Poster A26, Tuesday, August 20, 2019, 10:15 am – 12:00 pm, Restaurant Hall
Disentangling syntactic and semantic components in basic adjective-noun composition: an MEG study
Arnold Kochari1,2, Ashley Lewis1,3, Herbert Schriefers1;1Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, 2Institute for Logic, Language and Computation, University of Amsterdam, 3Haskins Laboratories
The possibility to combine smaller units of meaning - e.g. words - to create new and more complex meanings - e.g., phrases and sentences - is a fundamental feature of human language. While natural language utterances are clearly more complex, we can take composition of a minimalistic phrase as a starting point for investigating brain dynamics supporting linguistic combinatory processing. Specifically, in this project we investigated semantic and syntactic composition of adjective-noun phrases using MEG data. Processing a noun in a basic compositional context (“white horse”) as opposed to a non-compositional context (“zgftr horse”) has been found to be supported by the left anterior temporal lobe (LATL) with its activity peaking at 200-250 ms after noun onset (e.g., Bemis & Pylkkänen 2011; Westerlund & Pylkkänen 2014). Such early activity reflecting compositional processing was found to be modulated by adjective type: when the adjective is scalar (meaning it depends on the noun, like “large” in “large house” vs. “large mouse”), semantic composition seems to happen later than when the adjective is intersective (like “wooden” whose meaning does not depend on the noun; Ziegler & Pylkkänen 2016). In our study, we first attempted a conceptual replication of this finding in Dutch. Our secondary goal was to isolate syntactic composition in such adjective-noun phrases, which has not yet been investigated in similar MEG studies. In Dutch, adjectives have to agree with the grammatical gender of the noun that they modify (“een klein paard” [a small horse], but “een kleine vogel” [a small bird]), and we made use of this feature. In one of the experimental conditions, instead of real adjectives, nouns were combined with nonwords that had an ending agreeing with the grammatical gender of the noun - pseudoadjectives (“een #derige paard”). In these phrases syntactic composition was possible based on morphosyntactic features but semantic composition was impossible since pseudoadjectives lacked meaning (established in a pre-test). Participants (N= 40) saw an ‘adjective’ followed by a noun word-by-word and answered a comprehension question after each trial. We presented 80 nouns in 4 adjective conditions (scalar, intersective, pseudoadjective, letter string [no composition]). We recorded MEG signals using a whole-head MEG system with 275 axial gradiometers and obtained individual MRI scans for each participant. Source activity was estimated using minimum norm estimates. We looked at the ROI (BA21) and time-windows where the original study reported effects. Compatible with the basic compositional processing taking place at 200-250 ms, we observed larger activity for compositional as opposed to non-compositional contexts, although in the scalar adjective condition rather than intersective as in the original study. However, we did not observe more LATL activity for intersective as opposed to scalar adjectives; this effect thus does not seem to be robust. Given that the previous studies often reported an effect with a slightly different ROI and time-window, our failure to replicate these effects highlights the need to do a systematic investigation of the regions that support composition in adjective-noun phrases.
Themes: Meaning: Combinatorial Semantics, Syntax
Method: Electrophysiology (MEG/EEG/ECOG)