The aim of the current study was to examine the influence of visual feedback on compensatory variations in force within and between trials. In Experiment 1, the task was to maintain a constant force of 5 N for 15 s using both index fingers. In Experiment 2, the task was to produce discrete force pulses such that the peak value of the sum of the two finger forces was 5 N. In both experiments, there were three conditions that manipulated the amount of visual feedback of the force trace. Results showed that the within-trial correlations between the finger forces increased when feedback was degraded, indicating that the two fingers were increasingly constrained toward acting as a single unit. Similarly, between-trial correlations showed that the amount of error compensation decreased when feedback was removed, indicating that participants produced a smaller set of goal equivalent solutions. It appears that feedback has a dual role in coordination-removing the constraints on the degrees of freedom within trials, and facilitating the utilization of redundancy between trials. The distinction between these two classes of variation is central to understanding the redundancy problem in motor control.