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

Poster B38, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Neural Correlates of Semantic Coherence in English and Chinese Speakers during Natural Language Comprehension

Jixing Li1, Christophe Pallier2, Yiming Yang3, John Hale1;1Cornell University, 2INSERM-CEA Cognitive Neuroimaging Unit, 3Jiangsu Normal University

Introduction: Semantically incoherent words influences sentence comprehension. For example, in the sentence “He spread the warm bread with socks”, although “socks” is grammatically correct (i.e., noun), it hinders comprehension and elicits a large N400 effect (Federmeier & Kutas, 1999). Despite a large body of work on semantic expectation in sentence comprehension (e.g., Brothers et al. 2015; Federmeier & Kutas 1999; Lau et al. 2013), the neural correlates of semantic coherence remain unclear. It is also unknown whether there would be a cross-linguistic difference in the location of semantic coherence in the brain. The current study correlated time-series predictions from semantic coherence and fMRI data recorded while both English and Chinese participants listened to a same story in the scanner. The results showed that both English and Chinese speakers activated the precuneous cortex in response to semantic coherence, consistent with previous studies reporting the precuneous cortex for semantic processing (e.g., Binder et al. 2009; Leshinskaya et al. 2017; Lundstrom et al. 2005). Methods: 31 English speakers (20 female, mean age = 21.5) and 33 Chinese speakers (16 female, mean age = 20.7) listened to the whole audiobook of "The Little Prince" (in English or Chinese translation, respectively) for about 100 minutes across nine sections, and completed four quiz questions after each section. BOLD functional scans were acquired using a multi-echo planar imaging (ME-EPI) sequence with online reconstruction (TR=2000 ms; TE’s=12.8, 27.5, 43 ms; FA=77 ; matrix size=72 x 72; FOV=240.0 mm x 240.0 mm; 2 x image acceleration; 33 axial slices, voxel size=3.75 x 3.75 x 3.8 mm). Preprocessing was carried out with AFNI version 16 and ME-ICA (Kundu et al., 2011). Semantic vector representations of contents words in the story were generated by Mikolov et al.’s (2013) skipgram model. The training corpora for the English and Chinese models are all the English articles (n=4275675) and Chinese articles (n=744150) on Wikipedia. We then quantified semantic coherence between a word and its previous context as the negative cosine between the word and the mean of the previous 10 words’ vectors. Higher negative cosine value indicates lower semantic coherence. We also included four nuisance variables into the GLM analysis, performed using SPM12: word rate marks the offset of each spoken word in time; word frequency gives the log-frequency of each word in movie subtitles (Brysbaert & New, 2009; Cai & Brysbaert, 2010); intensity and f0 are marked at every 10 ms of the audio. These regressors were not orthogonalized. Results: Semantic coherence is associated with the precuneous cortex for both English and Chinese speakers (p < 0.05 FWE, k > 20). Chinese speakers showed additional bilateral anterior superior temporal gyrus (aSTG) activation. Direct comparison of the two groups confirmed stronger activity in the left aSTG for Chinese speakers than for English speakers (see Table 1 and Figure 1). Conclusion: We provided cross-linguistic evidence for semantic coherence in the precuneous cortex, suggesting that semantic coherence is a cognitively different process from other semantic processes, such as semantic combination (Bemis & Pylkkänen, 2011).

Topic Area: Meaning: Lexical Semantics

Back to Poster Schedule