When people move, they organize large, abundant sets of elements (limbs, joints, digits, muscles, motor units, etc.) in a task–specific way by the central nervous system. Such organizations (synergies) ensure action stability, which is crucial given the varying internal body states and external forces. Action stability is controlled in a task-specific way. In particular, stability is reduced in a feed-forward manner (anticipatory synergy adjustment, ASA) if a person plans to perform a quick change of a salient performance variable. The importance of controlled stability for everyday movements is exemplified by studies of neurological patients who show deficits in both aspects of controlled stability: reduced stability during steady-state actions and small/delayed ASAs in preparation to a quick action. The physical approach to movement synergies has been developed using two theoretical frameworks. One of them is the idea of control with spatial referent coordinates (RCs) for salient variables. The other is the idea of intention-specific stability of redundant systems developed within the uncontrolled manifold (UCM) hypothesis. The RC and UCM hypotheses have been united into a single theory incorporating the idea of hierarchical control. This theory is able to account for results of a variety of studies that used perturbations of ongoing movements, analysis of variance across repetitive trials, and analysis of motor equivalence. Recent studies have provided links of this theory to neurophysiological structures and provide tools sensitive to impaired stability and agility of movements in patients with various neurological disorders.