language to the child. Swedish and Finnish are linguistically very different, but monolingual and bilingual families are similar with respect to other demographic variables, such as parental education.
Our results are comparable to those of Paradis et al. (14) who reported no added negative effects of bilingualism in children with SLI on a test of syntax. The present study assessed other aspects of language and provides a complement to the study by Paradis et al. The findings contradict those by Cheuk et al. (15) who found that a bi- or multilingual home significantly increased the risk of SLI.
The contradiction might be related to differences in how children were enrolled and which statistic tests were used in these two studies. However, the subjects in the Cheuk study were very young. It cannot be ruled out that the adverse effects of bi- or multilingualism could be expressed only during the first years of life and then disappear.
Future studies should definitely clarify the interaction of bilingualism and language problems on children with different ages and with various measures of vocabulary (receptive and expressive).
The present study was a continuation of the study by Westman et al. (16). It was therefore expected that the results would correspond to each other. The present study, however, included larger study groups. Further, the LANG group was selected to correspond to definitions of SLI. Stricter criteria were applied that were operationalized in terms of test score cut-offs, instead of representing merely children who obtained the lowest 20% scores on a language screening.
Assessing and diagnosing bilingual children with language problems is a complicated issue as language tests may be sensitive also to bilingualism. This may lead to miss-diagnosing bilingual children as LI or SLI. In the present study, one of six subtests used as criteria for language problems, the body part naming subtest, was sensitive to bilingualism, independent of group (LANG or CONTR). It is therefore theoretically possible that the classification of children into the LANG group may partly reflect bilingualism. According to our results, the proportions of bilingual children did not differ significantly between the LANG and CONTR groups, although the distribution of bilingual and monolingual children on the groups appeared somewhat skewed. Further, five subtests of six did not show a significant effect of bilingualism. If some children from the bilingual LANG group were false positives and bilingual rather than had language problems, then the bilingual LANG group would have been expected to perform stronger than the monolingual LANG group on other subtests than the body part naming subtest. This was, however, not the case (Fig. 1).
It should be noted that the children were selected based on operational criteria, not diagnoses suggesting SLI. Owing to scarceness of speech and language services in Swedish in Finland, part of the children were identified through a screening or a questionnaire. Nor were the children assessed using a diagnostic, comprehensive language test, as no such test had been translated to Swedish and standardized at the time of this study (18). The NEPSY (21,22) is a comprehensive and standardized neuropsychological assessments intended to analyse children’s problems. In the diagnostic assessment of a child with suspected SLI, the NEPSY is, however, not sufficient. Although assessment protocols and methods vary in Finland and Sweden (19,24), a speech and language specialist should always be involved. Moreover, Bishop and McDonald (25) demonstrated that results from language tests and parental report do not always match; therefore, both should be used in the diagnostic process. The operational criteria for the LANG group were set to allow also relatively mild cases of language problems. Still, the present LANG group was representative of children with diagnoses indicating SLI, as there was no difference between diagnosed and undiagnosed LANG cases with respect to language performance level.
Proficiency in both Swedish and Finnish was assessed by questionnaires and not objective tests. All bilingual children were, however, reported by their parents to speak both Swedish and Finnish and both languages were consistently spoken to the child at home. Monolingual children spoke only Swedish and were not exposed to Finnish at home or in kindergarten. All children attended Swedish-speaking preschool or kindergarten.
The question of the effects of simultaneous bilingualism in children with language problems has practical and clinical relevance. Owing to the uncertainty concerning the effects of bilingualism, it is not uncommon that parents whose native languages are different are advised to choose only one language when communicating with a child with language problems (1). The present results indicate that bilingualism does not overtax these children’s language capacity. Still, the fact that simultaneously bilingual children from both groups did obtain slightly poorer scores on the body part naming subtest could be seen as a reason to protect children with language problems from this negative effect, using only one language at home. However, this might restrict the interaction of a parent with the child. Further, bilingual children also have an asset in acquiring two languages, which both are practiced by family and relatives. Their vocabulary may be seen as distributed over two languages rather than being restricted. Thus, we concur with Paradis et al. (1,14) that a bilingual upbringing is an option also for children with primary language problems and SLI.
A further implication of the findings is that bilingualism may add to the likelihood that children are misdiagnosed as being language impaired if assessments rely heavily on measures tapping vocabulary. Tests of comprehension, syntax (14), speeded naming, nonword repetition, narrative memory and working memory may provide less biased methods. The NEPSY verbal subtests, with the exception of the body part naming subtest, do not seem to be sensitive to bilingualism and are thus suitable as parts of the assessment of verbal abilities in bilingual children with language problems.
We gratefully acknowledge the financial support of the Signe and Ane Gyllenberg Foundation and the Åbo Akademi University Foundation as well as the help of MPsy Petra-Ann Salminen, MPsy Anu Haavisto, MPsy Pernilla Landen and MPsy Raija-Leena Holmberg in data acquisition. We are particularly indebted to MA Pekka Lahti-Nuuttila for help and advice with statistical calculations.