Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Morphological decomposition in bilateral fusiform areas in Bangla: An MEG study.
Poster A38 in Poster Session A, Tuesday, October 24, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Swarnendu Moitra1, Dustin A. Chacón2, Linnaea Stockall1; 1Queen Mary University of London, 2University of Georgia
Reading morphologically complex words involves identifying the word's constituent morphemes (Rastle & Davis 2008). MEG evidence suggests that these processes occur in the first 200ms of reading, localizing to the left fusiform gyrus (Solomyak & Marantz 2011). However, this work predominantly relies on European languages written in the Roman script. Here, we show evidence of morphological decomposition in bilateral fusiform gyri in Bangla, an understudied language written in an abugida. We develop a functional localizer to identify word-specific brain responses, and demonstrate that stem-to-word transition probability correlates with MEG activity in this region. [METHODS] N=22 Bangla speakers participated in two tasks. The first task involved reading 50 Bangla words and 50 non-linguistic symbols, embedded in two noise levels (Gwilliams, Lewis, & Marantz 2018; Tarkiainen, Helenius, Hansen, Cornelissen, & Salmelin 1999). The second was a lexical decision task, involving reading 152 grammatical and 152 ungrammatical morphologically complex words. Brain signals were recorded with a 208 axial gradiometer MEG. [RESULTS] For data from both tasks, we conducted two-stage regression analyses, fitting a regression to each time and source point. We identified significant clusters using spatio-temporal cluster-based permutation tests, computed on t-statistic from one-sided t-tests on regression beta values. Regressions for the Localizer task consisted of String Type (words, symbols) and Noise Level (low, high), spatio-temporal clustering was conducted in left and right occipitotemporal regions, 130-180ms (~"M170" time window) and 150–200ms and 170-220ms (based on visual identification of peaks in sensor data). Marginal effect of Noise was observed 130-153ms in right-posterior fusiform gyrus (p = 0.10). Effect of String Type was observed 150–182ms, in the same region (p < 0.01). Analyses of MEG data from the lexical decision task included regressions of stem-to-word transition probability, with clustering in bilateral occipitotemporal regions from 150–182ms, i.e., the same time coordinates as the String Type cluster in the localizer task, plus an exploratory 170-220ms time window. A marginally significant cluster of transition probability was identified in right-posterior fusiform gyrus, 150–178ms (p = 0.09) in the localizer time window, and 203–220ms in left middle fusiform gyrus in the later time window (p= 0.025). Further analyses were conducted for effects of stem frequency and whole-word frequency in 200–500ms time window, in the left temporal lobe. Significant clusters of whole word frequency were identified in left anterior superior temporal regions, 449–500ms (p = 0.02). [CONCLUSION] Our results are consistent with Neophytou, Manouilidou, Stockall, & Marantz's (2018) morphological processing model, in which visual form-based morphological decomposition occurs first in fusiform regions, followed by morphological 'recomposition' and whole-word look up in anterior temporal regions. Unlike previous findings in English and Greek, however, the visual word form-responses and morphological decomposition processes initiate in right hemisphere regions, and are subsequently detectable in left-hemisphere. Although other findings show bilateral activity in fusiform gyri for word-form responses (Tarkiainen et al. 1999) and morphological decomposition (Zweig & Pylkkänen 2009), these responses are left-lateralized in European languages. We suggest that hemisphere differences between previous findings and ours may arise from morphological properties of Bangla, or its orthography.
Topic Areas: Morphology,