Language testing of preschool-aged children is important in the identification of a wide range of developmental disabilities. Particularly when children are between one and five years of age, it is critical to supplement standardized tests with analysis of a naturalistic language sample (Language Sample Analysis, LSA), using a broad set of lexical and grammatical measures developed over the past forty years. Clinical training emphasizes the importance of using such measures in clinical decision- making. Unfortunately, this type of analysis is time-consuming, and clinicians often perform only a cursory analysis. Additionally, the psychometric properties of most of these language sample analysis (LSA) measures are quite weak, primarily because relatively few children, all of whom spoke mainstream English (rather than common dialects such as African-American English [AAVE], contributed to the development of reference scores. This situation is undesirable given educational mandates requiring that developmental assessment be done using valid, reliable and unbiased instruments and procedures. This project will utilize data from over 1500 children?s language samples compiled and curated at the Child Language Data Exchange System (CHILDES) open research data repository (www.childes.talkbank.org ) to validate, extend and strengthen the psychometric properties of LSA measures, for use in clinical and educational practice. We will integrate findings from large cross- sectional corpora with smaller numbers of children followed intensively over early development to verify trajectories in the mastery of a broad array of conventionally used LSA measures. Additional aims include development of dialect-sensitive measures and reference values, as well as those specific to gender, since current reference values may disadvantage girls. We will additionally explore impacts of socio-economic status on LSA values. Weak ?norms?, dialect variation, gender and social class may significantly impact appropriate referral for clinical and educational services and we attempt to remediate this. We will validate our resulting ?norms? by examining the accuracy with which they correctly classify an additional set of corpora from children with known developmental delays, including AAVE speakers. In sum, this initiative will translate a large body of primary research data gathered in the study of children?s typical mastery of expressive language skills to a freely-available utility that may be used by practicing clinicians to more easily, accurately and fairly identify preschool children in need of intervention services.
Early detection of language delays is critical to a host of developmental disabilities; however, normative data for expressive language sample analysis (LSA) are based on extremely small samples. This proposal uses the resources of more than 1500 language samples from by the CHILDES Project to validate and re- norm the most commonly utilized clinical LSA measures and create dialect-sensitive alternatives. Our final product, a free, easy-to-use software program, should enable easier and better identification of early language delays and inform appropriate intervention.
