The problem of motor redundancy has been one of the fundamental, albeit elusive, problems in motor control. Traditionally, it has been viewed as a computational problem for the brain, solved with either optimization methods or by introducing additional constraints to motor tasks. This review suggests that the problem was wrongly formulated, and that the abundant degrees of freedom are not to be eliminated but used to ensure dynamic stability of motor performance, which is vital given the unpredictable intrinsic states and external forces. The idea of synergies as mechanisms ensuring action stability is introduced based on the uncontrolled manifold hypothesis and the theory of control with spatial referent coordinates. The importance of controlled stability is illustrated with the phenomena of anticipatory synergy adjustments. This approach is productive for both basic and applied fields as illustrated, in particular, by changes in motor synergies with neurological disorder and exercise.