Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams

Test-retest reliability of TRF-derived measures of cortical tracking of speech

Poster A16 in Poster Session A, Thursday, October 6, 10:15 am - 12:00 pm EDT, Millennium Hall

Heather Dial¹, Lokesh Pugalenthi², Junyi Jessy Li²; ¹University of Houston, ²University of Texas at Austin

Introduction: Cortical tracking of speech refers to the alignment of neural oscillations to quasi-rhythmic fluctuations in speech. Temporal response function (TRF) modeling has emerged as an ecologically-valid tool for estimating cortical tracking of speech and may prove useful in clinical populations (Gnanateja et al., 2022). For example, aberrant cortical tracking of the speech envelope has been observed in individuals with hearing loss (Fuglsang et al., 2020) and logopenic variant primary progressive aphasia (Dial et al., 2021). However, the clinical utility of this approach is dependent upon its test-retest reliability. Recent work confirmed the reliability of tracking amplitude modulated noise (Cabral-Calderin & Henry, 2022), but researchers have yet to report on the reliability of TRF-derived measures of cortical tracking of the speech envelope. The current study sought to fill this gap. Methods: Nine neurotypical adults (mean age = 26.92 years, SD = 10.68) listened to 30 one-minute segments of an audiobook (Who Was Albert Einstein?) at two timepoints separated by at least one week (mean = 27.78 days, SD = 10.81). Cortical tracking of the multi-band speech envelope was estimated for neural oscillations in delta (1-4 Hz), theta (4-8 Hz), and delta-theta (1-8) bands using the mTRF toolbox (Crosse et al., 2016, time lags = -100 to 1000 ms). Cortical tracking was operationally defined as prediction accuracy of the TRF model, quantified as Pearson’s correlation between the observed and TRF-predicted EEG. Test-retest reliability for prediction accuracy was assessed using Pearson’s correlation (with 95% confidence intervals derived via bootstrapping) and Wilcoxon signed-rank tests across the two sessions. Results: Prediction accuracy (r) for the multi-band envelope was significantly above chance in delta (mean r = 0.08, SD = 0.03), theta (mean r = 0.06, SD = 0.02), and delta-theta (mean r = 0.06, SD = 0.02) bands. A strong correlation was observed across sessions in the theta band (r = 0.88, 95% CI [0.59, 0.97], p < 0.001), with weaker correlations observed in the delta (r = 0.42, 95% CI [-0.39 – 0.87], p = 0.23) and delta-theta (r = 0.55, 95% CI [-0.04, 0.93], p = 0.10) bands. Wilcoxon signed-rank tests identified no significant differences across sessions for any of the three frequency bands (p’s > 0.20). Discussion: The findings of this study indicate that theta-band tracking of the multi-band speech envelope has high test-retest reliability in neurotypical adults. Weaker evidence was provided for the reliability of delta and delta-theta band tracking of the multi-band speech envelope. This is a promising finding regarding the clinical utility of this method. That said, for populations with language disorders, this method will likely be of most value when estimating cortical tracking of linguistic features of speech. Thus, our next step is to examine the reliability of cortical tracking of linguistic features of speech, such as reported in Gillis et al. (2021). In ongoing work, we are collecting data from adults across the lifespan and individuals with stroke-induced aphasia to determine whether test-retest reliability of TRF-derived measures differs as a function of age and/or neurological disorder.

Topic Areas: Prosody, Morphology