Testing Language in the Presence of Dialect: Variation, Risk, and Clinical Validity - PROJECT SUMMARY/ABSTRACT
Language learning is a key component of various lifelong economic and health outcomes (e.g, reading,
employment, long-term college and career success). Strong, mainstream language skills are particularly
important to performance on standardized language assessment. Children who speak nonmainstream dialects
may be particularly challenged by the complexities created by dialect differences in language testing
environments. There is a large body of work documenting these challenges for African American children.
While modifications of existing language tests have been recommended and implemented for quite some time,
these modifications usually involve either giving children credit for items influenced by dialect or skipping these
items altogether. While such accommodations may avoid invalidly punitive test scores, such accommodations
also may also impede clinical diagnosis and needed intervention services, thereby exacerbating an already
serious health concern for AA children in potential need. Our overall objective is to understand the strengths
and weaknesses of language and dialect in AA children from a rigorous measurement perspective. It is not
clear that the measurement of language can be separated from the measurement of dialect, which is a
variation on that language. Our central hypothesis is that items of existing language tests measure crucial
features of language and that some items measure crucial features of dialect, but the relation between the
factors of language and dialect remains unclear. Linguistically, we know many features of language impairment
overlap with AA dialect. We therefore propose a secondary data analysis of a large sample of AA children (n =
890) who have been administered items of two standardized English language tests, and from whom were
obtained naturalistic language samples. Accordingly, this project has three specific aims:
1. Evaluate the psychometrics of the DELV items for the extent to which they measure dialect-relevant
features versus mainstream English features. We hypothesize that like prior investigations and our own
pilot analyses, that the items will conform to multiple task-specific factors, indicating diverse aspects of
language performance.
2. Evaluate the dialect-relevant item responses on a standardized measure of mainstream English, the
Test of Language Development. We hypothesize that the items of the TOLD will show strong validity for
measuring understanding of mainstream English and that dialect responses will show reasonable validity
for measuring the linguistic rules represented by AA dialect.
3. Compare the developed factors of language proficiency and/or dialect from the DELV and the TOLD
to naturalistic language measures. We hypothesize that measures of mainstream English proficiency will
show high convergent validity and that measures of dialect may cluster in rule-specific subskills, useful to
diagnose language disorder in dialect-speaking AA children.
