Factors influencing bilingual expressive vocabulary size in children with autism spectrum disorders

Research in Autism Spectrum Disorders 8 (2014) 1079–1089

Contents lists available at ScienceDirect

Research in Autism Spectrum Disorders Journal homepage: http://ees.elsevier.com/RASD/default.asp

Factors influencing bilingual expressive vocabulary size in children with autism spectrum disorders Catherine Hambly *, Eric Fombonne 1 Oregon Health & Science University, Institute on Development & Disability, 707 SW Gaines Street, Portland, OR 97239, United States

A R T I C L E I N F O

A B S T R A C T

Article history: Received 11 December 2013 Received in revised form 26 May 2014 Accepted 28 May 2014

This study explored bilingual exposure, language, social impairment and cognitive factors that could influence second language (L2) expressive vocabulary size as measured on the MacArthur–Bates Communicative Development Inventories (various languages) in 33 children (mean age = 60 months) diagnosed with ASD. In the 23 children with L2 vocabularies, recent language exposure estimates accounted for 69% of the variation in L2 vocabulary size, and the VABS-II expressive scale score explained an additional 13% of the difference. The complete sample was then subgrouped into three levels of L2 vocabulary size to compare children with no L2 vocabularies (NON-B, n = 10), low L2 word counts (LOW-B, n = 11) and high L2 counts (HIGH-B, n = 12), as determined by a median split procedure. The HIGH-B group had significantly larger L1 vocabularies than both the LOWB (p = .045) and the NON-B (p = .003) groups, and higher VABS-II expressive scale scores than both the LOW-B (p = .008) and the NON-B (p = .012) groups. Social impairment did not significantly differ across groups and cognitive impairment did not preclude the development of L2 vocabularies. Expressive bilingualism in this population appears related to high levels of recent direct L2 exposure in combination with stronger dominant language abilities. ß 2014 Elsevier Ltd. All rights reserved.

Keywords: Autism Language Vocabulary Bilingualism Child

1. Introduction A growing body of research has already shown that bilingual exposure does not negatively impact language development for children with a variety of diagnoses associated with language-learning delays including specific language impairment (Korkman et al., 2012; Paradis, Crago, Genesee, & Rice, 2003), Down syndrome (Feltmate & Kay-Raining Bird, 2008; KayRaining Bird et al., 2005), and ASD (Hambly & Fombonne, 2012; Ohashi et al., 2012; Petersen, Marinova-Todd, & Mirenda, 2012; Valicenti-McDermott et al., 2013). These studies also provide data to show that some children with significant impairments are acquiring two languages in childhood. The development of bilingualism in children with ASD, who present with social impairments often accompanied by language and cognitive delays, has only been minimally discussed in the research examining the effects of bilingual exposure. No study has examined the factors influencing the acquisition of second language vocabularies in children with ASD. Published research shows that children with ASD have a wide range of bilingual outcomes in early childhood. Seung, Siddiqi, and Elder (2006) reported a case study of a child with ASD who went from monolingual single word use to bilingual

* Corresponding author at: 14 Gulf Lane, Galveston, TX 77550, United States. Tel.: +1 409 939 4937/503 494 8364. E-mail addresses: [email protected], [email protected] (C. Hambly). 1 Tel.: +1 503 494 8364. http://dx.doi.org/10.1016/j.rasd.2014.05.013 1750-9467/ß 2014 Elsevier Ltd. All rights reserved.

1080

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

sentence use across a two year period from ages 3–5. Kay-Raining Bird, Lamond, and Holden (2012) described a sample of children aged 2–22 years of age where 36% of bilingually exposed children were reported to ‘‘speak two languages equally well’’: the exact levels reached varied with the subjects’ age since language levels were rated on a 1–5 scale relative to ‘‘a fluent adult or same-age peers.’’ Petersen et al. (2012) provided data on vocabulary size for 14 English–Chinese bilinguals aged 43–73 months with homogenous exposure histories. Hambly and Fombonne (2012) found that 62% of the bilingually exposed children aged 38–76 months in their sample spoke words in a second language, but noted that only a few of them were reported to have bilingual phrase use. Since bilingual acquisition is clearly possible for some children with ASD, it is clinically important to learn what factors contribute to the development of bilingual language abilities so that families can make evidence-based decisions about encouraging bilingualism in their young child with ASD. Studies of children with typical development and children with language impairment suggest that two main factors could contribute to bilingual outcomes for children with ASD: amount of exposure to the input languages and level of proficiency in a first language. Additionally, since no studies of bilingualism have been done in populations characterized by social impairments, the level of social impairment must also be explored as a factor. First, current language exposure appears to relate more strongly to bilingual vocabulary size than does the timing of onset, such as from infancy vs. a later time in early childhood. Bedore et al. (2012) found in a very large sample of 1029 Spanish–English bilingual pre-kindergarten students with a range of bilingual proficiencies that current language exposure accounted for more of the variance in language dominance than the age of first exposure to a language. David and Wei (2008) reported a significant correlation between language exposure and vocabulary size, and found that the presence of translation equivalent word pairs related to language exposure with the most balanced lexicon found in children with more balanced input. In a study of 191 young children with a mean age of 59 months, Hammer et al. (2012) reported that children’s vocabulary scores in each language were related to their exposure to and their usage of each language. A small number of reports suggest that this relationship between input and bilingual acquisition also holds for children with language impairments. Feltmate and Kay-Raining Bird (2008) suggested that current frequency of second language exposure could be a key factor in explaining levels of acquisition across two languages for the children with Down syndrome in their study, and they noted that the length of exposure over time did not predict any measure of second language ability. Gutie´rrez-Clellen, Simon-Cereijido, and Sweet (2012) found that levels of English use predicted differences in English acquisition in bilingual children with specific language impairment (SLI). Factors other than overall exposure need to be taken into account when describing bilingual environments. The communicative purpose of the language input also appears to play a role in second language outcome in typical development, with direct speech to the child playing a larger role than indirect speech: Oller (2010) found in a detailed case study using all-day recordings that language spoken directly to a child had dramatically more effect on vocabulary size than language that was overheard but not directed to him. Hoff et al. (2011) also discussed the communicative function of bilingual input (e.g., indirect exposure vs. child-directed speech) as well as the sources of input such as the one speaker: one language model vs. the one speaker: two languages model. Place and Hoff (2011) studied a large sample of children with naturally occurring bilingual exposure and found that unique sources of variance in children’s language proficiency were due to exposure factors including the number of different speakers providing input and the native-level language proficiency of the speakers. Baron-Cohen and Staunton (1994) demonstrated a related situation in a sample of older children with ASD raised by mothers who spoke to them in accented, non-native English: children with ASD were more likely than their nonASD siblings to retain their mothers’ accents. These children may have been more influenced by direct caregiver language models than by indirect exposure from others in their home or from their peers at school. These studies on bilingual input suggest that bilingualism in an ASD population should focus on estimates of direct speech to the child, rather than overall estimates of language use in the home, because there may be large amounts of second language use that are not directed to the child with ASD. Both Kremer-Sadlik (2005) and Yu (2013) discussed the phenomenon of children being raised as monolinguals in a bilingual home. Hambly and Fombonne (2012) also found that mothers of bilingually exposed children frequently spoke a non-native language to their child, which indicates that there may have been sources of L2 exposure indirectly in the home from overheard conversations between parents and other adults or siblings. In addition to language exposure variables, children’s language learning abilities clearly play a role in second language acquisition: children who have less severely impaired language skills are more likely to be more proficient in a second language than their peers with more severe language impairments. Verhoeven, Steenge, and van Balkom (2012) studied Turkish–Dutch bilingual children ages 7–11 with specific language impairment and found that second language proficiency levels could be explained by first language proficiency levels, even when controlling for working memory and non-verbal intelligence. Gutie´rrez-Clellen et al. (2012) also found a similar relationship in their 8 month longitudinal study of English– Spanish preschoolers with specific language impairment: the children’s Spanish first language skills were a key factor predicting English acquisition. These early studies of bilingual children with specific language impairment at different points in childhood suggest that there appears to be a close and enduring relationship between first and second language abilities in the context of language impairment that should also be found in children with ASD. Some parents of children with ASD have expressed concerns that children with lower cognitive skills should only be exposed to one language (Kay-Raining Bird et al., 2012); and some parents and clinicians may believe that bilingual word use requires more advanced cognitive skills than required for monolingual word use. It is important to note that signs of

