Introduction

There is an urgent need to understand the impact of bilingual exposure on the language development of young children with Autism Spectrum Disorders (ASDs) since there is a prevalent belief in clinical settings that bilingually-exposed children with ASDs may experience additional delays in language development. There is little data to support or refute this assumption since early bilingual exposure in young children with autism has only been explored in presentations (Hambly and Fombonne 2009; Leadbitter et al. 2009; Kremer-Sadlik 2005; Yu 2007), a manuscript currently under review (Petersen et al. personal communication) and a case report (Seung et al. 2006).

There is no evidence that bilingual exposure causes additional language delays for children with language impairments. Research on bilingual exposure in children with specific language impairment (Gutierrez-Clellen et al. 2008; Paradis et al. 2003) and Down Syndrome (Feltmate and Kay-Raining Bird 2008; Kay-Raining Bird et al. 2005) found no additional language delays in bilingually-exposed participants. These studies had small numbers of participants in the language-impaired bilingual exposure groups (e.g., 4–11), and participation was generally limited to children with ‘intensive’ bilingual exposure and productive expressive bilingual abilities. To paraphrase Genesee (2006, p. 51), these kinds of studies indicate that some children with language impairments and bilingual exposure acquire language to the same level as monolinguals, but they do not suggest that all bilingually-exposed children with language impairments do.

Children with ASDs have severe social impairments that are not present in children with specific language impairment or Down Syndrome. The social impairments associated with ASDs include deficits in attention to voices and joint attention: these core skills are so critical to language learning that these deficits could potentially limit the ability of children with ASDs to learn language in bilingual environments at the same rate or to the same level as children with ASDs in monolingual environments.

First, a preference for non-speech over speech sounds reduces automatic orienting to speakers in the environment (Kuhl et al. 2005) resulting in fewer opportunities for children with ASDs to analyze and sort the bilingual auditory input. Along with reduced auditory input, children with ASDs may also not attend to facial movements during speech that provide important visual cues for sorting languages (Weikum et al. 2007). There may also be a mingling of environmental noises with the speech signal that could mask the contrastive phonemes and stress patterns needed to sort languages and parse input within each language (Russo et al. 2009). These additional auditory processing demands can have an impact on language development in infancy: experimental studies have demonstrated that typically-developing bilingually-exposed infants are slower than monolingual infants to encode and retrieve the phonetic details that help them make novel object-word associations (Fennell et al. 2007). Word and eventually sentence comprehension and production could be affected if bilingual input cannot be readily or accurately perceived or processed. Although bilingually-exposed infants may display enhanced cognitive control abilities in bilingual processing tasks compared to monolingual infants (Kovács and Mehler 2009), it is not known if any cognitive advantages related to bilingual exposure during infancy would also be present in infants who are later diagnosed with ASDs.

Secondly, impaired joint attention abilities (Adamson et al. 2009) could also potentially worsen receptive and expressive language delays for bilingually-exposed children. Children with ASDs have reduced use of joint attention cues such as referential pointing and eye gaze that help them ‘map’ word labels to the appropriate referent (Parish-Morris et al. 2007). In addition to the mapping tasks that all children must perform, bilingual children must also recognize that two or more different word labels (e.g., apple/pomme) can map to a single concept or even multiple concepts across languages. For example, the English third person plural pronoun “they” maps to both “ils” and “elles” (masculine or feminine third person plural) in French, but a child learning English and French would also need to recognize that the French singular pronouns “il” and “elle” sound identical to their plural counterparts. Receptive and expressive vocabulary skills could be delayed in both languages if word mapping is perceived to be more inconsistent by bilingually-exposed children with ASDs, and additional grammatical delays could be a product of further perceived inconsistencies in grammatical systems (e.g., differing use of verb tense and pronoun morphemes in English vs. French) or structural order across languages (e.g., ‘the blue car’/‘une voiture bleue’).

