The automation of mechanical assembly requires analysis of both force and position information in order to monitor, guide, and guarantee the success of assembly tasks. This information may be obtained by force and position sensors on the assembly device, and used to make small corrections in the paths of parts while they are in contact and coming together. Two issues are involved: what path corrections to make, and how to make them in a stable dynamic fashion, given that such motions will alter the subsequent information. This paper presents a formal representation of vector force feedback strategies and shows that they must be expressed as control problems. Stability bounds are derived based on simple linear actuator models and sampled data control. These bounds are verified using computer simulation and laboratory test apparatus.