Proceedings of the Annual Meeting of the Cognitive Science Society

Cognitive tests used to measure individual differences are gen-erally designed with equality in mind: the same “broadly ac-ceptable” items are used for all participants. This has unknownconsequences for equity, particularly when a single set of lin-guistic stimuli are used for a diverse population of languageusers. We hypothesized that differences in language varietywould result in disparities in psycholinguistically meaningfulproperties of test items in two widely-used cognitive tasks, re-sulting in large differences in performance. As a proxy for in-dividuals’ language use, we administered a self-report surveyof media consumption. We identified two substantial clustersfrom the survey data, roughly orthogonal to a priori groups re-cruited into the study (university students and members of thesurrounding community). We found effects of both populationand cluster membership. Comparing item-wise differences be-tween the clusters’ language models did not identify specificitems driving performance differences.