The timing of bilingual exposure could also impact the outcomes of language learning for children with ASDs. The infant perception literature supports a general perceptual narrowing from 6 to 12 months of age as infants tune into the specific sensory input that characterizes their linguistic environment (Pons et al. 2009). Many studies have demonstrated that typically-developing infants are sensitive to acoustic contrasts across languages that older children do not perceive; there appears to be a critical window for heightened speech perception and discrimination during the first year of life (Werker and Byers-Heinlein 2008; Werker et al. 2009; Werker and Tees 1984). During this period of broad perceptual sensitivity, bilingually-exposed infants differentiate and sort linguistic input based on features including vowel contrasts (Sebastián-Gallés and Bosch 2009; Sundara and Scutellaro 2010), consonant contrasts (Sundara et al. 2008), and visible articulation cues (Pons et al. 2009) to develop a foundation in two language systems. The perceptual ‘tuning’ that occurs before the end of the first year of life (Gervain and Werker 2008) reflects the infants’ attention to contrasts that are relevant in their language system(s) and has an impact on word learning in typically-developing children: very young monolinguals appear to use different learning mechanisms to acquire words than their bilingually-exposed peers (Werker et al. 2009). It is possible that perceptual abilities and word-learning strategies may differ depending on whether a child with an ASD received bilingual or monolingual exposure during the first year of life.

We hypothesized that the social impairments characteristic of ASDs could cause additional language delays in bilingually-exposed children with ASDs compared to monolingually-exposed children with ASDs. These additional delays would manifest in smaller expressive vocabularies, lower levels of language comprehension and production, and later onset of early language milestones for bilingually-exposed children with ASDs. Within this general hypothesis, we posited that children with ASDs whose bilingual exposure started after infancy might show greater language delays than children who received simultaneous bilingual exposure beginning during infancy.

Methods

Participants

Families of 75 children with ASDs aged 36–78 months were recruited from Quebec (N = 71) and Ontario (N = 4). 65% of the children were patients at the Montreal Children’s Hospital (MCH) ASD clinic, recruited via mailings or during clinical visits. 35% were recruited from publicity in community organizations and service providers. The participants from Quebec were diagnosed by child psychiatrists or developmental pediatricians; the participants from Ontario were diagnosed at hospitals or a university-based ASD research unit. Diagnostic reports were available for 89% of all participants (96% of children diagnosed at the MCH, 82% of those diagnosed elsewhere), with the Autism Diagnostic Observation Schedule-Generic (Lord et al. 2000) used in at least 66% of participants’ assessments. Three 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 within the autism category (n = 59; 56 diagnosed with autism, 3 listed as “ASD”). One child diagnosed with Asperger syndrome was grouped with the Pervasive Developmental Delay—Not Otherwise specified category (n = 16; 15 diagnosed as PDD-NOS, 1 diagnosed with Asperger syndrome).

Enrollment was limited to children whose spoken vocabulary was in French, English, Chinese, Farsi, Hebrew, Italian, Romanian, Spanish and/or Tamil, since the expressive vocabulary questionnaire was available in these languages. Eleven children with trilingual exposure were included because their expressive vocabularies could be measured using available language forms. Nonverbal children were included since we had theorized that bilingualism could be detrimental to language development in the presence of an ASD: if true, we would have introduced a sampling bias in the bilingually-exposed group by excluding nonverbal children. Seven participants spoke fewer than 10 words, but only two of these children were nonverbal. The informant caregivers all spoke and read English or French well.

Participants were divided into groups based on their language exposure history (monolingual vs. bilingual) with the bilingually-exposed children sub-grouped based on dual-language exposure before or after 12 months of age. This age cut-off was selected based on changes in infant speech perception that occur around this age, and after consideration of the various methodological approaches to studying bilingualism in young children. There is little consensus in the literature on how to subgroup bilingual children: Bialystok (2001, p. 225) highlights the subjectivity and methodological disputes around this issue.

