The base space (upper left space) defines the situation and states the goals of the situation. One entity interacts with the agent and requests some task. The presentation (middle) space displays the task from the base space. This is a generic plan that defines roles, relating to the kind of task activated in the base space, most likely defined by the type of verb used in the request. For example, if the verb is get, then the generic plan would involve some motion to and a grabbing. The reference space (the space on the right) is the perception, the map that the agent finds itself in. The generic plan is superimposed onto the map, blending the presentation and reference space, and creating  a specific map with the goal defined in the context of the map. However, this is not enough to create an action routine relating to the objects in the pragmatic space. In order for this to happen, there needs to be some sort of relevance mechanism.

     This mechanism defines how we can interact with the objects being discussed in the base space and what kinds of sequences it takes in order to do this. This logic mechanism, or relevance making device, fills the roles that the presentation space suggests. It defines a sequence of events that would reach that goal in this particular situation. This creates a sequened action plan, as displayed in the explanation of the logic box. This information would have to be activated based on the categorical knowledge provided in the pragmatic space. For example, getting a cup is different from getting a ball. Cups are in a cabinet, requiring a few different steps. Implementing this logic device is a topic that needs to be exlored further by us in implementing this system. Next, the blended space is carried out and the first step is accomplished.

    Elaborating on this, each step in the action plan sequence could have a predicted map, giving the agent something to compare its current results and determine if the step has been successful. Additionally, if this process is happening in time, each step would have its own mini task, a kind of subordinate blending system resulting from the action space in the initial blend. In this new subordinate system, each individual step could have a task schema, a plan outline and a current map input, blended to create the real situation and made relevant by a system defining with what speed or force to do the action. This would be a more fine-grained system for each step in the sequence. This addition would effectively be holding the overarching goal state in mind while create little mini goal states for each step of the way.