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

1081

bilingual word use can be detected in cognitively impaired children with mental ages less than 36 months, as in Kay-Raining Bird et al. (2005). Studies in typical development show early bilingual word use from the earliest stages of language development (see discussion in De Houwer, Bornstein, and De Coster, 2006), when children’s cognitive development permits the mapping of a word to an object. Bialystok (2001) suggests that early word learning is ‘yoked’ to cognitive development (p. 26), with words mapped onto existing concepts or cognitive structures. Cognitive impairment should not preclude the development of bilingual vocabularies when a child has the capacity to map words to objects, although it may impact developing more advanced language skills. So, children with severely impaired language, as already seen in the population of children with Down syndrome, may be able to learn words in two languages, mapping two different words to the same concept. Children with ASD show a wide range of cognitive profiles, with cognitive levels distributed across both the low and high ends of measures. In monolingual children with ASD, expressive vocabulary size is related to both age and non-verbal mental age (Luyster, Lopez, & Lord, 2007). Replication of this study in a bilingual population would be challenging because the variability introduced by exposure factors adds to the challenge of recruiting large numbers of bilingually exposed children with ASD at different cognitive levels. One methodological solution was used by Kjellmer, Hedvall, Fernell, Gillberg, and Norrelgen (2012) when they classified children with ASD aged 24–63 months by cognitive level into one of three groups: intellectual disability, learning problems/developmental delay, and normal intelligence. Using these groupings, they found cognition and age explained 71% of the variance in spoken vocabulary size in monolinguals. In bilinguals, this grouping approach could be used to identify the presence of second language vocabularies in children at different levels of cognitive impairment. Finally, the contribution of social impairments to second language acquisition needs to be addressed since social deficits are not part of the diagnostic profile of previously studied samples of children with specific language impairment or Down syndrome. Social impairments could influence bilingual acquisition in multiple ways: social impairments could result in reduced opportunities to directly interact with peers or family (Baron-Cohen & Staunton, 1994; Kremer-Sadlik, 2005; Yu, 2013), potentially resulting in large amounts of indirect exposure to a second language but little bilingual acquisition. At a more fundamental level, the development of some bilingual word-learning strategies may depend on children’s level of social awareness and interaction. For example, typically developing bilingual children have an enhanced ability compared to monolinguals in using speaker factors as well as object characteristics to assign new labels to words (Brojde, Ahmed, & Colunga, 2012). We do not know if young bilingually exposed children with ASD develop this same speaker-supported word learning mechanism or other socially mediated language-learning strategies. Studies of bilingualism in ASD have not looked for relationships between levels of social ability and levels of second language outcomes. Neither Petersen et al. (2012) nor Ohashi et al. (2012) presented any data on the levels of social impairment of their participants. Hambly and Fombonne (2012) did not find differences between bilingual exposure groups on two measures of socialization, but they did not relate social ability to bilingual outcomes. Social abilities as measured on broad tools can make a small contribution to vocabulary size in monolinguals above and beyond age and cognitive level (Kjellmer et al., 2012), so it is possible that social abilities could also impact on second language acquisition. This study will explore the hypotheses that second language vocabulary size in bilingually exposed children with ASD will be associated with the amount of direct bilingual exposure received as well as with children’s dominant language skills. The contributions of social abilities and cognitive level will also be explored.

