Presentation

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

Sentence Context and Working Memory Load: Tracking Context Maintenance with ERPs

Poster C4 in Poster Session C, Friday, October 7, 10:15 am - 12:00 pm EDT, Millennium Hall

Megan A. Boudewyn¹, Cameron S. Carter²; ¹University of California, Santa Cruz, ²University of California, Davis

Introduction: Language comprehension relies in part on specialized processes, like word decoding, but also on non-linguistic cognitive functions such as working memory. Working memory refers to the mental representation of limited amounts of information that is currently in attentional focus. Working memory is central to language processing and comprehension, as it allows us to keep the current language context in mind, and to use that developing context representation to interpret incoming words. Electrophysiological studies of working memory in which an array of items (such as non-word letter sequences) is briefly displayed to participants and then held in memory over a delay period have observed a slow, negative-going wave (NSW) during the delay (Ruchkin et al., 1990; 1992; 1997). The amplitude of the NSW is thought to reflect the mental representation of the items in working memory, such that a larger amplitude corresponds to higher working memory load. The goal of this study was to use the NSW to investigate changes in working memory load as a function of sentence context, measured during the pause between sentences. Methods: In this study, 30 healthy adult participants listened to sentence pairs while EEG was recorded. The relative working memory load of a sentence can be quantified in a number of different ways. Here, we compared ERPs time-locked to the onset of the pause after Sentence 1 (analogous to the delay period of a working memory experiment) as a function of the following Sentence 1 characteristics: duration (Short vs. Long), predictiveness (Not Predictive vs. Predictive of a specific word), ease of reading (Low Gunning-Fog Index (Easier) vs. High Gunning-Fog Index (Harder)), and average word frequency (Higher Frequency content vs. Lower Frequency content). Previous studies have found the NSW to be broadly distributed across electrode sites, but following from the pattern observed by Ruchkin et al. 1997 in response to auditory stimuli, we focused our analyses on 8 fronto-central electrodes (F3, F4, FC1, FC2, C3, C4, FZ, CZ). Results: The results showed a significant difference in ERP amplitude during the pause between Sentence 1 and Sentence 2 as a function of duration (larger NSW for long than short Sentence 1 trials; p<0.05), predictiveness (larger NSW for Sentence 1 trials strongly predictive of a specific word than for unpredictable Sentence 1 trials; p<0.05 and ease of reading (larger NSW for relatively difficult to read Sentence 1 trials than for relatively easy to read Sentence 1 trials; p<0.05). The comparison as a function of average Sentence 1 word frequency was not statistically significant. Conclusions: These results suggest that the NSW, an ERP component typically associated with working memory maintenance during a delay period following intentional manipulations of working memory load, can be used to track changes in working memory load that result from variability in several different aspects of language context. The results demonstrate the sensitivity of this neural marker of working memory operations during the comprehension of naturally produced speech.

Topic Areas: Control, Selection, and Executive Processes, Meaning: Discourse and Pragmatics