Some foundational assumptions for DCM

The foundational constraints or initial conditions on which DCM is built can all be traced to the topology of space-time and to human as an embodied and embedded system

One Action at a Time

“One F/G at a time” implies “one POV at a time,” which, for an organism embodied in its environment, translates into the fact that only one action can be controllably performed at a time—or, more precisely, that a single agent cannot engage simultaneously in two actions that are not congruent or compatible with each other. For example, RELEASE, MOV, and TRADE are all congruent and can therefore be executed simultaneously: although they are linked, they are related by composition (thus co-oriented). Likewise, CUSTODY and RELEASE are congruent because they are unlinked (their vectors are in different spaces).
Different actions, however, require different framings, which may be conflicting or mutually exclusive if attempted simultaneously.

This limitation arises from being embodied in a body with specific limbs and muscles that have flexible but predefined degrees of freedom, and which is itself embedded in a three-dimensional world. This bodily space of possible actions shapes how action is cognized, while the dynamics of the body also shape the dynamics of the brain. From an evolutionary perspective, it would not have made sense to expend resources on speeding up a nervous system capable of operating orders of magnitude faster than the body can move. Accordingly, the motor neuronal system only needs to run fast enough to control muscles, which have a limited contraction speed.

Since each action requires its own specific motor program, having a fixed number of limbs that can move at fixed speeds means that the body can efficiently control only one action at a time. The deployment of a specific action schema—engaging neural assemblies in the brain—frames both the organism’s body and the world as perceived by the organism in a way that suits that action. This deployment orients the organism toward achieving a specific goal attainable through that action. Thus, it is impossible for an organism to orient itself simultaneously in different directions (for example, to pick a cup and to throw a stone at the same time).

Human cognition is limited to deploying one figure-ground construal at a time, or more specifically, to deploying one vantage at a time on a given target. This follows from the fundamental space-time constraint : an object cannot be at multiple locations simultaneously (as observed from a single perspective).

Note that what the space itself constitutes is also part of the vantage (construed space). For example, when location becomes abstracted away in one space, it may still be localized in another. When we abstract away the identity of a BUYER, the BUYER is still localized within the TRADE frame as some agent. But for the action the TRADE defines, the identity does not matter—what matters is that the BUYER is distinct from PRODUCT and CASH and that it is releasing the latter and capturing the former.

All action is local. This principle can also be traced to the topology of space-time. All action on physical objects is local—no "spooky action at a distance" is allowed. (There may be cases construed as action at a distance, which still qualify as instances of direct perception, but these will be discussed elsewhere.)

Furthermore, “local” means that an action must be resolved with respect to all its participants (localized within the frame’s space): the source of force (or cause)—an agent—and the target to which the force is applied—the patient (in semantic role terms).

Because DCM is organized in spaces, an agent has immediate access to any of its subspaces but not to anything outside of them. Thus, the scope of action for each agent is limited to its subspaces—in this sense, the action of each agent is local.

The containment relation between subspaces serves as the localization criterion, specifying which actions are executable(i.e. local) and which are virtual and therefore cannot be executed directly, requiring a realisation (i.e. model-to-realisation relation).

In other words, if an agent cannot localize itself in space-time, it cannot deterministically relate to the target of its action, making the action impossible to execute.
The agent (subject) needs to be co-embedded with the patient (object), and the position of the agent within this embedding space must be resolved. Otherwise, the position of the agent relative to the patient remains indeterminate. And having an indeterminate relation to the target of an action renders the action itself non-executable. For example: if I don’t know where I am in a room, how can I grab the cup on the table—an act that requires not only locating but also orienting my body towards  the cup?