Co-Principal Investigator: Schmitt, Mary Beth

Purpose: The purpose of this project is to explore the role of speech-language pathologists' (SLPs) talk in facilitating the language growth of children with language impairment receiving therapy in the public elementary schools. Language ability undergirds most areas of school achievement, including reading, writing, math, and pro-social functioning. It is also important in its own right, as language ability is foundational to one's ability to communicate. There is a great need for educational research focused on mechanisms by which to influence children's language growth, particularly for children with disabilities. In this study, the research team will use previously collected data to study the role of SLPs' child-directed talk across three domains (quantity, complexity, responsivity) as a potential mechanism for language growth in children with language impairment. The team will also conduct a feasibility study to test whether an adaptive feedback system can affect the complexity of SLP talk during therapy sessions.

Project Activities: The research team will transcribe, code, and analyze SLPs' talk during 811 recorded therapy sessions collected in a prior IES study and examine relations between SLPs' talk and children's language growth as measured at five times over the academic year. They will also conduct a feasibility study using a small, randomized trial to examine whether an adaptive feedback system can affect the complexity of SLP talk during therapy sessions.

Products: The project will result in an understanding of the associations between the characteristics of SLP talk and language outcomes in elementary school children with language impairment as well as data on the feasibility of an adaptive feedback system for SLPs. It will also result in a shared final dataset, peer-reviewed publications and presentations, and products that reach education stakeholders such as practitioners and policymakers. The research team will disseminate the open-science tools used to automatically transcribe and code SLP talk freely to the research community.

Structured Abstract

Setting: The primary study was conducted in public schools from 2009 to 2013 in Ohio and Tennessee. The feasibility study in Year 4 will be conducted in an urban pediatric hospital.

Sample: The 811 recorded speech-therapy sessions represent a sample of 73 certified SLPs and 293 children in kindergarten, first, and second grade. The children had language impairment and were receiving ongoing therapy from an SLP to target language-specific goals. The Year 4 feasibility study will involve 20 SLPs from an urban pediatric hospital and children with language impairment on their caseloads.

Factors: Factors of interest in this exploration study are three domains of SLPs' child-directed talk: quantity, complexity, and responsivity. The research team hypothesizes that one or more of these domains will serve as a key mechanism of language growth for children based on input-based theoriesof language development.

Research Design and Methods: The primary research aims will be addressed by transcribing, coding, and analyzing previously collected therapy videos in a prospective longitudinal study. In that study, therapy was delivered largely in pull-out rooms designated for speech and language services. Up to five regularly scheduled therapy sessions were recorded for each included child. Transcripts of SLP talk during these sessions will be automatically generated, using an existing open-source tool created by the research team. Child talk in these sessions will be hand-transcribed by trained research assistants.  For the Year 4 feasibility study, the research team will use an under-powered randomized control trial in which 10 randomly assigned SLPs are provided adaptive feedback targeting the complexity of their talk in three separate one-on-one therapy sessions with children with language impairments.

Control Condition: In the feasibility study, the SLPs in the control condition will conduct therapy as usual without adaptive feedback.

Key Measures: Outcomes and factors will be measured with language sample analysis of transcribed, previously recorded video. Software will automatically capture from the transcript the total number of SLP utterances, total number of words, and words per minute, as measures of quantity of SLP talk. The mean length of SLP utterances, clause density, number of different words and type-token ratio will be coded by computer to measure complexity of SLP talk. The number of times an SLP recasts a child's utterance and the percent of child utterances that are recast will be hand coded by research assistants to measure responsivity of SLP talk. Child language growth will be measured with four distinct indices: mean length of utterances in morphemes, total number of words, clauses per utterance, and words per utterance. Information collected in the previous study, including surveys of SLP demographics and caseload, and child language skills and nonverbal cognition measured with Clinical Examination of Language Fundamentals-4 and Kaufman Brief Intelligence Test will be included in the dataset.

Data Analytic Strategy: Three data analytic strategies will be employed in this study. First, the research team will use multilevel confirmatory factor analyses to examine individual variation in the quantity, complexity, and responsivity of SLP talk. Second, they will use a multilevel longitudinal regression model to explore SLP-level characteristics that are associated with dimensions of their talk. Third, to explore relations between dimensions of SLP talk, the team will estimate cross-lagged panel models aimed at capturing the reciprocal relationship between each dimension of SLP talk and child language over time, controlling for child-level attributes (such as gender) and SLP-specific random effects. In the feasibility study, researchers will compare the prescribed and delivered complexity of SLP talk using an appropriate nonparametric analysis.

Related IES Projects Language Growth and Therapy Characteristics for Early Elementary Students (R324A090012)