A task has been developed to investigate the ability of cats to switch the direction of an ongoing target-reaching forelimb movement with the aid of a visual cue. The cats were standing in front of two horizontal tubes (internal diameter 30 mm; shoulder level) with food. The entrances of the tubes were closed with opaque trap doors but during illumination inside a tube its trap door was unlocked allowing the cat to retrieve food with the paw. When the cats had learnt to select the illuminated tube for insertion the next step was to switch the illumination to the other tube during ongoing target-reaching. Limb lifting was performed when the light was switched on in one of the tubes and time was measured from breaking electrical contact between the paw and the floor. After 25-75 ms, illumination was shifted to the other tube and the latency to the earliest change in movement trajectory was measured. The trajectory was recorded with the aid of cameras detecting the position of infrared light emitting diodes fixed to the dorsal part of the wrist. Every 3 ms the position was fed into a computer, and the movement trajectory (horizontal and sagittal planes) was displayed graphically. The velocities in the direction of cartesian coordinates x, y and z (protraction, adduction-abduction, lifting) were also computed. Single tube trials and switching trials from either tube were made in a random series. In order to switch, the cats used a combination of braking the protraction and a sideways movement. Initially there was often some retraction of the paw to avoid hitting the trap door of the first illuminated tube, but with more proficiency braking decreased and the movement path became smoothly curved. During braking of protraction there was also deceleration of lifting but not enough to maintain a constant movement path in the sagittal plane. In sessions with single tube trials, the movement paths in the horizontal plane were reasonably straight. In sessions with intermixed switching trials the single tube paths became segmented or curved, seemingly in order to facilitate switching. The mean switching latency in four cats ranged from 83 to 118 ms. In the fastest cat the switching latency ranged from 70-106 ms.(ABSTRACT TRUNCATED AT 400 WORDS)