2. Method 2.1. Participants The 33 subjects for this study were selected out of a larger database of children with ASD from Quebec (94%) and Ontario (6%). More than half of the participants (64%) were recruited directly from the Montreal Children’s Hospital ASD clinic via mailings or during clinical visits, with the remaining 36% recruited via publicity in community organizations and service providers. All subjects had an informant caregiver who could read and speak English or French well in order to complete the interview and questionnaires and all the children’s spoken vocabularies could be measured using an available language version of the main study questionnaire. The children were ages 3 and 7 years and had been diagnosed with an ASD prior to enrolling in the study: the participants from Quebec were diagnosed by child psychiatrists or developmental pediatricians; the participants from Ontario were diagnosed at hospitals or at a university-based ASD research unit. Subjects were included in this study on the basis of any history of bilingual exposure. This exposure criterion acknowledges the highly heterogeneous nature of bilingual exposure and fits with the study’s goal of identifying factors— including the role of varying amounts of language exposure—that influence the development of second language skills. This decision also permitted us to recruit a relatively large number of participants with ASD and a history of bilingual exposure within a small geographic region. Children also needed to have a spoken vocabulary of at least 50 words in a language. The 50 word benchmark was used so that participants were at the level—in at least one language—where phrases could be expected to emerge based on data from typically developing monolinguals (Fenson et al., 1993). This benchmark was set slightly lower than the 100 word minimum

1082

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

used in Kay-Raining Bird et al. (2005) in order to include the largest number of children for whom bilingual phrase acquisition could theoretically be present. 2.2. Measures Measures of general language and social abilities were selected to be appropriate for the anticipated population of dominant French and English-speaking children and were already available in French translations approved for research use. French or English questionnaires and interviews were used according to each family’s needs. 2.2.1. ASD diagnostic measures Caregivers were asked to sign releases allowing the research team to obtain information on their child’s ASD diagnosis. Diagnostic reports were obtained for 97% of all participants, with the Autism Diagnostic Observation Schedule-Generic (Lord et al., 2000) used in at least 73% of participants’ assessments. In the sample used for this study (n = 33), 22 children were diagnosed with ‘autistic disorder’ and two participants had ‘Autism Spectrum Disorder’ listed in their reports without further classification. However, based on information (such as ADOS scores) within the reports, we grouped these children together within the ‘‘autistic disorder’’ category (n = 24). One child diagnosed with Asperger syndrome was grouped with the pervasive developmental delay—not otherwise (PDD-NOS) Specified category (n = 9; 8 diagnosed as PDD-NOS, 1 diagnosed with Asperger syndrome). 2.2.2. Cognitive measures Caregivers were given the option to sign releases allowing the research team to review children’s cognitive test reports, if available. All available reports were reviewed: a variety of measures were used, but reports typically included a descriptive term summarizing the child’s overall functioning that could be used to classify the child’s level of cognitive functioning per Kjellmer et al. (2012). Children with terms including ‘‘above average,’’ ‘‘average,’’ and ‘‘low average’’ cognitive functioning were grouped together, as were children with ‘‘mild delay’’ or ‘‘below average’’ descriptors. Children with terms including ‘‘moderate delay,’’ ‘‘extremely low,’’ and ‘‘severe delay’’ were grouped together. Three children had cognitive testing reports that did not provide a usable summary term or test scores, and the remaining fourteen children had not been tested yet or had reports that could not be accessed. The data from the sixteen children with known cognitive functioning levels are presented by the three groupings described above. 2.2.3. Bilingual exposure: language environment interview All families participated in a language environment interview (LEI) that collected data on home language environment history and generated language exposure estimates for each 6 month period of the child’s life and produced an overall estimate. The LEI questions were administered by a trained research assistant during a 30 min phone interview with an informant, typically the child’s mother. The informant provided a detailed history of all caregivers’ amount and duration of care for each six month period of the child’s life. The informants also described the languages these caregivers used with other family members (to identify indirect exposure sources) and with the child (to quantify the amount of direct exposure). Language exposure information was gathered for daycare or school environments where relevant: the informant was asked what languages the educators spoke directly to the child as well as what languages were used when the educators spoke to the group of children. Data for each caregiver was summarized for each six month period of time. Each caregiver’s language input was weighted using a scale based on the relative amounts of direct (e.g., one-to-one) caregiver–child communication. For example, input from a parent with a full-time job received a lower weight than input from a parent with primary inhome childcare responsibilities. The input scores resulted in a summary exposure estimate for each six month period since birth. The final lifetime ratio (LR) represents a composite average exposure (e.g., 75% French/25% English) in direct caregiver–child interactions; it does not reflect the amount of passive exposure via indirect/overheard conversations in the home or community. Questions and scoring for the LEI are described in detail in Hambly and Fombonne (2012). 2.2.4. Categorizing bilinguals’ language dominance Languages are categorized as first vs. second languages based on their historical usage data on the LEI, with the child’s ‘‘L1’’ being the language used most often spoken to the child across his lifetime and the ‘‘L2’’ the language spoken to the child less often. As noted in Hambly and Fombonne (2012), many children with ASD experience changes in their bilingual environments in early childhood so it can be misleading to categorize bilingual exposure based on what language they heard ‘first’ (a time of onset definition). Additionally, many children with ASD will not fully develop even one language during the preschool years, so if a child with delayed language begins hearing a second language after age three or four we cannot clearly say that they are learning a second language ‘after’ learning a first language; the terminology from studies of typically developing bilinguals would be misleading. For these reasons, we refer to L1 and L2 based on overall levels of exposure, which coincided with the children’s actual language levels in all but two cases. The two exceptions were children who had balanced bilingual exposure for at least the last 12 months’ period (48%/52% and 54%/46% exposure to the L1/L2 respectively) and whose word counts across the two languages were also relatively equal (302 L1/370 L2 words

