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

The neural implementation of semantic and syntactic composition: ERP evidence from adjective and verb processing

Poster D6 in Poster Session D with Social Hour, Friday, October 7, 5:30 - 7:15 pm EDT, Millennium Hall

Lia Calinescu1, Giosuè Baggio1, Gillian Ramchand2, Staale Slungård1; 1Norwegian University of Science and Technology, 2The Arctic University of Norway

Linguistic theory posits distinct combinatorial operations for predication (verb-noun) and modification (adjective-noun) structures. Previous work by Olstad et al. (2020) suggests that the two theoretical operations trigger different ERP responses. Moreover, lexico-semantic properties of words are also known to impact composition: Intersective (grey), subsective (small) and privative (fake) adjectives, when combined with a noun like elephant, differentially affect the meaning of the resulting phrase. The present study employed a novel paradigm to test whether syntactic and semantic properties of words impact the neural implementation of composition differently. The Cut-Compose paradigm compares full sentences (Compose) with sentences where composition is prevented by a naturalistic sentence boundary (Cut). This allows the comparison of the same element (a noun in our case) preceded by matched lexical material in combinatorial and non-combinatorial contexts. We conducted an EEG experiment asking whether the brain employs different mechanisms for combining nouns in predication vs. modification contexts, as well as for combining nouns with different classes of adjectives. We compared well-formed Norwegian sentences involving the composition of intersective (1.a.), subsective (2.a.) and privative (3.a.) adjectives with a noun (modification), as well as composition between verbs and nouns (4.a.) (predication) to corresponding baseline Cut conditions where the two words were part of different sentences and separated by punctuation. Cloze probabilities were matched across conditions. These examples are approximate translations of the original Norwegian the word 'elephants' was the point where composition was measured: MODIFICATION: 1.a. Intersective-adj Compose: Some birds must sit on grey elephants to clean them. 1.b. Intersective-adj Cut: Some birds have wings that are grey. Elephants are sometimes white. 2.a. Subsective-adj Compose: Both brown monkeys and small elephants life in Afrika. 2.b. Subsective-adj Cut: Monkeys are usually small. Elephants can step on them. 3.a. Privative-adj Compose: Animal-stories in childrens books depict fictional elephants that fly. 3.b. Privative-adj Cut: Animal-stories for children are fictional. Elephants are real. PREDICATION: 4.a. Verb Compose: In some countries people ride donkeys and feed elephants every day. 4.b. Verb Cut: At night people the horses feed. Elephants find food by themselves. We find different ERP responses involving modification and predication. All Compose–Cut contrasts modulate early components (N1-P2) but responses diverge later on. Verb Compose vs. Cut results in a higher N400 amplitude at the noun, while modification Compose–Cut do not significantly differ in the N400 time-window for any adjective class. The semantic environment also impacts processing of the noun differently. Composition with privative adjectives reliably modulates the N400 component compared to composition with intersective and subsective adjectives as shown by the Privative Compose vs. Intersective Compose and Privative Compose vs. Subsective Compose contrasts. Our results are novel in linking well established ERP components to theoretically posed constructs. More specifically, we show that the composition of words as neurally implemented is impacted differently by different syntactic and semantic environments, with processing costs incurred by obligatory argument saturation as well as by composition with privative adjectives that substantially alter the original denotation of the noun.

Topic Areas: Meaning: Combinatorial Semantics, Syntax

SNL Account Login

Forgot Password?
Create an Account