Abstract
During reading, saccadic eye movements aiming at word centers are generated to shift words into the fovea for lexical processing. However, where the eyes land within words is, on average, a function of the distance of the center of the target word from the prior fixation location (i.e., the launch site). In effect, the optimal first-fixation position at the center of new words is realized only from an optimal launch-site distance, while close or distant word centers are systematically over- or undershot (McConkie et al., 1988). A similar pattern was found in saccades aiming at single targets, but the launch-site contingent shift of saccadic end points appears to be much smaller than in reading (Kapoula, 1985). Here we show that the launch-site effect in both paradigms can be quantitatively reproduced by a recently developed Bayesian model of saccade planning (Engbert & Krügel, 2010). The model is based on the assumption that subjects use Bayesian estimation for optimal computation of target position based on noisy sensory information. We demonstrate that the pronounced launch-site effect in continuous reading compared to single saccade paradigms is a consequence of less precise visual information about the location of the saccade target. Our results bridge the gap between basic oculomotor research and research on eye movements in reading. Implications for reading theories are discussed.
Meeting abstract presented at VSS 2012