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

1083

and 412 L1/586 L2 words). The 31 other children had a clear dominant language based on both exposure history and vocabulary size data. Six children had some limited exposure and spoken language skills in a third language (L3); this language was heard less often and the children’s spoken vocabulary sizes in the third language were lower than in their L1 and L2. The sample of trilinguals was too small for analyses of factors related to third language acquisition. 2.2.5. Language diary Families completed a week-long language diary by charting the child’s location (e.g., home, daycare), communication partners (e.g., parents, siblings) and language exposure estimates throughout each day for a week. Parents estimated their language usage at five different points in each day (e.g., early morning, morning, early afternoon, late afternoon, evening) and these estimates were summarized into a weekly average. These data were collected to corroborate information on recent (past 6 month) estimate of language use collected on the LEI; estimates were expected to be similar but not identical due to the different methodologies and since the LEI focused only on language used directly with the child while the language diary did not ask parents to make a distinction between direct and indirect language usage across the day. 2.2.6. Family Background Information Questionnaire and list of services Caregivers completed a Family Background Information Questionnaire (FBIQ) which elicited marital status, household income (detailed in units of $10,000 up to the ceiling level on the questionnaire of ‘‘>$80,000’’), highest level of education obtained, employment status, and ethnic/cultural heritage for parent caregivers. Educational level was later recoded into a binary variable (education above or below a university coursework level). Due to the diversity of ethnicities represented, we followed Leadbitter and Hudry (2009) in categorizing participants’ ethnic/cultural heritage responses as ‘Caucasian’ vs. ‘nonCaucasian’. A ‘List of Services’ elicited treatment program histories including intensive behavioral interventions, speechlanguage, occupational, and physical therapy, and descriptions of other therapies. Intensive behavioral service descriptions were subsequently categorized as ‘no services’ vs. ‘some services’. 2.2.7. Adaptive language and social behavior The Vineland Adaptive Behavior Scales-Second Edition (Sparrow, Cicchetti, & Balla, 2005) measures skills exhibited in everyday life. The receptive and expressive communication subdomains were appropriate to administer to bilingually exposed participants since lower-level items do not directly reference language-specific content and since all participants with phrase speech had English or French as their dominant language; the measure was administered only once. The Interpersonal subdomain of the VABS-II was selected as a variable of interest due to its focus on interpersonal relationships and social skills. Test-retest reliability for the VABS-II has been established: subdomain reliability coefficients are excellent with most values exceeding .85 (Sparrow et al., 2005). 2.2.8. Social responsiveness The Social Responsiveness Scale (SRS) measures the severity of ASD symptoms within children’s natural environments (Constantino, 2002); it is appropriate for use in children with ASD aged 4 years and older. Since the items measure the degree of social impairment, it is used in this study as an indicator of children’s social skills alongside the VABS-II Interpersonal scale. Scoring procedures and standardization data followed the technical guide (Constantino & Gruber, 2005). 2.2.9. Vocabulary measures The MacArthur Communicative Development Inventory: Words and Sentences (MCDI) (Fenson et al., 1993) is a wellstudied parent report measure of vocabulary; it is designed for children aged 16–30 months, but can be used without normative data with older developmentally delayed children. Total expressive vocabulary on the MCDI is correlated with observational data for children with and without language impairment (Fenson et al., 1993), and for both monolingual and bilingual typically developing children (Patterson, 2000). MCDI language adaptations (Dale, Fenson, & Thal, 1993) used in this study included English (Fenson et al., 1993), Quebec French (Frank, Poulin-Dubois, & Trudeau, 1997), Spanish (JacksonMaldonado et al., 2003), Hebrew (Maital, Dromi, Sagi, & Bornstein, 2000), Chinese (Wu, 1997), and Romanian (Geangu & Benga, 2006). Children with bilingual exposure histories during the initial screening call were sent all relevant versions of the MCDI (e.g., English, French, and Spanish forms if all these languages were heard in the home). The MCDI allows counts of vocabulary items to create a total vocabulary score in each language. Concepts can also be mapped across languages to identify translation-equivalent pairs. 2.3. Procedures All interested families were screened by phone for eligibility then mailed a questionnaire package including one or more language versions of the MCDI as well as the SRS, Language Diary, FBIQ, List of Services, consent forms for participation and release of diagnostic and cognitive reports, and return postage. The Language Environment Interview was administered by phone after enrollment. Administration procedures and reliability indices are described in Hambly and Fombonne (2012).

1084

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

Table 1 Sample characteristics.

% Male % With autistic disorder % Received some behavioral intervention Mother: % speaking a majority language to their child Mother: % non-Caucasian Mother: % Canadian-born Mother: % with some university education

Family income ($1000) Age (months, SD) Age of first parental concerns (months, SD) Age at diagnosis (months, SD)

NON-B (n = 10)

LOW-B (n = 11)

HIGH-B (n = 12)

p

80% 80% 50% 70% 30% 60% 80%

91% 64% 45% 91% 45% 73% 73%

75% 75% 50% 75% 27% 58% 75%

Chi-square .604 .685 .896 .465 .628 .742 .925

64.0 60.3 19.7 35.4

(20.0) (12.6) (9.0) (8.8)

60. 5 (17.4) 59.6 (11.8) 21.6 (15.3) 42.0 (16.4)

74.5 60.4 19.8 37.9

(15.7) (8.4) (9.3) (11.2)

ANOVA .193 .984 .905 .485

2.4. Bilingual grouping Three levels of bilingual outcomes were established for the 33 bilingually exposed children because a wide range of L2 vocabulary counts was found. Ten of the 33 children did not have any second language vocabulary usage; they were grouped as non-bilinguals (NON-B). The remaining 23 children had second language expressive vocabularies ranging from 2 words to 559 words, with a median L2 vocabulary of 70 words. The wide range of vocabularies combined with a low median count pointed to different levels of expressive bilingual achievement: we chose to split the 23 bilingual word users into two separate groups based on the median second language vocabulary score of 70, a median-split procedure used in other studies of bilingualism (Goldstein, Bunta, Lange, Rodriguez, & Burrows, 2010). Children with second language vocabularies from 1 to 69 words were classified as low bilinguals (LOW-B; n = 11) and those with vocabularies from 70 to the maximum score were classified as high bilinguals (HIGH-B; n = 12). 2.4.1. Sample characteristics by groups The sample characteristics of the three groups are presented in Table 1, and no statistical differences were found among the groups on variables such as age (current, at the time of diagnosis, or when first concerns were noted), intervention history, or maternal characteristics that would indicate any fundamental differences among the groups. Each group contained children from a variety of bilingual language combinations, with bilingual French–English environments the most common occurrence across groups. Children with no second language vocabulary (NON-B) were exposed to bilingual French and English exposure (60%), bilingual exposure with French or English and a minority language (30%), and trilingual exposure with French, English and a minority language (10%). LOW-B children were divided among bilingual French and English exposure (73%), trilingual exposure with French, English and a minority language (18%), and bilingual exposure with French or English and a minority language (9%). HIGH-B children were divided among bilingual French and English exposure (58%), trilingual exposure with French, English and a minority language (33%), and bilingual exposure with French or English and a minority language (8%). 2.5. Statistical analyses Only 23 children out of the 33 bilingually exposed participants had any second language vocabulary, so two types of analyses were done. First, correlates of second language vocabulary size were examined in the 23 children with L2 vocabularies using Pearson correlation coefficients and regression analysis. Then, we examined group differences in the complete sample among participants with no L2 vocabularies (NON-B), low L2 vocabularies (LOW-B) and high L2 vocabularies (HIGH-B) to identify factors that could explain the three different levels of bilingual outcomes. Group differences on categorical variables were examined using Chi-square tests of independence. Group differences on continuous variables were analyzed using one-way analysis of variance (ANOVA). Statistical significance was set at the p < 05 level. Post hoc testing using the least significant difference test was performed on any variable with a p value of <0.5, and effect sizes were calculated using Cohen’s d. No statistical tests were used in discussion of cognitive groupings as the sample sizes in each group were too small due to limited number of participants with cognitive test data. Statistical analyses were performed using SPSS 14. 3. Results 3.1. The relationships among second language vocabulary size, dominant language abilities, and second language exposure This study hypothesized that the second language expressive vocabulary size would relate to how much direct second language exposure children received and to their overall expressive language abilities in their dominant language. To test the