One common subgrouping methodology is based on McLaughlin (1978) who used a cutoff age of 3 years—an age at which typically-developing children have phrase-level expressive language abilities—to differentiate between children who learn two languages simultaneously and those with later exposure who learn their languages sequentially. However, many children with ASDs have a protracted language acquisition period compared to typically-developing peers, and some children with ASDs may never develop spoken language, much less phrase-level expressive language abilities. Any age cutoff based on expected language levels at a given age would be meaningless in a population of children with significant communication delays and variable expressive language achievement, so this approach was not selected for this study.

Another common methodological approach is to assess current bilingual exposure and to select only children who receive a certain percentage of bilingual input each week (as in Gutierrez-Clellen et al. 2008). However, a major limitation of this approach is that it does not account for changes in language use over time. Many families in this sample made changes in language use proportions from year to year, so bilingual exposure was not constant across time. Bilingual families often reduced their use of a minority home language (e.g., from speaking it 50 to 20% of the time, for example) or spoke it among other family members but not directly to the child (Yu 2007) when their children’s language delays became apparent. In this scenario, the bilingual exposure was present since birth, but the child’s exposure to the language changed from direct to indirect exposure at home. So, grouping by current direct language exposure levels would have masked the large exposure changes over time that may be unique to this population.

An additional methodological challenge to categorizing participants based on language exposure emerges from the presence of major language changes, which in this study was defined as either the addition of a novel language to the child’s environment, or conversely the complete loss of a language in direct communication with the child. For example, families from monolingual minority language homes often reported adding a second language at home after becoming aware of their child’s language delay and subsequent need for majority-language therapy services. Some bilingual families stopped speaking one of their languages to their child with an ASD, or reduced a trilingual home environment to a bilingual one. These changes were tracked in this study: although the cut-off for subgrouping by bilingual exposure was placed at 12 months of age, the simultaneous and sequential exposure groups actually had far more than 12 months of exposure differentiating them from each other.

Our decision to subgroup children based on infant language exposure was ultimately guided by the research on infant speech perception, but the use of an early cut-off also allowed us to differentiate between families that had an original intent to provide a bilingual environment for their children and those families who did not. Since the cut-off for simultaneous bilingual exposure was placed before the age at which language delays are usually detected—an event that can be a trigger for major language environment changes—the chosen subgrouping methodology effectively separated families who intended to raise their children bilingually from those families who had planned for a monolingual childhood. It acknowledges the possibility that families who ‘begin bilingual’ may provide more natural and fluent language and socialization models and more indirect bilingual exposure than families who add a second language to help their child access majority-language therapy or schooling (Yu 2007). In other words, the qualitative differences between the simultaneous and sequential exposure groups may be as important as any perceptual advantage from infant exposure or any quantitative difference in levels of direct exposure between the bilingually-exposed groups.

Table 1 displays the characteristics of the three exposure groups.

Table 1 Sample characteristics by language-exposure groups
Full size table

Monolingually-exposed children (MON, n = 30) had no history of bilingual exposure. Bilingually-exposed children (n = 45) presented with a range of exposures, but were sub-grouped based on whether they received simultaneous bilingual exposure from before 12 months of age (SIM, n = 24) or had second language exposure added sequentially into the environment after 12 months of age (SEQ, n = 21). The age of the earliest major language changes for each group (SIM = 25.6 months, n = 14, SEQ = 31.0 months, n = 21) highlights that major alterations in language environment were typically made after the age of first concerns: in the SIM group, the changes included a shifts from (a) trilingual to bilingual direct exposure and (b) bilingual to monolingual direct exposure. The age of earliest major language changes also documents that on average children in the SIM group were bilingually-exposed for the first 25 months of life, whereas children in the SEQ group were monolingually-exposed for the first 31 months of life. So, while a cut-off of 12 months of age was used to distinguish between the groups, the average difference in early monolingual versus bilingual exposure between the groups was significantly more than 12 months.

Children in the SIM group whose bilingual or trilingual exposure was reduced or eliminated over time were retained in the SIM group due to their significant early bilingual exposure. Children whose first language of exposure was completely replaced by a second language in caregiver-child interactions remained in the SEQ group due to the high likelihood of ongoing indirect bilingual exposure (e.g. between family members) at home.

