PROJECT SUMMARY
In stroke-induced aphasia, deficits in auditory comprehension are prevalent, which has negative consequences
for everyday communication and quality of life, as well as speech-language assessment and treatment.
Auditory comprehension is not a monolithic process, but rather relies on processing multidimensional acoustic
and linguistic cues contained in spoken language. There are few clinical tools available for identifying level(s)
of auditory comprehension deficit(s) in aphasia, with typical assessments limited to single words and
sentences. While such assessments may allow for precise identification of where deficits exist, they lack
ecological validity. Moreover, these tasks require overt responses which utilize cognitive resources beyond
those required for auditory comprehension, impeding precise characterization of deficits. Recent computational
advances allow for objective examination of neural correlates of auditory comprehension across levels of
processing (e.g., spectrotemporal, lexical) using an ecologically valid task wherein individuals listen to
continuous speech while electroencephalography (EEG) responses are collected, with no overt responses
required. This approach, temporal response function (TRF) modeling, involves fitting a linear function to map
multivariate features of the continuous speech stimulus (e.g., spectrogram, lexical frequency) onto the EEG
data. The resulting TRF is used to derive a predicted EEG, and the relation between the TRF-predicted EEG
and observed EEG provides a measure of the fidelity of neural processing of that feature. TRF modeling has
shown promise for use in clinical populations. However, researchers have yet to assess validity and test-retest
reliability of TRF-derived measures of auditory comprehension in individuals with language disorders, severely
limiting their clinical utility. The proposed study thus has two specific aims: to examine the validity (Aim 1) and
test-retest reliability (Aim 2) of TRF-derived measures of auditory comprehension from the level of
spectrotemporal processing through semantic and syntactic processing. To this end, 40 individuals with stroke-
induced aphasia and 40 older adult control participants will complete a comprehensive cognitive-linguistic
battery comprising tightly controlled tasks and standardized assessments designed to measure different levels
of cognitive-linguistic processing. They will also listen to a continuous narrative while EEG responses are
recorded at two timepoints. TRF modeling will be used to derive measures reflecting neural correlates of
auditory comprehension which will be compared to performance on the cognitive-linguistic battery (Aim 1) and
across the two timepoints (Aim 2). The proposed study has the potential to improve characterization of auditory
comprehension in aphasia. Moreover, knowledge of normal variability across sessions for TRF-derived
measures will help researchers to make informed inferences about treatment-related outcomes or
spontaneous recovery so that test-retest variability is not mistakenly attributed to meaningful change.