You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster D51, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Regions that preferentially respond to verbs or nouns are more sensitive to semantic differences among words in their preferred grammatical class: An MVPA fMRI study.

Giulia V. Elli1, Connor Lane1, Marina Bedny1;1Johns Hopkins University

Numerous neuropsychological and neuroimaging studies have identified cortical areas that are preferentially involved in processing verbs and nouns (Martin et al., 1995; Shapiro et al., 2000). The left middle temporal gyrus (LMTG) and the left inferior frontal gyrus (IFG) respond preferentially to verbs, whereas the left inferior parietal lobule (LIP) and the left inferior temporal cortex (LIT) are more engaged during noun processing (Shapiro & Caramazza, 2003; Bedny et al., 2008). Does this activation reflect preferential involvement in representing verbs and nouns? Or is it related to greater processing demands for one category (e.g. morphological complexity of verbs)? If these regions represent verbs and nouns, we might expect them to be more sensitive to semantic distinctions among words from their preferred class. However, the mean level of activity in a region is not sensitive to such fine-grained lexical information. We used multivoxel pattern analysis (MVPA) to investigate whether verb- and noun-preferring regions are more sensitive to distinctions among words in their preferred class. We hypothesized that LMTG and LIFG would be more sensitive to differences among verbs, whereas LIP and LIT would be more sensitive to differences among nouns. Participants (N=13, Mage=34, STDage=10) judged the semantic relatedness of pairs of words from 8 semantic categories – 4 verb categories: sound emission (“to boom”), light emission (“to sparkle”), mouth action (“to bite”), hands action (“to caress”); 4 noun categories: birds (“the crow”), mammals (“the lion”), natural places (“the marsh”), manmade places (“the shed”). We identified subject-specific regions of interest (ROIs) in the LMTG and LIFG (verbs>nouns), and in the LIP and LIT (nouns>verbs). Within each ROI, a linear support vector machine (SVM) classifier was trained on half of the data (e.g. even runs), and tested on the other half (e.g. odd runs) to decode among verbs and among nouns. We also inspected the group confusion matrices to better characterize the semantic categories’ discriminability profiles. There was a double-dissociation, as the classifier was significantly more accurate for verbs than nouns in LMTG (t(12)=2.11, p=.05), and for nouns than verbs in LIP (t(12)=2.67, p<.05) and LIT (t(12)=3.51, p<.01). There was no difference in classification accuracy between verbs and nouns in LIFG (t(12)=1.12, p=.29). However, classification performance for both verbs and nouns was significantly above chance (25%) in all ROIs (Ps<.01). In the LMTG, LIP and LIT the classifier distinguished not only the categories with the grossest differences (emission vs. action verbs, animals vs. places), but also some of the fine-grained distinctions (LMTG hand vs. mouth actions P<.05, light emission vs. sound emission P=.06; LIP and LIT manmade vs. natural places P<.05). These findings suggest that the LMTG and LIP/LIT are preferentially involved in representing verbs and nouns lexical-semantic information, respectively. By contrast, the LIFG appears to be equally involved in representing verbs and nouns and activation differences may reflect greater processing demands associated with verbs. Notably, all the studied regions are sensitive to semantic distinctions among both verbs and nouns, suggesting that selectivity within the lexical system is not all or none.

Topic Area: Meaning: Lexical Semantics

Back to Poster Schedule