MON children were exposed to English (53% of participants), French (43%), or Spanish (3%). SIM children were divided between bilingual French and English exposure (54%), trilingual exposure with French, English and a minority language (33%), and bilingual exposure with French or English and a minority language (13%). SEQ children were divided between bilingual French and English exposure (48%), bilingual exposure with French or English and a minority language (38%) and trilingual exposure with French, English and a minority language (14%).

Instruments

Most instruments were selected to categorize general and/or adaptive language levels with minimal to no referencing of specific language features. Instead of standardized test measures which have limited availability outside of English or Spanish and generally can only test a single language at a time, we selected parent report measures which are commonly used in bilingualism research with young children. French translations of the measures were used with French-speaking families.

All families participated in a Language Environment Interview (LEI) that collected data on home language environment history and generated a language exposure estimate. The LEI questions were administered by a trained research assistant during a 30 min phone interview with an informant; most informants (83%) were the children’s mothers. Questions and scoring are described in detail in “Appendix 1”. To summarize, the informant provided a detailed history of all caregivers’ amount and duration of care for each 6-month period of the child’s life. The informants also described the languages these caregivers used with other family members and with the child. Additional information on countries of birth, a report on caregivers’ fluency in all languages spoken at home, and reasons for language changes were collected for family caregivers. Language exposure information was gathered for daycare or school environments where relevant.

Following the interview, major language exposure changes (e.g., languages added or dropped) were identified, and data for each caregiver was summarized for each 6-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 child-care 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 exposure in the home. The language heard by the child most often during their lifetime was labeled their dominant exposure language.

Current socio-communicative levels and ages of early language milestones were elicited using questions excerpted from the Autism Diagnostic Interview-Revised (Le Couteur et al. 2003) and included: pointing for joint attention initiation and response (#42), attention to voice (#46), comprehension of simple language (#29), overall level of language (#30), age of first words (#9), and age of first phrases (#10). Interviewers were trained using ADI-R guidelines, and coding procedures were followed with one exception: question 42 was modified to separate initiation and response levels. Scores on some ADI-R questions were merged to facilitate statistical analyses; consult “Appendix 2” for details on questions and scoring.

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. Data were summarized into a weekly average. The correlation for bilinguals between the Language Diary and the last six-month period assessed on the LEI was highly significant (r = 0.773, p < 0.001).

The Social Responsiveness Scale (SRS) measures the severity of autism symptoms within children’s natural environments (Constantino 2002); it is appropriate for use in children with autism aged 4 years and older. Scoring procedures and standardization data followed the technical guide (Constantino and Gruber 2005).

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 et al. (2009) in categorizing participants’ ethnic/cultural heritage responses as ‘caucasian’ versus ‘non-caucasian’.

A ‘List of Services’ elicited treatment program histories including intensive behavioral interventions, speech-language, occupational, and physical therapy, and descriptions of other therapies. Intensive behavioral service descriptions were subsequently categorized as ‘no services’ versus ‘some services’.

The MacArthur Communicative Development Inventory: Words and Sentences (MCDI) (Fenson et al. 1993) is a well-studied parent report measure of vocabulary; it is designed for children aged 16–30 months, but can be used 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 et al. 1993) used in this study included English (Fenson et al. 1993), Quebec French (Frank et al. 1997), Spanish (Jackson-Maldonado et al. 2003), Italian (Caselli and Casadio 1995), Hebrew (Maital et al. 2000), Chinese (Wu 1997), Farsi (Kazemi et al. 2008), Romanian (Geangu and Benga 2006), and Tamil (Sethuraman 2008). 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).

Scoring for bilinguals followed the procedures outlined in Pearson et al. (1993) for computing Total Conceptual Vocabulary (TCV): concepts named in both languages are mapped to each other so that translation-equivalent word forms are only counted once even when used in two languages. Cross-language mappings were developed for this study for English and French with Chinese, Hebrew, Italian, Spanish, and Romanian. A ceiling score of 680 concepts (selected to match the ceiling on the English form; other forms vary between 611 and 680 words) was applied to the bilingual exposure groups since their TCV scores could exceed the maximum ceiling available to the MON group.

