Local and global semantic integration in an argument structure: ERP evidence from Korean

• Neurolinguistic studies have compared semantic integrations at different syntactic levels.
• We provide new ERP evidence on local and global semantic integration in Korean.
• ERPs to local semantic mismatch vary depending on the verb position.
• ERPs to global mismatch do not vary across different lexical/structural conditions.

The neural responses of Korean speakers were recorded while they read sentences that included local semantic mismatch between adjectives (A) and nouns (N) or/and global semantic mismatch between object nouns (N) and verbs (V), as well as the corresponding control sentences without any semantic anomalies. In Experiment 1 using verb-final declarative sentences (Nsubject [A–N]object V), the local A–N incongruence yielded an N400 effect at the object noun and a combination of N400 and a late negativity effect at the sentence final verb, whereas the global N–V incongruence yielded a biphasic N400 and P600 ERP pattern at the verb compared with the ERPs of same words in the control sentences respectively; in Experiment 2 using verb-initial object relative clause constructions ([Nsubject _V]rel [A–N]object …..) derived from the materials of Experiment 1, the effect of local incongruence changed notably such that not only an N400 but also an additional P600 effect was observed at the object noun, whereas the effect of the global incongruence remained largely the same (N400 and P600). Our theoretical interpretation of these results specifically focused on the reason for the P600 effects observed across different experiment conditions, which turned out to be attributable to (i) coordination of a semantic conflict, (ii) prediction disconfirmation, or (iii) argument structure processing breakdown.