How learning to read visual Braille co-opts the reading brain

Abstract

Learning to read assigns linguistic value to an abstract visual code. Whether regions of the reading network tune to visual properties common to most scripts or code for more abstracted units of language remains debated. Here we investigate this question using visual Braille, a script developed for touch that does not share the typical explicit shape information of other alphabets, yet maps onto the same phonology and lexicon as other more regular scripts. First, we compared univariate responses in visual Braille readers and a naive control group and found that individually localized Visual Word Form Area (VWFA) was selectively activated for visual Braille when compared to scrambled Braille only in expert Braille readers. Multivariate analyses showed that linguistic properties can be decoded from Latin script in both groups and from Braille script in expert readers in an extended network of brain regions including the early visual cortex (V1), the lateral occipital region (LO), the VWFA and the left Posterior Temporal area (l-PosTemp). These results suggest that the tuning of an extended reading network to orthography relies more on the linguistic content of the script rather than their specific visual features (e.g. line junctions). Nevertheless, cross-scripts generalization was significantly lower than within-script decoding and failed to reveal common representations across Latin and Braille in experts in all regions except the l-PosTemp. These results suggest that V1, LO and VWFA encode orthographic representations in a script-specific manner, whereas l-PosTemp encodes abstracted linguistic information.