Many stimuli used to measure individual differences in cognitive abilities require participants to process linguistic stimuli as a matter of course. Although these stimuli may be carefully normed with respect to statistical properties that influence language processing, recent analyses demonstrate large, systematic variability in language usage by age, gender, race, region, and even political views. The goal of this proposal is to develop a technique for generating appropriate stimuli for participants by sampling corpora in a way that approximates their reported experience with language, and then generating stimulus sets that are matched according to these personalized statistics.
In Aim 1, we will survey a large, diverse population regarding their media habits and use this information to identify coherent patterns of variation in language experience. We will then examine stimuli from standardized tests of working memory and speech perception in noise for systematic bias by comparing lexical and syntactic statistics as computed across groups of participants, and test for the impact of potential bias on overall test performance and reliability.
In Aim 2, we will develop techniques for sampling this ?corpus of corpora? in order to approximate the language experience of a participant on the basis of a short questionnaire.
In Aim 3, we will apply this approach in a pilot study using the reading span and speech in noise tests. We predict that these tests will greater reliability when stimuli are normed based on personalized corpus. We further predict better performance for speakers of varieties that less closely approximate mainstream American English, and for less participants disadvantaged with respect to education, mobility, and/or participants who overwhelmingly consume media targeted to a specific audience. Finally, the results of this research will be provided as an open source research tool that can be used to address related issues across a wide variety of tests using linguistic stimuli.
Current standards for evaluation of language disability urge caution in distinguishing between language difference ? i.e., differences in lexical and grammatical properties of the same language as spoken by distinct communities, based on region, race, ethnicity, or other factors ? and language disability. Many tests used to measure individual differences in cognitive processing, however, use language stimuli that are standardized and normed without consideration for language differences. We propose to develop and pilot a technique that, if successful, could be applied generally to reduce the potential impact of language difference in measures of cognitive abilities.
