The potential contribution of preplanned refixations to the preferred viewing location

On the basis of recent evidence for the preplanning of refixation saccades in reading-like tasks (e.g., Beauvillain, Vergilino, & Dukic, 2000), we proposed that for a proportion of words refixated during natural reading, the refixation sequence is planned before the word is fixated, and that the saccade target for these cases is near the beginning of the word. The preferred viewing location (PVL) phenomenon can then be explained as a mixture of two populations of planned saccades: those that initiate a multiple-fixation sequence and single fixations. This account supplants saccadic undershooting as the primary determinant of the PVL. We used a corpus of eye movement data to show that the probability of refixation is strongly correlated with the estimated perceptibility of the target word from the current fixation location, and that the observed (re-)fixation behavior is justified when assessed in terms of a lexical uncertainty measure. Refixations may be preplanned in order to increase the chances of successful word identification.