MEMORY REPRESENTATIONS UNDERLYING MOTOR COMMANDS USED DURING MANIPULATION OF COMMON AND NOVEL OBJECTS

1. While subjects lifted a variety of commonly handled objects of diff erent shapes, weights, and densities, the isometric vertical lifting f orce opposing the object's weight was recorded from an analog weight s cale, which was instrumented with high-stiffness strain gauge transduc ers. 2. The force output was scaled differently for the various object s from the first lift, before sensory information related to the objec t's weight was available. The force output was successfully specified from information in memory related to the weight of common objects, be cause only small changes in the force-rate profiles occurred across 10 consecutive lifts. This information was retrieved during a process re lated to visual identification of the target object. 3. The amount of practice necessary to appropriately scale the vertical lifting and gri p (pinch) force was also studied when novel objects (equipped with for ce transducers at the grip surfaces) of different densities were encou ntered. The mass of a test object that subjects had not seen previousl y was adjusted to either 300 or 1,000 g by inserting an appropriate ma ss in the object's base without altering its appearance. This resulted in either a density that was in the range of most common objects ( 1. 2 kg/ 1) or a density that was unusually high (4.0 kg/ 1). 4. Low vert ical-lifting and grip-force rates were used initially with the high-de nsity object, as if a lighter object had been expected. However, withi n the first few trials, the duration of the loading phase (period of i sometric force increase before lift-off) was reduced by nearly 50% and the employed force-rate profiles were targeted for the weight of the object. The force scaling was still adapted to the object's weight whe n lifting the same object 24 h later.5. In contrast, lifting a novel o bject with a more common density yielded stable loading phase duration s and peak-force rates across consecutive trials, beginning with the f irst lift. Thus, for unfamiliar objects, subjects infer the object's w eight assuming a default density that is within a range of commonly en countered densities and used in combination with size cues. 6. It is c oncluded that humans use anticipatory control to scale motor commands to the weight of familiar objects. The memory information is robust an d can be retrieved through visual identification of the target object, and accurate memory representations related to the weight of novel ob jects develop quickly. Such anticipatory control is highly purposeful because it allows quick and accurate manipulation during everyday task s that is not subject to limitations imposed by a strict dependence on ''moment-to-moment'' sensory feedback control.