Project Summary/Abstract
Anomia is a core feature of aphasia, a disorder affecting 2.5–4 million Americans. Sensitive metrics are
needed to support clinical decision making, treatment research, and investigations of the cognitive
mechanisms underlying anomia. The first goal of this project is to test a novel psychometric model that will
jointly estimate anomia severity and processing speed while taking into account information that is available
during the clinical diagnostic process. The second goal is to test a key distributional assumption of response
times (RTs) during confrontation naming as a first step towards developing a chronometric model of anomia.
Most testing situations possess rich and diverse sources of information, ranging from the
psycholinguistic properties of the items to the patient’s accuracy and RTs during various tasks designed to tap
a range of cognitive-linguistic processes. Recently, unified analytical frameworks for clinical data have been
proposed that integrate multiple sources of information to enhance clinical inference. To this end, in Specific
Aim 1, a hierarchical Bayesian item response theory model will be used to simultaneously quantify anomia
severity and processing speed based on accuracy and collateral RTs during confrontation naming. To assess
the validity of a clinically useful model, proxy RTs automatically captured by readily available software and
hand-annotated RTs will be used. In Specific Aim 2, the model will be extended to infuse predictive information
about items (i.e., psycholinguistic properties) and persons (i.e., performance on other tasks) into the model to
both enhance its precision and understand the relations among its various components. In Specific Aim 3, we
will assess the fit of RTs during confrontation naming to an ex-Gaussian distribution with partial pooling using
Bayesian hierarchical modeling with random effects.
Consistent with Theme 5 in NIDCD’s 2023-2027 Strategic Plan, this proposal will generate the first
measurement model of processing speed during word retrieval, a construct crucial for successful
communication yet prohibitively costly to assess in fast-paced clinical settings. Further, the joint estimation of
speed and severity will lead to more efficient measurement, thereby decreasing barriers to adopting rigorous
psychometric practices in aphasiology and creating an IRT framework that could be easily implementable in
clinical and research applications. Harnessing the possibilities of this framework will be a critical step towards
computer adaptive testing batteries that will reach high levels of precision while minimizing testing burden for
both clinicians and patients and will ultimately allow each task to be administered dynamically and tailored
uniquely to each person with aphasia. Finally, assessing the distributional properties of RTs will allow
researchers to build in silico models of word retrieval that can account for both the underlying cognitive-
linguistic processes and the temporal dynamics of word retrieval in post-stroke anomia.
