Poster C81, Sunday, March 26, 5:00 – 7:00 pm, Pacific Concourse
The neural representation of verbs and nouns meaning
Giulia V. Elli1, Connor Lane1, Marina Bedny1; 1Johns Hopkins University
Prior studies found that partly distinct brain regions process nouns and verbs (Martin et al., 2995; Bedny & Caramazza, 2011). Do these regions encode fine-grained distinctions among words? We asked whether noun-responsive regions are more sensitive to differences among nouns, whereas verb-responsive regions are more sensitive to differences among verbs using multivoxel pattern analysis (MVPA). Participants judged the similarity of pairs of words – verbs: sound emission (e.g. “to boom”), light emission (e.g. “to sparkle”), mouth action (e.g. “to bite”), hands action (e.g. “to caress”); nouns: birds (e.g. “the crow”), mammals (e.g. “the lion”), natural places (e.g. “the marsh”), manmade places (e.g. “the shed”). We identified regions in the left hemisphere showing larger responses for verbs – the posterior middle temporal gyrus (MTG) and inferior frontal cortex (IF) – and regions showing larger response for nouns – the inferior temporal (IT) and the inferior parietal (IP) cortices. A linear support vector machine (SVM) classifier was trained on half of the data (e.g. even runs), and then tested on the other half (e.g. odd runs). We successfully decoded among verbs and nouns in all ROIs (all p<0.01), the classifier was significantly more accurate for nouns in IP (t(12)=2.67, p<0.05) and IT (t(12)=3.51, p<0.01), and for verbs in MTG (t(12)=2.11, p=0.05). There was no difference in the classifier accuracy in IF (t(12)=1.12, p=0.29). These results suggest that verb-responsive regions (lMTG) are more sensitive to semantic differences among verbs, whereas noun-responsive regions (IP and IT) are more sensitive to distinctions among nouns.
Topic Area: LANGUAGE: Semantic