Reading in the Brain Revised and Extended: Response to Comments

Reading in the Brain (Les neurones de la lecture, 2007) examined the origins of human reading abilities in the light of contemporary cognitive neuroscience. It argued that reading acquisition, in all cultures, recycles preexisting cortical circuits dedicated to invariant visual recognition, and that the organization of these circuits imposes strong constraints on the invention and cultural evolution of writing systems. In this article, seven years later, I briefly review new experimental evidence, particularly from brain imaging studies of illiterate adults, which indicates that reading acquisition invades culturally universal cortical circuits and competes with their prior function, including mirror-invariant visual recognition and face processing. In response to my critics, I emphasize how brain plasticity and brain constraints can be reconciled within the Bayesian perspective on learning. I also discuss the importance of a newly discovered gesture system in reading and writing. Finally, I argue that there is consistent evidence for deep cross-cultural universals in writing systems, as well as for the multiple subtypes of dyslexia that are expected given the broad set of areas recruited by the reading task. It is a rare privilege for an author to see his work reviewed and dissected by a panel of experts. I am therefore extremely grateful to Anne Castles, Max Coltheart, Greg Downey, Naama Friedmann and Richard Menary for their thorough critical reading of Reading in the Brain (hereafter RITB). In this reply, I first briefly review some relevant recent findings on the brain mechanisms of reading, most of which have appeared since RITB was published in 2007. Using this data as background, I then briefly respond to each of my critics. 1. Reading in the Brain: An Update RITB’s main thesis is that during reading acquisition, a specific region of the ventral visual occipito-temporal cortex, the visual word form area (VWFA), gets ‘recycled’ and specializes from the recognition of written characters. Since 2007, several fMRI studies have confirmed the central role of the VWFA in reading acquisition. In an fMRI study of adults with highly variable levels of literacy, ranging from complete illiterates to expert readers, the amount of VWFA activation to letter strings increased in direct proportion to the subjects’ reading score (Dehaene, Pegado et al., 2010). Similarly, when comparing fMRI activations to written words in six-year-old children who had or had not learned to read, the main site where a difference was seen was the VWFA (Monzalvo, Fluss, Billard, Dehaene and Dehaene-Lambertz, Address for correspondence: Inserm-CEA Cognitive Neuroimaging Unit, NeuroSpin center, CEA/SAC/DSV/I2BM Bât 145, Point Courrier 156 F-91191 Gif/Yvette, France. Email: [email protected] Mind & Language, Vol. 29, No. 3 June 2014, pp. 320–335. © 2014 John Wiley & Sons Ltd Reading in the Brain: Response to Comments 321 2012). At an even younger age, in kindergarten, Brem et al. (2010) found that a few hours of exposure to the GraphoGame, a software that trains grapheme-phoneme correspondences, was sufficient for activation to emerge around the VWFA site. In adults too, training to read a new script systematically leads to changes in the VWFA, regardless of script and language (Hashimoto and Sakai, 2004; Song, Bu, Hu, Luo and Liu, 2010; Yoncheva, Blau, Maurer and McCandliss, 2010). An essential postulate of RITB is that the VWFA participates in a universal reading circuit, the same in all cultures and which comprises both phonological and semantic routes. fMRI studies comparing English and Chinese readers have largely confirmed this hypothesis (Booth et al., 2006; Szwed, Qiao, Jobert, Dehaene and Cohen, 2014). In fact, the reading circuit is so determined that the VWFA is activated even in congenitally blind adults when they read Braille or when they identify letters through the auditory modality using a sight-to-sound sensory substitution device (Reich, Szwed, Cohen and Amedi, 2011; Striem-Amit, Cohen, Dehaene and Amedi, 2012). In RITB, I speculated that the surprisingly fixed localization of the VWFA arises from an intersection of two types of constraints: visual constraints that render this location particularly appropriate for discriminating small letter shapes, and connectivity constraints that interconnect it with the left superior temporal and inferior frontal language areas. In support of the first type of constraint, Szwed et al. (2011) found that configurations of line junctions, which are universally used in writing systems throughout the world (Changizi, Zhang, Ye and Shimojo, 2006), specifically promote activation in the ventral fusiform part of the visual system. In support of the second type of constraint, several papers have reported a consistent co-lateralization of the VWFA to the dominant hemisphere for spoken language and particularly to the superior temporal sulcus, suggesting a tight functional interconnection (Cai, Paulignan, Brysbaert, Ibarrola and Nazir, 2010; Pinel and Dehaene, 2009). Improvements in diffusion tensor imaging have recently allowed tracking the main fiber tracts that originate from the VWFA. As predicted, they specifically target left-hemispheric superior temporal and inferior frontal language areas, more so than other lateral or antero-posterior control sites such as the fusiform face area (Bouhali et al., 2014). If the VWFA is recycled for reading, what is its primary function in evolution? Much of my recent research has been targeted at elucidating this point. By scanning adult illiterates with a battery of visual stimuli, we observed that prior to reading, the VWFA site is particularly responsive to faces and to checkerboards and that these responses decrease with increasing literacy (Dehaene, Pegado, et al., 2010). Remarkably, reading ability correlates with an increased displacement of face responses: activations to faces decrease in the left hemisphere and increase in the right-hemispheric fusiform face area (FFA), suggesting that a competition for cortical space between faces and written words occurs in the left occipito-temporal sulcus. These findings, which were replicated in 6-year-old and 9-year-old children (Monzalvo et al., 2012), strongly suggest that invariant visual recognition, particularly for faces, is the prior function of the VWFA and gets progressively tuned out as this cortical region specializes for letters and their combinations (for © 2014 John Wiley & Sons Ltd