For a participant to control a situation means that he has the power to start, continue and stop the situation. As said above, the possibility of control is, in the first place, a definitional feature of a type of situation, viz. of an act or action. On the part of the participant in question, conditions for control are fuzzier. While the prototypical controller is a human being, nothing prevents, in principle, an animal, a machine or even a celestial body from controlling a situation.

Actions are goal-directed. In the case of situations of cognition and communication, the goal is generally the solution of a problem. Goal-directedness presupposes intention; and intention presupposes consciousness. Consequently, in addition to the definitional power mentioned, prototypical control involves the following set of relational features:

The criterion of intentionality is the basis of the test frames used in the previous section. Consciousness is a polysemous word; what is susceptible of a definition is one of its senses: ‘x is conscious of y’ means that y is the object of x’s thinking in the same way as it would be if x was speaking about y. This implies that we are conscious of what we are speaking about. This is so because communication and cognition are problem-solving activities; and the solution of a problem requires thinking.

On the other hand, human beings undergo processes in which nobody or nothing but themselves are involved. For instance, they sneeze or slide. Some of these are not amenable to control, for instance purely physiological occurrences like heartbeat, digestion and dreaming. Others are amenable to control, but normally proceed without control, for instance, breathing. Uncontrolled processes happen automatically. What happens in the inanimate world are primarily processes; only if we impute control to an inanimate participant are they conceived as its actions.

Control is in many respects a gradual notion. If x forces y to act, x exerts a higher degree of control than if he asks y to act. Also, x may have the alternative of either doing z or causing y to do z. In either case, x has the highest control of z; but in the second case, his control is mediate. If x does a certain action z to achieve some goal, then he controls z. However, doing z involves subordinate steps. For instance, I want to enter a certain room. Having pressed down the handle, I pull the door towards me. In doing this, I take a step back in order to get into an appropriate position against the now open door. This latter step is, in principle, controllable; but normally it will run automatically. In the sense here relevant, it is controlled mediately. Suppose that you are standing behind me and in opening the door in the way described, I step on your foot. Although nobody assumes that I did it willfully, I am nevertheless expected to apologize, which implies I am held responsible for the event. Thus, in doing a certain action, I trigger a chain of subordinate processes which I do not monitor and which I trust will work automatically. Thus, “the contrast between conscious and automated processing is not a single discrete division, but rather a hierarchic, multi-level, scale.” (Givón 1989:258)

The two types of processing have been investigated and been established as a “dual processing theory” in psychology for a long time (s. the brief research history in Schneider & Chein 2003, §1). The literature on pedagogic psychology tends to see a dichotomy between controlled and automatic processes, where the latter are not controlled by intentions, strategies and plans. This is too simple. Processes occupy an inferior position in a hierarchy. Neumann 1984:256 characterizes thus a more adequate theory: “It conceives automatic processing not as lacking control, but as being controlled at levels below the level of conscious awareness.”