Lexicon development in bilingual toddler

Bilingualism, not monolingualism, is the global norm (UNESCO, 2003). We have a good understanding of how the minds of adult bilinguals store words in their two languages: for example, we know that a proficient French-English bilingual, when hearing the English word 'dog', unconsciously inhibits its French equivalent, 'chien'. However, there is little known about how the words are organised in the mind of a young bilingual. This is not a trivial problem as even monolingual children's mental processing of words is different from that of adults.
Understanding how bilingual children deal with words has important implications for assessment of their language learning abilities. Bilinguals tend to have delays in critical stages of language development, which can lead to over-referral for Speech and Language Therapy. The opposite problem is also prevalent, under-referral on the basis that some delay, relative to monolinguals, is expected. Such failings in the assessment system are costly for society in the short-term and the long-term, as a child with undetected language problems will struggle academically and socially.
The absence of bilingual-specific tools for language assessment can be attributed to (1) lack of knowledge about the architecture of the developing bilingual lexicon (2) the multiple social and linguistic influences on bilingual language development, which are poorly understood, particularly in combination. For example, we know that the linguistic distance between the languages being learnt is very important, but we don't have tools to measure linguistic distance in children.
Here we propose to produce a new computer-based model of the developing bilingual lexicon, based on experimental data, that will provide (1) significant advances in our theoretical understanding of bilingualism, and (2) pave the road for the development of a universal tool for the assessment of young bilinguals' language development.
This project is organised in three streams: two three-year data-collection streams and a final-year modelling stream. Children will be aged 18 months, which usually marks the onset of a sharp increase in vocabulary development.
In the first data-collection stream, we will run a series of experiments examining the organisation of words in bilingual infants. In collaboration with the Oxford Babylab, pairs of spoken words related or unrelated in meaning ('chien'-'pig' or 'chien'-'boat') will be presented accompanied by appropriate pictures. Children's looking times for pictures associated with related versus unrelated pairs will inform us about the strength of connections between words.
In the second data-collection stream, we aim to measure the interactions between the external factors influencing vocabulary development in bilinguals. Two hundred and fifty bilingual toddlers will be assessed in five different sites in the UK (Plymouth, Bangor, Birmingham, Kent and Liverpool) to get a large range of specifics in terms of language background and socioeconomic status. Infants' amount and mode of exposure to their two languages will be checked through appropriate questionnaires. Their level of vocabulary in English and the additional language will be evaluated through language-specific MacArthur-Bates Communicative Development Inventories (CDI), vocabulary questionnaires with versions for 60 languages. The data from the CDIs will help us to design child-adjusted metrics of distance between the two languages being learned.
In the final, modelling stream, the findings from the previous data-collection streams will be used to create a computational model of bilingual lexicon development, which will be able to predict language development in toddlers in different social and linguistic environments.
This project will considerably advance our understanding of language processing in young bilinguals, and prepare the ground for the design of a universal language assessment tool for this currently overlooked population.

Fighting inequalities and ensuring equal access to health care is at the heart of the UK government policy (Health and Social Care Act 2012). However, at present health care practitioners lack the tools to effectively screen the (growing) population of bilingual children in their two languages for early language delays (Kohnert, 2010). The detection of these delays, allowing the provision of adequate care is of vital importance for the social, emotional and cognitive development of children (Snowling et al., 2002). The research plan of this project have been specifically tailored towards the provision of the empirical and modelling data that would be required for the construction of the first universal vocabulary assessment tool for young bilinguals (VATB).

Who will benefit from this research?
There are estimated to be around 105,400 multilingual 2-year-olds in England (extrapolated from UK National Statistics sources and the 2010 School Census, Department for Education), of whom 7-15% (the same proportion as monolinguals) can be expected to experience a delay in language acquisition due to pervasive developmental disorders (including autism), sensory or specific language impairment (SLI). Unfortunately these children are currently at significant risk of misdiagnosis and under referral, with Crutchley (2000) finding that a far larger proportion of the bilingual parents (45%) stated that professionals had initially failed to diagnose their child's difficulties or take note of their worries than did the monolingual parents (18%).
In theory, recommendations by the Royal College of Speech and Language Therapists (RCSLT, 1998) and the American Speech-Language-Hearing Association (ASHA, 1999) are that bilingual pre-schoolers should be screened in their two languages. However, children's assessment in their Additional Language (AL) can be challenging as a great many languages lack standardised tests, and even if such a test does exist the monolingual norms cannot be directly applied to bilingual children as they usually underperform monolinguals (Hoff et al., 2012). Even use of interpreters is problematic as translation styles may vary from the general to the literal word-for-word (Stow & Dodd, 2003). Parental assessment (Paradis et al., 2010) suffers from variability in parental opinion or experience, and the highly promising dynamic assessment (Gutiérrez-Clellen & Pena, 2001) requires a significant experience and knowledge (Kohnert, 2010), making it unsuitable for wider screening applications at 18 months.

How will they benefit from this research?
Providing SLTs and health care professionals with a flexible, easy-to-use and reliable tool to evaluate vocabulary of 18-month-old bilingual toddlers will introduce a massive change in the access to care for the bilingual populations. The VATB would be implemented as a user-friendly computerised interface, allowing health care professionals (health visitors, crèche assistants) to enter a wide range of information about the child together with her results at English and Additional Language vocabulary tests. This would allow them to make an informed decision regarding the child's need for a referral from as early as 18 months, without unnecessarily increasing SLTs' waiting lists (which can be up to 24 months in the UK). This first assessment would be done at an age where a 'watch-and-see' approach (Paul, 1996) would be very appropriate and not cost-effective to prepare for a later intervention. The principles of the VATB could set a benchmark for all countries in need of a tool for bilingual children screening.