Slide Slam J10 Sandbox Series
What are the sources of predictability and plausibility? A neurophysiological study of sentence comprehension in Korean
Boyoung Lee1, Say Young Kim1; 1NeuroCognition of Language Laboratory, Hanyang University, Seoul, Korea
Predictability and plausibility based on semantic context are two critical features that determine how a word is processed and utilized to construct a message-level representation of a sentence. In previous ERP studies, the two factors have generally been associated with two patterns of biphasic negativity-late positivity: an N400-frontal positivity (fPNP) for unexpected but plausible sentence continuations and an N400-posterior positivity (pPNP) for unexpected and implausible continuations (DeLong, Quante, & Kutas, 2014; Kuperberg, Brothers, & Wlotko, 2020; Van Petten & Luka, 2012). This suggests that there are distinct ERP responses to conflicts, either induced by prediction (i.e., via the N400) or plausibility (i.e., via the divergence in the late positivity distribution). However, the question of whether different sentence constituents (e.g., argument nouns, modifiers, verbs) vary in their vulnerability to these conflict-prone processing features has remained unsolved. Thus, the present ERP study is designed to answer this question by comparing ERP responses to conflicts induced by two different verb-containing constructions (i.e., noun argument-verb vs. adverb modifier-verb) in Korean sentences. Taking advantage of the verb-final word order in Korean, nouns and adverbs preceding the shared sentence-final verb are separately manipulated to generate either a prediction or plausibility violation when combined with the verb. Thus, with the manipulation of three factors (i.e., constituent construction, expectancy, and plausibility), a total of five experimental conditions can be developed: 1) Control [CON] (e.g., ‘Mary-nun (Mary-NOM) – elyewun (difficult-Adj) – sihem-ul (exam-ACC) swuepsi (numerously-Adv) – chiluko (take-V) …’ which means ‘Mary took the difficult test numerously and …’), 2) Adverb-verb unexpected congruent [A-UCNG] – replaces ‘swuepsi (numerously-Adv)’ with ‘pwuncwuhakey (busily-Adv)’), 3) Adverb-verb (unexpected) anomalous [A-ANM] – replaces ‘swuepsi (numerously-Adv)’ with ‘mwulkkulemi (blankly-Adv)’), 4) Noun-verb unexpected congruent [N-UCNG] – replaces ‘sihem-ul (exam-ACC)’ with ‘swuswulel (surgery-ACC)’, and 5) Noun-verb anomalous [N-ANM] – replaces ‘sihem-ul (exam-ACC)’ with ‘mwunceylul (problem-ACC). The switched nouns and adverbs in all unexpected sentences lead the reader to predict a verb different from the one actually encountered, producing a conflict. High contextual constraint confirmed by high cloze probabilities and (in)congruency validation through plausibility ratings will be obtained and matched between conditions before the ERP experiment. The experimental procedure will consist of a segment-by-segment presentation (SOA = 800ms), and an acceptability judgement task to maximize participants’ sensitivity to anomalies. Following the notion of N400 reduction reflecting processing benefits from previous semantic context, all four unexpected conditions are predicted to produce large N400 amplitudes. Within the unexpected conditions, there are two possible outcomes. In contrast to the noun-verb construction conflicts that are expected to yield significant negativity-positivity patterns, the adverb-verb conflicts may yield smaller, or even no late positive components. Such results would imply the more significant role of the mandatory argument-verb relation information (e.g., selectional restriction) – relative to the optional modifiers - in message-level representation development of a sentence and judgement of its overall plausibility. Alternatively, if the ERP responses to adverb-verb and noun-verb structures are comparable, it can be suggested that the construction of sentence meaning is not differentially affected by the two constructions.