DISORDERED-SPEECH · Early Identification of Dyslexia and DLD using Speech Production Measures with Children: An Innovative Cross-Language Approach

Language problems are common in children, with ~7% incidence of significant spoken language problems (Developmental Language Disorder, DLD) and a further ~7% for written language problems (Developmental Dyslexia, DD). While there can be overlap in behavioural indices of DLD and DD (eg prosodic difficulties), no accepted causal theories satisfactorily explain such overlaps, and early identification remains difficult: only half of “late talkers”, toddlers who are not speaking by age 2, go on to develop DLD. Here we combine our innovative neural theory of child language acquisition (temporal sampling [TS] theory) with a novel speech production task to open a new research front in developmental language disorders, based on speech rhythm and the accurate production of prosodic patterns. We test both pre-school children aged 2.5 – 6 years and children with DD and DLD in 4 languages varying in rhythm type, English, French, Spanish and Basque. We apply our in-house state-of-the-art methods for measuring the similarity between the child’s speech amplitude envelope (AE) and pitch contour (PC) of multisyllabic items to oral targets. Prior English data show impaired AE and intact PC for DD, impaired AE and impaired PC for DLD, and consistent links between AE/PC production in typically-developing (TD) children and phonology/vocabulary. We test whether similar patterns characterise TD, DD and DLD across languages. Our focus on the accurate production of the AE/PC is entirely novel, exploits an automatic aspect of linguistic processing, applies to all languages, and may suggest new avenues for remediation. The speech production data acquired from >1000 children will further enable a machine learning approach to early diagnosis based on AE/PC. Our goal is to develop new tools for identifying DD and DLD which can apply across European languages. The frontier vision is that as AEs and PCs describe words in all languages, our approach may reveal new universals in language disorders.