Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The development of lexical access and linguistic prediction: Adapting story-based EEG tasks to understand comprehension in preschoolers
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
This poster is part of the Sandbox Series.
Briony Waite1, Anthony Yacovone1,2, Tanya Levari1, Jesse Snedeker1; 1Harvard University, 2Boston University
Introduction. Language comprehension involves rapidly retrieving the meaning of each word as a sentence unfolds. In adults, this process is facilitated by prediction, or the pre-activation of representations before they appear in the input (Kuperberg & Jaeger, 2016). This facilitation manifests as smaller N400s (an ERP index of lexical processing) to predictable relative to unpredictable words (Federmeier & Kutas, 2011). EEG research has begun using more ecologically-valid stimuli to assess lexical processing in naturalistic contexts with adults (e.g., Brennan et al., 2016) and school-aged children (Levari & Snedeker, 2024). Here, we extend these paradigms to preschoolers to better understand the development of lexical access. Only a few studies have used EEG to investigate sentence processing in this age range, and they assessed preschoolers’ neural responses to overt violations. The present study uses a story listening EEG paradigm with three-year-olds to understand the role of various word features and contextual cues during lexical access. Specifically, we ask whether the N400 is correlated with word predictability during naturalistic listening. Method. Children listen to audio recordings of picture books while EEG is recorded. We analyze the N400 to each of the 1,477 content words across centro-parietal electrodes (Cz, C3/4, Pz, P3/4, Fc1/2, Cp1/2). Pilot data exhibited a longer lasting negativity than the canonical N400 time window, thus, we analyze two time windows: earlier (350–550 ms) and later (550–850 ms). The current sample includes 12 three-year-old children (M age = 3.6, intended n = 35). We measured word predictability using a word-by-word cloze task, and collected frequency (SUBTLEXUS) and local semantic association (word2vec of each word with the prior content word) values. We control for phonological neighborhood density, concreteness, acoustic length, word position, and effects of immediately surrounding words. Preliminary results. We used linear mixed effects models to assess which word and contextual features best predict the N400 in each of our two time windows. In the first window we find significant effects of Frequency (b = 0.4, SE = 0.17, t = 2.3, p < 0.05) and Local Semantic Association (b = 0.41, SE = 0.19, t = 2.21, p < 0.05). In the later window we find significant effects of Cloze (b = 0.43, SE = 0.16, t = 2.66, p < 0.01) and Local Semantic Association (b = 0.43, SE = 0.17, t = 2.45, p < 0.05). Thus, cloze significantly predicts the amplitude of the N400, in line with adult data patterns, but the effects of frequency and semantic association remain, in contrast to adult findings. Summary. These preliminary results suggest that lexical access is sensitive to word predictability as early as age three. This indicates that preschoolers are already engaging in lexical prediction during online language comprehension. In contrast to findings with older children and adults, we also find effects of local semantic association and frequency on the N400 to words in a rich context. Lexical access appears to therefore be driven by both top-down contextual features and bottom-up word features at early stages of language acquisition.
Topic Areas: Language Development/Acquisition, Meaning: Lexical Semantics