Presentation

Samantha Wray^1,3, Suhail Matar², Alec Marantz^2,3, Linnaea Stockall⁴; ¹Dartmouth College, ²New York University, ³New York University - Abu Dhabi, ⁴Queen Mary University London

The fusiform gyrus is the seat of several functional regions specializing in pattern recognition for various visual phenomena, from faces to words. The Visual Word Form Area (VWFA) in literate individuals has been demonstrated to discriminate an array of linguistic properties of the written word, including morphological complexity (Dehaene, et al. 2005, among others). Arabic exhibits an unusual root-and-pattern system that places it at the forefront of many lexical access studies investigating morphology. Because of this, the current study focuses on in-progress efforts to develop and validate a functional localizer in VWFA for reading Arabic text. Arabic is written using an abjad, an orthographic script that optionally indicates vowels with additional diacritics, and is written right-to-left. Arabic also exhibits an extreme heterogeneity of letter shapes; each character exhibits four different word-position-dependent forms. Additionally, the writing system typically connects letters in a form of cursive, though some letter combinations do not connect. Previous work probing Arabic readers’ sensitivity to this last property has implicated the VWFA (Taha et al. 2013). N=12 right-handed native Arabic speakers (aged 19-31, M=23.7) participated in a passive looking task, with concurrent MEG recording using a 1000 Hz sample rate on a 208-channel axial gradiometer system. Duration of the task was approximately 10 minutes. Eye blinks, heartbeat, and other motor artefacts were identified and removed using independent component analysis (ICA). Each trial displayed an image with the following properties for 200ms: (a) string type (non-linguistic symbol vs. letter/word); (b) length (singleton vs. length of 3); (c) noise mask (Gaussian level 1 vs. Gaussian level 24). Words were monomorphemic and did not include diacritics given they have been shown to inhibit reading speed (Asadi 2017) and recruit insula and inferior frontal regions (Bourisly et al. 2013). Materials were based on a design for English from Gwilliams et al. (2016), in turn from a localizer for Finnish text (Tarkiainen et al. 1999). Both English and Finnish writing systems enjoy a wealth of widely-used monospace fonts in which letters are uniform in their space allocation, easing the control of stimulus properties with respect to both size of letters compared to each other and compared to symbol strings; Arabic does not. Deviating from the English and Finnish designs, symbols were placed in close proximity, with random two symbol pairs touching to mirror the effect of cursive writing. A spatiotemporal permutation cluster analysis was performed on source estimates of the MEG data (Maris and Oostenveld, 2007) in bilateral occipitotemporal regions searching over a 120–200ms time window, based on previous M/EEG research identifying components related to the written word such as the M130, N170, and M170. A regression for each stimulus property above was fit at each time and source point. A cluster for “Type Two” (i.e. linguistically sensitive, Tarkiainen et al. 1999) response for level of Gaussian mask was identified (p=0.056). These preliminary results cement the role of VWFA in discriminating linguistic properties in early visual word recognition further, this time for a typologically distinct orthographic system.

Topic Areas: Reading, Writing and Spelling