The Vineland Adaptive Behavior Scales-Second Edition (Sparrow et al. 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 0.85 (Sparrow et al. 2005).

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, and return postage. In order to evaluate the reliability of the vocabulary measure, 17 families completed the MCDI 6 weeks after the baseline measurement. The first 16 families also received the VABS-II Parent/Caregiver Rating Form (Communication and Socialization domains) in their package. Since mailed VABS-II were often returned with incomplete information, the remaining 59 families were administered the VABS-II Survey form by phone by a trained research assistant. The LEI and ADI-R questions were administered by a trained research assistant within a single phone interview.

Statistical Analyses

Group differences on continuous variables were analyzed using one-way analysis of variance (ANOVA). Significant ANOVA results were subsequently assessed with post hoc t tests on any variable with a p value of <0.1 to identify which group pairings were significant. All categorical data were examined using chi-square tests of independence. Post-hoc testing employed chi-square or Fisher exact tests to identify significant contrasts between paired groups. Statistical significance was set at the p < 0.05 level, but post hoc testing was performed on any variable with a p value of <0.1 in order to explore trends. Statistical analyses were performed using SPSS 14.

Six-week test–retest reliability data for the MCDI was collected and analyzed: the Pearson coefficient of >0.9 corresponded to the test–retest reliability for the same period reported in the MCDI User’s Guide (Fenson et al. 1993) and supports the use of this measure in a bilingual population.

Results

Sample characteristics by exposure grouping are presented in Table 1. Children with simultaneous (SIM) versus sequential (SEQ) bilingual exposure had significantly different amounts of lifetime bilingual exposure (p < 0.001) but similar current bilingual exposure levels on both the language diary and the last 6–12 month subtotal of the Lifetime Ratio (LR). SEQ children were significantly more likely (p = 0.015) than MON to have a minority dominant-exposure language. Maternal characteristics were similar across groups except for a difference in mothers’ fluency: there were significant contrasts in maternal native fluency in child-directed language(s) between MON and both SIM (p = 0.002) and SEQ (p = <0.001) mothers, indicating that bilingual mothers were not always native speakers of the languages they use with their children. There was a difference approaching significance in receipt of intensive behavioral services, with significantly more SEQ compared to SIM (p = 0.038) children receiving services.

Social and language measures are presented for the exposure grouping in Table 2.

Table 2 Social and language measures by language exposure groups
Full size table

We had hypothesized that social impairments could negatively affect language-learning in bilingual environments for children with ASDs. Prior to examining language results, both general and specific aspects of social abilities were compared across groups to verify that these abilities were evenly distributed in this sample. The only significant difference between groups was found in the VABS-II Interpersonal subdomain (VINTP) scale score (F (2,63) = 3.4, p = 0.038), where the SIM group had the highest scores of the three exposure groups and the SEQ group had the lowest scores. Post hoc testing revealed significantly better performance for the SIM group compared to the SEQ (p = 0.025) group; there was an additional group difference at the threshold of significance between the MON and the SEQ (p = 0.052) groups.

Despite the significant group difference on the VINTP, there were no significant differences on any language variable, although a pattern of performance emerged whereby SIM and MON consistently scored higher than SEQ children. Neither Total Conceptual Vocabulary (TCV) nor dominant language counts differed significantly between groups. TCV and dominant language counts were similar because only 66% of the SIM and 57% of the SEQ children reported words in a second language: these bilingual children used many translation-equivalent pairs so their TCV scores were not raised much beyond their dominant language word counts. There were no significant differences on the other expressive and receptive measures. No differences were noted in the early language milestones data.

Discussion

The results suggest that bilingually-exposed children with ASDs do not experience additional delays in language development compared to monolingually-exposed children with ASDs. In addition, the timing of bilingual exposure from infancy versus the post-infancy period did not impact outcomes in the children’s dominant language abilities. While there may be a perceptual advantage to bilingual exposure during infancy, and while children with infant bilingual exposure heard more of their second language over their lifespan than their sequentially-exposed peers, there was no difference between the exposure groupings in dominant language outcomes. Second language outcomes will be examined in a separate paper.