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

1085

Table 2 Correlates of bilingual outcomes (n = 23 children). L2 vocabulary size Exposure: LEI—last 6 months L1 vocabulary VABS-II expressive scale scores VABS-II receptive scale scores

p

.828 .280 .605 .170

.000 .195 .005 .460

relationship among L2 vocabulary size and the factors of exposure and dominant language abilities, we examined the subset of 23 children who had spoken words in a second language (Table 2). First, we correlated four factors with L2 vocabulary size. We correlated the VABS-II receptive language scores with L2 vocabulary size but found no significant relationship. We examined L1 expressive vocabulary size and overall adaptive expressive language scores on the VABS-II: only overall expressive language scores on the VABS-II—not vocabulary size in the L1—were significantly correlated with second language vocabulary size (r = .605, p = .005). Finally, we correlated the recent exposure variable (as measured on the LEI by reporting the percentage of exposure to the child’s dominant language in a six month period): a significant correlation was found between smaller amounts of recent L1 exposure (e.g., higher amounts of L2 exposure) and larger L2 vocabulary counts (r = .828, p = .000). Finally, we entered the LEI recent exposure variable and the VABS-II expressive language scores into a stepwise regression analysis: recent language exposure as measured on the LEI accounted for 69% of the variation in L2 vocabulary size, and the VABS-II expressive scale score explained an additional 13% of the difference. This supports our hypothesis that both exposure and dominant expressive language abilities play a role in second language vocabulary development and provides data to show that exposure to a second language may play a determining role in second language vocabulary acquisition in this population. The children without L2 vocabularies (the NON-B group) were not included in the correlation and regression analyses, so these findings do not explain why some children with exposure histories and adequate L1 expressive language skills did not develop any L2 word usage. We conducted additional analyses to explore language and exposure factors among all three outcome groups. 3.2. Exposure analyses In order to understand the differences in exposure history that could contribute to the development of L2 vocabularies, we performed a series of ANOVA tests on lifetime vs. recent exposure data across the three bilingual outcome groups. Lifetime bilingual exposure was not significantly different among groups. The recent exposure to L2 as measured using the last 6 months’ period on the LEI showed a trend toward difference (F(2, 30) = 2.58, p = .093) among groups on ANOVA testing: the NON-B and HIGH-B children received on average 25% and 27% of their recent language input in their L2, but the LOW-B group only received about 6% exposure to the L2. The language diary, a measure designed to complement the data on the LEI recent exposure variable, also showed a similar pattern of children in the LOW-B group receiving on average the least amount of second language exposure although group differences that did not reach statistical significance. As an additional exploratory measure, we examined historical trends in the exposure data for the NON-B, LOW-B and HIGH-B groups. Patterns of language ratios over time as measured by the LEI showed that children in the HIGH-B group had second languages added to their environments (e.g., they were enrolled in second language/bilingual childcare or an adult caregiver began speaking the language to them at home) but they did not ‘lose’ existing languages from their environments. In contrast, nearly half of the other children (five in each of the LOW-B and NON-B groups) experienced language loss based on caregivers no longer speaking directly to them in the L2. It should be noted that since the LEI measures only direct communication to the child, some of the children who have ‘lost’ a language per the LEI may still receive some exposure in everyday overheard conversations or learn isolated L2 words from their families, childcare workers or from TV (Table 3). 3.3. Dominant language ability and social impairment analyses Language variables were also examined across all three groups to explore why bilingual outcomes varied and some children with bilingual exposure did not use any L2 words. First, expressive vocabulary size in the children’s dominant

Table 3 Exposure measures by bilingual vocabulary groups.

Lifetime ratio on language environment interview (LEI): % dominant language exposure LEI: % last 6 months Language diary: % current exposure

NON-B (n = 10)

LOW-B (n = 11)

HIGH-B (n = 12)

p

84.6 (12.7)

79.6 (16.6)

72.3 (18.0)

.208

75.7 (36.6) 77.6 (32.0)

94.3 (9.5) 88.9 (8.0)

73.0 (20.0) 68.7 (20.9)

.093 .107

1086

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

Table 4 Language and social measures by bilingual vocabulary groups.

Expressive vocabulary and language MCDI: words in L1

NON-B (n = 10)

LOW-B (n = 11)

HIGH-B (n = 12)

p

LSD post hoc p

366 (176)

440 (160)

569 (103)

.010

NON-B vs. HIGH-B = .003 LOW-B vs. HIGH-B = .045 NON-B vs. LOW-B = ns NON-B vs. HIGH-B = .012 LOW-B vs. HIGH-B = .008 NON-B vs. LOW-B = ns

VABS-II expressive scale scores

10.5 (2.0)

10.5 (2.6)

13.8 (2.9)

.012

Receptive language VABS-II receptive scale scores

10.6 (3.4)

12.7 (2.4)

13.5 (2.4)

.089

Social measures SRS T score VABS-II Interpersonal scale scores

81.4 (19.1) 11.1 (2.4)

74.5 (12.5) 10.6 (2.0)

69.1 (12.9) 12.4 (2.2)

