Abstract
Contemporary thinking is that visual short-term memory of absolute object location is rather imprecise. We investigated whether humans use a target's remembered location to help plan reaching movements when visual information about the target is available, and whether the extent to which they rely on remembered and visual information is optimal; that is, is predicted by the relative reliabilities of the two sources of information.
Subjects sequentially picked up and moved two different, virtual, “magnetic” target objects from a target region into a virtual trash bin with their index fingers. Targets could be of high or low contrast. In some trials, the position of the second target was shifted 1cm while the finger was transporting the first target to the trash (masked by screen flicker, so subjects were unaware of the shift). We regressed subjects' initial movement trajectories against the initial and shifted locations to infer the weights that subjects gave to remembered and visual location information. In a second experiment, we measured the reliability of vision and memory information in the task by adding conditions in which the second target only appeared after subjects put the first target into the trash (vision-only condition) and in which the second target was initially present, but disappeared during the transport of the first target (memory-only condition).
When both visual and remembered location information were available, subjects' initial movement trajectories were biased to the remembered target location. Moreover, they gave more weight to remembered location for the low contrast targets. The observed variances in the memory and vision only conditions of the second experiment are consistent with the different weights that subjects gave to memory and visual information in the low and high contrast conditions, suggesting that memory is used near-optimally to help plan pointing movements.