We had suggested theoretical reasons why children with ASDs might react to language learning in bilingual environments differently than children with SLI and Down Syndrome, but it appears that children with ASDs are capable of similar language achievements regardless of whether their environment is monolingual or bilingual. Impairments or differences in speech perception, joint attention, and other foundational skills for language learning may not produce additional vulnerabilities for bilingually-exposed children compared to their monolingually-exposed peers with ASDs.

This study documented that some bilingually-exposed children with ASDs are acquiring bilingual vocabulary in early childhood: 62% of the bilingually-exposed children spoke words in a second language, but their second language vocabularies were much smaller than their dominant language vocabularies, and few children had progressed to phrase-level speech in a second language. So, while there does not appear to be any risk related to bilingual exposure in infancy or early childhood, the benefits of this exposure may not be fully apparent in early childhood. Parents who wish to maintain a bilingual environment should be counseled that bilingual acquisition is an unstudied area for children with ASDs. For example, we do not know if children with ASDs acquire second language vocabulary more easily than second language phonology, morphology, or syntax, but learning across all these language domains is necessary to become bilingual. Longitudinal research designs will be critical to understanding bilingual acquisition in this population.

Finally, the variety of bilingual experiences and the types of changes noted in this sample were more diverse than expected. For example, while some families did display the commonly-cited pattern of switching from bilingual to monolingual exposure when their children’s language delays became apparent, many other families introduced a second language after the age of first concerns. Patterns of language changes and reasons driving parental decisions will be explored in a separate paper. However, families of children with language delays may make a variety of language choices that have not been described in the literature to date. These choices may be influenced by parents’ own language abilities, their rights as immigrants (e.g., French language education is mandatory in Quebec for immigrants, but not for families with a Canadian-born parent whose children can attend English-language schools), their choices for special education (e.g., inclusionary policies in English schools in Quebec vs. ‘exclusionary’ autism class options in French schools), or childcare options (e.g., family caregivers vs. public childcare). Clinicians and researchers need to be aware that bilingual experiences are diverse.

It is important to acknowledge several limitations to this study. First, the recruitment approach allowed families to self-select for participation but our large sample of children encompassed a broad and representative range of experiences and language levels, and the results may be more generalizable than results from studies which selected participants based on restricted levels of exposure or verbal skills. Secondly, the study was designed in order to detect a mean vocabulary difference of 50 words between groups: it is possible that a larger sample size would have been able to detect smaller group differences in vocabulary size or differences in subgroups, such as within children with autism versus PDD-NOS. Third, the relationship between the amount of bilingual exposure and language outcomes was not examined in this study because most bilingual families made multiple exposure changes that complicated quantification attempts. Fourth, we did not analyze cognitive test data although intellectual functioning plays an important role in the acquisition of language. Finally, caution should be exerted when evaluating the direction and causal nature of the results.

This study suggests several directions for research. Research on bilingualism in children with ASDs must take into account that there is a longer language acquisition period and more variable language outcomes for children with ASDs than for typically-developing children, and that methodologies based on typical language development may not be appropriate in this population. Future research on bilingual experiences in children with special needs could explore the methodological challenges in grouping bilingually-exposed participants who have language delays. Next, this study’s findings should be replicated using independent assessment measures of language outcomes. Furthermore, it is critical to study how factors such as parental language fluency and family language choices could affect children in bilingual environments and potentially influence child language outcomes, particularly when a child has language delays. Finally, it would be beneficial to follow bilingually-exposed children longitudinally in order to understand the pace and patterns of language growth.

In conclusion, the data suggest that there is no language delay associated with bilingual exposure for children with ASDs regardless of whether the children were exposed to two languages from infancy or from early childhood. Caregivers should not be discouraged from maintaining bilingual environments or introducing a second language where necessary for the child or family, although little is currently known about the pace of learning or bilingual achievements for children with ASDs.