.177 .192

languages on the MCDI were compared. ANOVA testing revealed significant group differences (F(2, 30) = 5.41, p = .01). Post hoc testing revealed that the HIGH-B group had significantly larger L1 vocabularies than both the LOW-B (p = .045) and the NON-B (p = .003) groups. The HIGH-B group had vocabulary sizes that on average were near the ceiling for the MCDI, with a mean score larger than the 50th percentile for the highest age range in the normative sample. Their full vocabulary size was likely underestimated by this tool and this means that the actual difference between the HIGH-B group and the other two groups may be larger than the MCDI could measure. There was no significant difference between the L1 vocabulary size of the NON-B and LOW-B groups (Table 4). The VABS-II expressive scale scores were considered next, since this measure was not subject to the ceiling effect seen on the MCDI. ANOVA testing found a significant difference (F(2, 25) = 5.33, p = .012) among groups, and post hoc testing showed that the HIGH-B group scored significantly higher than both the LOW-B (p = .008) and the NON-B (p = .012) groups. The HIGH-B group scores fell in the ‘‘adequate’’ range of adaptive functioning, whereas children in the other two groups averaged scores in the ‘‘moderately low’’ range. No significant differences between the LOW-B and NON-B groups emerged that could explain the different levels of second language vocabulary use. Language comprehension was also examined. VABS-II receptive scale score differences did not reach a level of statistical significance (p = .089) on ANOVA testing in this small sample although the average score of the HIGH-B children placed them in the ‘‘adequate’’ range on this measure while the average scores of the other two groups fell in the ‘‘moderately low’’ range. Levels of social impairment and responsiveness were examined. No significant differences among groups were found. All three groups had mean scores that fell in the moderately low range of adaptive functioning on the VABS-II. 3.4. Cognition Finally, we explored whether cognitive impairments would prevent the development of second language vocabularies. Children across the three cognitive groupings demonstrated large L1 expressive vocabularies, and notably all three cognitive groups—including those with moderate-severe impairments—had children who were acquiring second language vocabularies. The moderate–severe impairment grouping included a child who spoke 200 L2 words and had nearly 30% of named concepts labeled with translation-equivalent word pairs: this child was clearly developing second language abilities despite the reported cognitive limitations. The other two children with L2 vocabularies in this group had only 2% and 6% of their concepts named bilingually, but their exposure data also showed significantly decreased exposure to a second language over time. Children in the mild impairment group showed a range of L2 vocabulary size (7–559 words) and a range of translation-equivalent pairs (1–66% of concepts): some children with mild impairments are clearly acquiring a second language. Overall, these data demonstrate bilingual vocabulary development in children with ASD in the context of measured significant cognitive impairment (Table 5).

Table 5 Language variables by cognitive groupings.

MCDI: words in L1 MCDI: words in L2 VABS-II expressive scale scores VABS-II receptive scale scores

Average or better (n = 5)

Mild impairment (n = 7)

Moderate–severe impairment (n = 4)

Missing (n = 17)

533 (164) 39 (39) 11 (1.9) 14.4 (2.9)

522 (148) 155 (228) 12.8 (1.9) 13.2 (2.6)

417 (182) 86 (100) 8.7 (1.5) 12 (1.7)

431 (172) 200 (188) 12.0 (3.4) 11.5 (2.9)

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

1087

4. Discussion Several findings emerge from this study. First, the amount of recent direct language exposure plays a critical role in the acquisition of second language vocabulary in the subset of children who had some L2 vocabulary usage, as seen in the HIGHB group. The presence of 10 children with no L2 vocabulary despite second language exposure histories and some ongoing indirect second language exposure demonstrates that passive exposure does not guarantee development of expressive bilingual vocabularies. Further exploration of the data revealed that children in the NON-B and LOW-B group had little to no direct caregiver communication in their L2 at the time of the study although historically they had been spoken to in both languages. Many were from bilingual homes but were experiencing language loss or were treated as monolinguals within their bilingual family. If in childcare, they were either currently enrolled in monolingual childcare sites or addressed by adult caregivers in only their dominant language. Children’s dominant language abilities contributed to second language vocabulary size in the regression analysis, and the HIGH-B children scored significantly higher than children in the other two groups on their overall VABS-II expressive language scores and their dominant language expressive vocabulary size. This finding of relatively stronger expressive language skills could explain why none of the children in the HIGH-B group experienced language loss: caregivers may not have felt it necessary to make large changes in their children’s language environments if the children were showing adequate development in expressive language even in the context of diagnosed ASD. This stands in marked contrast to the patterns of language restriction and loss experienced by children in the other two groups: it is possible that caregivers’ choices about whether or not to expose a child to bilingual input depends to some degree on the child’s level of functioning at an earlier age. Any studies of bilingualism in this population will need to be aware that selecting children based on consistent bilingual exposure since birth may bias the sample toward children with higher skills and exclude children with more significant impairments in early childhood. Levels of receptive language were not significantly different among groups. This finding may reflect the limitations of using broad, parent-report measures for studies with relatively small numbers of subjects. For example, both the NON-B and LOW-B groups’ receptive scale scores indicate that their language comprehension fell in the ‘‘moderately low’’ range, in contrast to the HIGH-B group’s functioning which was within the ‘‘adequate’’ range, but this difference did not reach the level of statistical significance. The limited number of questions used to assess receptive language may have masked real differences between the LOW-B and NON-B groups and the role of receptive language should be explored further. The lack of statistical differences in social impairment among groups was intriguing, since all three groups had VABS-II scores that placed them in the ‘‘moderately low’’ range of adaptive functioning on the Interpersonal scale. This, combined with the limited data on cognitive impairments across the groups, suggests that social and cognitive impairments did not dissuade some families from providing bilingual exposure to their children with ASD or preclude the development of bilingual vocabularies in some exposed children. However, we do not know if bilingual word users have developed any specific socially mediated word learning strategies (such as using speaker cues to map labels to objects) or have any cognitive advantage stemming from bilingual acquisition that could indicate any positive outcomes of bilingualism on social and cognitive abilities. These results suggest that—as in typically developing populations—higher levels of recent L2 exposure contribute to larger L2 vocabularies. Dominant language expressive language abilities explained a small part of the variance, but recent direct L2 exposure explained the majority of the variability in L2 expressive vocabulary size. Children in the HIGH-B group likely attained larger vocabularies than the other two groups because they had the most language exposure (both recently and over their lifetime) and the strongest language learning abilities. It is not clear from the data why children in the NON-B group did not display any L2 vocabulary use when they had similar language and social functioning to the LOW-B children and more L2 exposure. The LOW-B group presents an interpretative challenge: these children had expressive vocabulary and language skills were significantly lower than the HIGH-B group, and a lower amount of direct exposure to the L2 than the NON-B group, but they were nevertheless beginning to develop L2 vocabularies. Several explanations are possible. First, their small (e.g., 2–52 words) L2 vocabularies may be words directly taught to them by parents or teachers in a community where bilingualism is a valued outcome. Secondly, their emerging L2 vocabulary skills may be the result of their indirect exposure: children hearing an L2 in their home, in their classroom, or from media may learn some L2 words that are highly motivating or repeated often. Finally, the presence of a small L2 vocabulary does not necessarily equate with emerging bilingualism: only 4% of the LOW-B children’s total conceptual vocabularies were concepts with translation-equivalent pairs and it is quite possible that their expressive vocabularies are not differentiated in LOW-B children’s language systems at this time. In contrast, the HIGH-B children’s percentage of doublets was 36%, which is similar to the ratio found in typically developing children on the same measure (Pearson, Fernandez, & Oller, 1993). It would be important to monitor children in these language-learning environments over time to determine what factors influence the development of bilingual vocabularies, especially when exposure comes from indirect sources (e.g., overheard in the home or classroom). 5. Conclusions The preliminary clinical application of this study is that it is possible for some young verbal children with ASD to begin developing L2 expressive vocabularies when provided with direct exposure in communicative interactions. L2 expressive

