You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster D13, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Test-retest reliability of language evoked potentials

Matthew Walenski1, Elena Barbieri1, Brianne Dougherty1, Cynthia K. Thompson1,2,3;1Department of Communication Sciences and Disorders, Northwestern University, Evanston, IL, USA, 2Cognitive Neurology and Alzheimer’s Disease Center, Northwestern University, Evanston, IL, USA, 3Department of Neurology, Northwestern University, Evanston, IL, USA

Introduction. We examine the test-retest reliability of ERP components elicited in event-related potential (ERP) studies of sentence processing, the: left anterior negativity (LAN), N400 and P600. If these components are to be used as markers of intervention effectiveness – for example, assessing the efficacy of treatment interventions in individuals with language disorders – it is important to measure their stability over time in order to separate intervention effects from day-to-day variation. No prior published studies have reported on the reliability of these components in sentence comprehension. Method. We recorded scalp EEG from 32 electrodes while participants listened to grammatical passive sentences (The cook was scolded by the waiter on opening night), sentences with a semantic violation expected to elicit an N400 (The cook was printed by the waiter on opening night), and sentences with a syntactic violation expected to elicit LAN/P600 effects (The cook will scolded by the waiter on opening night). Nineteen healthy college-age participants were each tested twice on the same materials, in two sessions roughly one week apart. Data from 4 participants were excluded due to high rates of epochs with artifacts (> 25%) in one or both sessions. Thirty-eight sentences of each type were presented, along with 60 filler sentences (i.e., grammatical future tense sentences, ensuring ‘will’ did not serve as a cue to grammaticality) and active versions of the same 3 sentence types. Event-related potentials were time-locked to the onset of the verb in all experimental conditions. For each violation, we examined effects relative to the grammatical baseline in two time windows of interest: 450-700 ms for the LAN and N400, and 700 – 950 ms for the P600. To examine test-retest reliability across the 2 test sessions, we computed intra-class correlation (ICC) coefficients for each sentence type (grammatical, semantic violation, syntactic violation) at the electrode with the peak ERP effect for each violation in each time window. These coefficients have a maximum value of 1, with scores above .75 indicative of excellent reliability across sessions. Results. The semantic violations elicited a left-lateralized centro-parietal negativity in the early time window, which was maximal at electrode CP1. ICC scores were fair for the violation (.429), but poor for the grammatical sentence (.327). The syntactic violations elicited a left anterior negativity in the early time window, which was maximal at electrode FC5. This effect showed poor reliability for the violation (.281), but fair for the grammatical baseline (.548). In the late time window a significant positive effect was found, maximal at FC2. For this effect, reliability was excellent for the violation (.812), and fair for the grammatical baseline (.568). Conclusions. Reliability was generally higher for violations than for grammatical sentences, and was strongest for the P600 in the late time window. The results should inform analyses of intervention effects for these components. When ICC values are low, ICC values could be included as a covariate (e.g., ANCOVA), or alternate analyses could be pursued (e.g., structural equation modeling) to control for inter-session variability.

Topic Area: Grammar: Syntax

Back to Poster Schedule