PONG: A Computational Model of Visual Word Recognition Through Bihemispheric Activation

Orthographic processing is an open problem. Decades of visual word recognition research have fueled the development of various theoretical frameworks. Although these frameworks have had good explanatory power, various recent results cannot be satisfactorily captured in any model. In order to account for old and new phenomena alike, here I present a new theory of how the brain computes letter positions. According to PONG (which describes the Positional Ordering of N-Grams), each hemisphere of the brain comprises a set of mono- and multigram detectors. The crux is that the detectors for a given N-gram are activated to different extents in their respective hemispheres, depending on where in the visual field the N-gram is located. This differential activity allows the brain to estimate the leftness or rightness of that N-gram, whereby word activation is a function of the N-gram's identity plus its laterality relative to that of other activated N-grams. Simulations with PONG suggest that the framework effectively accounts for classic phenomena, as well as newer phenomena and cross-linguistic differences that cannot be explained by other models. I also reflect on the neurophysiological plausibility of the model and avenues for future inquiry.