1088

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

vocabulary size was significantly correlated with dominant expressive language scores, so clinicians and parents should discuss a child’s language functioning rather than discouraging bilingual exposure on the basis of the child’s ASD diagnosis or social impairments. It is also clear from the cognitive grouping data that some children with measured cognitive impairments can acquire bilingual vocabularies. The data suggest that families who want to provide their children with a basis for expressive bilingualism need to provide their children with sufficient direct L2 exposure, since the exposure-L2 vocabulary relationship described in this study is based on measures of estimated direct exposure in adult–child communication. Finally, since this study focused on spoken vocabulary in early childhood, parents and clinicians should remember that it is likely that many of the emerging bilinguals in this study (the LOW-B group) as well as some children in the NON-B group will develop L2 skills if direct communicative exposure to the L2 is provided. The conclusions of this study are limited by the need to rely on autism diagnoses from clinical reports, the relatively small sample size in each group, the lack of direct measures of exposure in the children’s environments, and the unavailability of direct standardized measures of language in children’s L1 and L2. Autism diagnoses were by necessity drawn from the children’s clinical case reports rather than through direct ascertainment. This limitation is tempered by the fact that a standardized assessment tool was used in 73% of the sample. Larger subject numbers are necessary in order to explore the interactions among language ability, social impairment and exposure levels. The limitations related to exposure measurement are common to other studies in the field, all of which rely on parental estimates of exposure either through diaries or recall. Technological advances have made it possible to capture and automatically analyze language input into differing languages of exposure in real-time (Oller, 2010); this will allow researchers to accurately calculate exposure amounts instead of relying on estimates. The limitations related to the measurement of L1 and L2 are also common to studies in this field: few standardized measures of language are available outside of English and Spanish (e.g., no measure normed on a French-speaking population was available at the time of the study). Using results from this study, parents and clinicians can begin a discussion of children’s bilingual environments and factors—like increasing the amount of direct exposure—that may promote eventual L2 acquisition and the child’s full participation in both home and community life. Further exploration of the factors influencing bilingual acquisition and the trajectories of bilingual development in children with ASDs is critical to providing better learning environments for children raised in bilingual communities or homes. References Baron-Cohen, S., & Staunton, R. (1994). Do children with autism acquire the phonology of their peers? An examination of group identification through the window of bilingualism. First Language, 14, 241–248. Bedore, L., Pen˜a, E., Summers, C., Boerger, K., Resendiz, M., Greene, K., et al. (2012). The measure matters: Language dominance profiles across measures in Spanish–English bilingual children. Bilingualism: Language and Cognition, 15, 616–629. Bialystok, E. (2001). Bilingualism in development: Language, literacy, and cognition. New York: Cambridge University Press. Brojde, C., Ahmed, S., & Colunga, E. (2012). Bilingual and monolingual children attend to different cues when learning new words. Frontiers in Psychology, 3, 155. Constantino, J. (2002). The Social Responsiveness Scale. Los Angeles: Western Psychological Services. Constantino, J., & Gruber, C. (2005). Manual: Social Responsiveness Scale. Los Angeles: Western Psychological Services. Dale, P., Fenson, L., & Thal, D. (1993). Some suggestions for the adaptation of the MacArthur Communicative Development Inventories to additional languages. Unpublished manuscript.. David, A., & Wei, L. (2008). Individual differences in the lexical development of French–English bilingual children. International Journal of Bilingual Education and Bilingualism, 11, 598–618. De Houwer, A., Bornstein, M., & De Coster, S. (2006). Early understanding of two words for the same thing: A CDI study of lexical comprehension in infant bilinguals. International Journal of Bilingualism, 10, 331–347. Feltmate, K., & Kay-Raining Bird, E. (2008). Language learning in four bilingual children with Down syndrome: A detailed analysis of vocabulary and morphosyntax. Canadian Journal of Speech-Language Pathology and Audiology, 32, 6–20. Fenson, L., Dale, P., Reznick, J., Thal, D., Bates, E., Hartung, J., et al. (1993). MacArthur Communicative Development Inventories: User’s guide and technical manual. Baltimore: Paul H Brookes Publishing Co. Frank, I., Poulin-Dubois, D., & Trudeau, N. (1997). Inventaires MacArthur du de´veloppement de la communication: Mots et e´nonce´s. Montre´al. Geangu, E., & Benga, O. (2006). Chestionarul MacArthur de Evaluare a Dezvoltarii Abilitatilor de Comunicare: Cuvinte si Propozitii. Unpublished Manuscript, ClujNapoca, Romania: Babes-Bolyai University.. Goldstein, B., Bunta, F., Lange, J., Rodriguez, J., & Burrows, L. (2010). The effects of measures of language experience and language ability on segmental accuracy in bilingual children. American Journal of Speech Language Pathology, 19, 238–247. Gutie´rrez-Clellen, V., Simon-Cereijido, G., & Sweet, M. (2012). Predictors of second language acquisition in Latino children with specific language impairment. American Journal of Speech Language Pathology, 21, 64–77. Hambly, C., & Fombonne, E. (2012). The impact of bilingual environments on language development in children with autism spectrum disorders. Journal of Autism and Developmental Disorders, 42, 1342–1352. Hammer, C., Komaroff, E., Rodriguez, B., Lopez, L., Scarpino, S., & Goldstein, B. (2012). Predicting Spanish–English bilingual children’s language abilities. Journal of Speech, Language, and Hearing Research, 55, 1251–1264. Hoff, E., Core, C., Place, S., Rumiche, R., Senor, M., & Parra, M. (2011). Dual language exposure and early bilingual development. Journal of Child Language, 39, 1–27. Jackson-Maldonado, D., Thal, D., Marchman, V., Newton, T., Fenson, L., & Conboy, B. (2003). MacArthur Inventarios del Desarrollo de Habilidades Comunicativas. User’s guide and technical manual. Brookes: Baltimore. Kay-Raining Bird, E., Cleave, P., Trudeau, N., Thordardottir, E., Sutton, A., & Thorpe, A. (2005). The language abilities of bilingual children with Down syndrome. American Journal of Speech-Language Pathology, 14, 187–199. Kay-Raining Bird, E., Lamond, E., & Holden, J. (2012). Survey of bilingualism in autism spectrum disorder. International Journal of Language and Communication Disorders, 1, 52–64. Korkman, M., Stenroos, M., Mickos, A., Westman, M., Ekholm, P., & Byring, R. (2012). Does simultaneous bilingualism aggravate children’s specific language problems? Acta Pediatrica, 101, 946–952. Kremer-Sadlik, T. (2005). To be or not to be bilingual: Autistic children from multilingual families. In J. Cohen, K. McAlister, K. Rolstad, & J. MacSwan (Eds.), Proceedings of the 4th international symposium on bilingualism (pp. 1225–1234). Somerville, MA: Cascadilla Press.

