Skip to main content

Cracking the orthographic code

Final Report Summary - O-CODE (Cracking the orthographic code)

Reading is one of the most complex skills that humans master without specific genetic predisposition, and the starting point of this complex task is orthographic processing – encoding the identities and the positions of the letters that compose the word being read. This project set out to develop, test, and refine a theoretical framework for orthographic processing that describes how visual features make contact with different types of letter representations that then provide access to whole-word representations and meaning. The hard question in orthographic processing is to understand how our brains keep track of the positions of the different letters in the word. We hypothesized the existence of location-specific letter detectors that “know” that a given letter is present at a given position relative to eye fixation. Another central hypothesis in our approach is that information about letter combinations is used in order to transform a location-specific orthographic representation into a location-invariant code. At this level of representation we know where letters are in the word independently of where the word is relative to eye fixation. Our computational modelling has revealed the superiority of this specific approach to orthographic processing compared with alternative approaches. We have identified the precise time-course of the hypothesized transformation of location-specific representations into a location-invariant orthographic code, using a combination of electrophysiological recordings (EEG) and priming methodology. We have also found empirical support for the hypothesized distinction between positionally flexible and more precise sublexical word-centered orthographic representations. The flexible code optimizes access to semantics from print by providing diagnostic information with respect to word identity, whereas the positionally more precise orthographic code facilitates graphemic and morpho-orthographic segmentation.

Learning to read is thought to involve adaptation of basic object recognition processes to the very specific characteristics of printed words. Our research has pointed to one key adaptation thought to involve a modification in the receptive field structure of location-specific letter detectors. Our research on crowding in letters and other kinds of stimuli suggests that the crowding zone is reduced for letter stimuli in order to optimize the parallel processing of letters in words. Our research on letter-in-string identification also suggests that receptive fields (that determine the crowding zone) are biased to the left for letters falling in the left visual field (for languages read from left-to-right) in order to give priority to the initial letters of words. A systematic comparison of the way we process strings of letters (typically random consonants – FHTRM) compared with other types of stimuli such as symbols (#%?@§), digits (72498), and simple shapes (e.g. star, cross, heart, etc.) – has helped identify the nature of the processing that is specific to strings of letters, and the nature of the adaptive mechanisms that are one key to the development of skilled reading behaviour. Furthermore, our investigation of orthographic processing in non-human primates (baboons) has helped bridge the link between general object processing mechanisms and the adaptation of these mechanisms to the specificities of letter string processing.

Our developmental research has allowed us to specify the developmental trajectories associated with the different types of orthographic code postulated in our theoretical framework. We have shown that the development of flexible sublexical orthographic representations is tightly related to the development of skilled reading behavior, as measured by standardized reading tests. On the other hand, our results point to a transition from two types of precise letter position coding during reading development – one type of coding involved in the slow process of phonological recoding used by beginning readers, and a more automatized process of grapheme-phoneme conversion used by more skilled readers. Finally, we have shown that the well-established initial letter advantage found in skilled readers gradually emerges during primary school education, hence supporting our hypothesis that this advantage is driven by adaptive mechanisms rather than serial processing.