Studies of saccadic suppression and induced motion have suggested sepa
rate representations of visual space for perception and visually guide
d behavior. Because these methods required stimulus motion, subjects m
ight have confounded motion and position. We separated cognitive and s
ensorimotor maps without motion of target, background, or eye, with an
''induced Roelofs effect'': a target inside an off-center frame appea
rs biased opposite the direction of the frame. A frame displayed to th
e left of a subject's center line, for example, will make a target ins
ide the frame appear farther to the right than its actual position. Th
e effect always influences perception, but in half of our subjects it
did not influence pointing. Cognitive and sensorimotor maps interacted
when the motor response was delayed; all subjects now showed a Roelof
s effect for pointing, suggesting that the motor system was being fed
from the biased cognitive map. A second experiment showed similar resu
lts when subjects made an open-ended cognitive response instead of a f
ive-alternative forced choice. Experiment 3 showed that the results we
re not due to shifts in subjects' perception of the felt straight-ahea
d position. In Experiment 4, subjects pointed to the target and judged
its location on the same trial. Both measures showed a Roelofs effect
, indicating that each trial was treated as a single event and that th
e cognitive representation was accessed to localize this event in both
response modes.