C. Hambly, E. Fombonne / Research in Autism Spectrum Disorders 8 (2014) 1079–1089

1089

Kjellmer, L., Hedvall, A., Fernell, E., Gillberg, C., & Norrelgen, F. (2012). Language and communication skills in preschool children with autism spectrum disorders: Contribution of cognition, severity of autism symptoms, and adaptive functioning to the variability. Research in Developmental Disabilities, 33, 172–180. Leadbitter, K., Hudry, K., & the Preschool Autism Communication Trial Consortium, (2009, May). Does bilingualism affect language development in autism? Poster session presented at the International Meeting for Autism Research, Chicago, IL.. Lord, C., Risi, S., Lambrecht, L., Cook, E., Leventhal, B., DiLavore, P., et al. (2000). The autism diagnostic observation schedule-generic: A standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders, 30, 205–233. Luyster, R., Lopez, K., & Lord, C. (2007). Characterizing communicative development in children referred for autism spectrum disorders using the MacArthur–Bates Communicative Development Inventory (CDI). Journal of Child Language, 34, 623–654. Maital, S., Dromi, E., Sagi, A., & Bornstein, M. (2000). The Hebrew Communicative Development Inventory: Language specific properties and cross-linguistic generalizations. Journal of Child Language, 27, 1–25. Ohashi, K., Mirenda, P., Marinova-Todd, S., Hambly, C., Fombonne, E., Szatmari, P., et al. (2012). Comparing early language development in monolingual- and bilingual-exposed young children with autism spectrum disorders. Research in Autism Spectrum Disorders, 6, 890–897. Oller, D. (2010). All-day recordings to investigate vocabulary development: A case study of a trilingual toddler. Communication Disorders Quarterly, 31, 213–222. Paradis, J., Crago, M., Genesee, F., & Rice, M. (2003). Bilingual children with specific language impairment: How do they compare with their monolingual peers? Journal of Speech, Language, and Hearing Research, 46, 113–127. Patterson, J. (2000). Observed and reported expressive vocabulary and word combinations in bilingual toddlers: Research note. Journal of Speech, Language and Hearing Research, 43, 121–128. Pearson, B., Fernandez, S., & Oller, D. (1993). Lexical development in bilingual infants and toddlers: Comparisons to monolingual norms. Language Learning, 43, 93–120. Petersen, J., Marinova-Todd, S., & Mirenda, P. (2012). An exploratory study of lexical skills in bilingual children with autism spectrum disorder. Journal of Autism and Developmental Disorders, 42, 1499–1503. Place, S., & Hoff, E. (2011). Properties of dual language exposure that influence 2-year-olds’ bilingual proficiency. Child Development, 82, 1834–1849. Seung, H., Siddiqi, S., & Elder, J. (2006). Intervention outcomes of a bilingual child with autism. Journal of Medical Speech-Language Pathology, 14, 53–63. Sparrow, S., Cicchetti, D., & Balla, D. (2005). Vineland adaptive behavior scales (2nd ed.). Minneapolis, MN: Pearson Assessment. Valicenti-McDermott, M., Tarshis, N., Schouls, M., Gadston, M., Hottinger, K., Seijo, R., et al. (2013). Language differences between monolingual English and bilingual English–Spanish young children with autism spectrum disorders. Journal of Child Neurology, 28, 945–948. Verhoeven, S., Steenge, J., & Van Balkom, H. (2012). Linguistic transfer in bilingual children with specific language impairment. International Journal of Language and Communication Disorders, 47, 176–183. Wu, J. (1997). Language, play and general development for Chinese infant-toddlers: Using adapted assessments (Unpublished doctoral dissertation) Boulder: University of Colorado. Yu, B. (2013). Issues in bilingualism and heritage language maintenance: Perspectives of minority-language mothers of children with autism spectrum disorders. American Journal of Speech-Language Pathology, 22, 10–24.