Characterizing accuracy and variability in speech sound productions across bidialectal and bilingual preschoolers - Project Summary/Abstract Speech sound disorder (SSD) is a high-incidence developmental disability that can result in long-term negative impacts on academic and career achievement. Although the impacts of SSD can be mitigated through evidence-based interventions, the challenge of diagnosing and managing SSD is greatly exacerbated when working with clients from linguistically diverse backgrounds. A clinician evaluating a child who uses more than one language daily must avoid over-diagnosing disorder in cases more accurately characterized as differences while also guarding against under-diagnosis of SSD in this population. This need to reduce misdiagnosis of SSD is particularly pressing in understudied contexts, namely children who are bidialectal or bilingual in two closely related languages. In these cases, production variation is ubiquitous yet might be flagged as being atypical in the diagnostic process. Languages interact differently depending on their typological properties, and the body of existing literature focused on a small set of relatively well-documented cases (e.g., Spanish- English) runs the risk of arriving at an overly homogenous model of speech development and disorders. By studying two groups of children – those acquiring African American English and Standardized American English (AAE/SAE) and those acquiring Jamaican Creole and Jamaican English (JC/JE) – this proposal not only benefits underserved populations but will also broaden the theoretical base for understanding speech development and disorders in linguistically diverse contexts. Although our team has made recent strides toward understanding trade-offs between accuracy and variability in a bilingual context, additional elaboration of the theoretical framing is needed to inform the diagnostic significance of production variation when languages are linguistically related. We propose to model the relationship between accuracy and variability in production using the Articulatory Map ([A-map]; McAllister Byun et al., 2016) and Gradual Learning Algorithm ([GLA] Boersma and Hayes, 2001). Application of the A-map and GLA are needed to account for typological patterns of phonetic and phonological learning across language contexts both theoretically (A-map) and computationally (GLA). The long-term goal of this research is to arrive at a coherent understanding of production variation to reduce misdiagnosis of SSD in underrepresented populations. Using transcription- based measures of accuracy and acoustic (spectral)-based measures of variability, we will test the predictions of the A-map and GLA in preschool-aged bidialectal AAE/SAE and bilingual JC/JE speakers. Statistical modeling of the relative contributions of acoustic-based measures of variability and transcription-based measures of accuracy to a gold standard diagnosis will be used to identify the optimal diagnostic criteria for disorder status. The data we generate will support a methodological template for protocols that reduce misdiagnosis of SSD in bidialectal and bilingual children to meet the NIDCD 2023-2027 Strategic Objective Theme 4: Goal 3 to advance understandings about communication disorders in underrepresented populations.
