Visually guided pointing, the Müller-Lyer illusion, and the functional interpretation of the dorsal-ventral split: conclusions from 33 independent studies.

Models of the human vision propose a division of labor between vision-for-action (identified with the V1-PPT dorsal stream) and vision-for-perception (the V1-IT ventral stream). The idea has been successful in explaining a host of neuropsychological and behavioral data, but has remained controversial in predicting that visually guided actions should be immune from visual illusions. Here we evaluate this prediction by reanalyzing 33 independent studies of rapid pointing involving the Müller-Lyer or related illusions. We find that illusion effects vary widely across studies from around zero to comparable to perceptual effects. After examining several candidate factors both between and within participants, we show that almost 80% of this variability is explained well by two general concepts. The first is that the illusion has little effect when pointing is programmed from viewing the target rather than from memory. The second that the illusion effect is weakened when participants learn to selectively attend to target locations over repeated trials. These results are largely in accord with the vision-for-action vs. vision-for-perception distinction. However, they also suggest a potential involvement of learning and attentional processes during motor preparation. Whether these are specific to visuomotor mechanisms or shared with vision-for-perception remains to be established.