A metanalysis of the effect of the Müller-Lyer illusion on saccadic eye movements: no general support for a dissociation of perception and oculomotor action.

Milner and Goodale's (1995) proposal of a functional division of labor between vision-for-perception and vision-for-action is supported by neuropsychological, brain-imaging, and psychophysical evidence. However, there remains considerable debate as to whether, as their proposal would predict, the effect of contextual illusions on vision-for-action can be dissociated from that on vision-for-perception. Meta-analytical efforts examining the effect of the Müller-Lyer (ML) illusion on pointing (Bruno, Bernardis, & Gentilucci, 2008) or grasping (Bruno & Franz, 2009) have been conducted to resolve the controversy. To complement this work, here we re-analyzed 17 papers detailing 21 independent studies investigating primary saccades to target locations that were perceptually biased by the ML illusion. Using a corrected percent illusion effect measure to compare across different studies and across experimental conditions within studies, we find that saccadic eye movements are always strongly biased by the illusion although the size of this effect can be reduced by factors such as display duration and between-trials variability in display length and orientation, possibly due to a process of saccadic adaptation. In contrast to some reports, we find no general support for differences between voluntary and reflexive saccades or between saccades performed in conjunction with a pointing movement and saccades performed without pointing. We conclude that studies on the effect of the Müller-Lyer illusion do not provide evidence for a functional dissociation between primary saccades and perception.