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

Poster E70, Friday, November 10, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

Learning a new alphabet: Identifying changes in neural representations

Robert W Wiley1, Brenda Rapp1;1Johns Hopkins University

Introduction: In alphabetic systems, learning to identify letters and map their identities to the sounds they represent is a key step to literacy. Recent work has investigated the neural underpinnings of various types of letter representations, both abstract and modality-specific (visual, phonological, and motoric). The longitudinal study reported here uses multivoxel pattern analysis (MVPA) to further our understanding of how different letter representations come to be instantiated in the brain by examining changes in the neural responses of adult second-language learners before and after training with Arabic letters. This fMRI study seeks to identify: 1) which brain areas show significant learning-related changes; 2) what types of letter representations are instantiated in these areas and 3) how the neural representations of Arabic letters compare to those for Roman letters. Methods: 16 native English speakers (ages 18-33) underwent a training study during which they learned to identify 34 Arabic letter-shapes representing 20 unique identities. Participants were trained in one-hour sessions until reaching criterion, or for a maximum of six sessions. Participants were scanned before and after training, completing a visual word form area (VWFA) localizer task and a symbol detection task, both of which included Arabic and Roman letter stimuli. Regions of interest (ROIs) were identified by comparing group activation from the VWFA localizer (Arabic >> checkerboards) at post vs. pre-training. MVPA was applied to the ROIs using data from the symbol detection task to identify and track different types of letter representations (visual, phonological, motoric and symbolic). Linear mixed-effects modeling (LMEM) was used to determine which types of representations best explain the observed patterns of neural activity, and to perform tests of statistical significance across regions, time-points, and alphabets. Results: 22 regions in which learning occurred were identified, indicating wide-spread neural changes in: bilateral medial and anterior ventral occipital-temporal cortex (vOTC), extrastriate cortex, Brodmann areas 44 and 45, supramarginal gyri, and left premotor and supplementary motor cortex. Among these regions, analyses revealed significant effects of time-point, hemisphere, and/or alphabet for different representational types. For example, bilateral vOTC represented visual information for Arabic letters and did so more after training, whereas this same ROI did not represent visual information for Roman letters. Additionally, there was a significant hemisphere by time-point interaction indicating an increase in the phonological/symbolic representations of Arabic letters in the left, but not the right, vOTC. Summary: The novel application of MVPA and LMEM analyses to a longitudinal letter learning study reveals that learning to identify Arabic letters: 1) produces changes in BOLD response in a widespread set of regions; 2) these regions show patterns of activity indicating the learning of multiple types of letter representations including visual, motoric, and phonological/symbolic; and 3) the regions that show the greatest change in response to learning Arabic letters do not necessarily represent the same informational content for both alphabets. The differences between the neural instantiation of representations for the two alphabets may be due either to differences in their unique properties, or the nature of letter representations at different learning stages.

Topic Area: Perception: Orthographic and Other Visual Processes

Back to Poster Schedule