Letter identity and position coding in the parafovea

Letter position coding has been extensively examined in studies of isolated word identification, spurring the development of computational models. However, these models are largely restricted to explaining word identification in foveal vision, despite the fact that early lexical processing during reading occurs in the parafovea. We report four experiments that examined the flexibility of parafoveal letter identity and position coding using a variant of the same–different match task. Participants matched transposed- and substituted-letter strings to reference words, with the former being displayed at various retinal eccentricities for 100 ms versus 300 ms to respectively preclude or allow eye movements. The first pair of experiments demonstrated the relative difficulty of coding parafoveal letter positions as compared to their identities, as well as the standard benefit in identifying words displayed in the right visual field. The second pair of experiments further demonstrated that the location of letter-position uncertainty (i.e., transposed letters) interacts with both eccentricity and visual field. Initial letter transpositions were more easily detected in the left visual field, whereas transpositions of the final letters were more accurately detected in the right visual field. As discussed, these results are challenging for existing models of reading, which can individually account for some of our findings but not the results in their entirety.