Although neurally based units of action (neuromotor synergies) have often been proposed as a possible basis for coordinated intentional as well as automatic actions, the idea has rarely been translated into sets of testable hypotheses. This essay examines four issues which should facilitate the development of such hypotheses: (a) definitions of neuromotor synergies, (b) criteria for recognizing and comparing synergies in automatic and intentional actions, (c) problems in representing systems of synergies, and (d) models for generating intentional actions from sets of neuromotor synergies. Limitations of, and support for the neuromotor synergy hypothesis are discussed, both in general and for the specific cases of postural synergies and cervico-spinal reflexes. Although current data do not provide conclusive support for or against the neuromotor synergy hypothesis, the problem can be formulated in ways open to experimental investigation.