How easily people can communicate. We listen and speak without the slightest sense of what's involved! One of us expresses an idea, the other understands it, and neither thinks anything complicated has happened; it seems as natural to talk as it is to walk. Yet both simplicities are illusions. To walk, you must engage a vast array of agencies to move your body down the street. To talk, you must engage a vast array of agencies to build new structures in another person's mind. But how do you know just what to say to affect that other person's agencies?
Let's suppose that Mary wants to tell Jack something. This means there is a certain structure p somewhere inside the network of Mary's agencies — and that Mary's language-agency must construct a similar structure inside Jack's mind. To do this, Mary will need to speak words that will activate appropriate activities inside Jack's agencies, then correctly link them together. How can she do that? Here is what we'll call the re-duplication theory of how we formulate what we say:
Mary proceeds, step by step, to construct a new version of p — call it q — inside her own mind. In doing this, she will apply various memory-control operations to activate certain isonomes and polynemes. As Mary performs each internal operation, her speech-agency selects certain corresponding verbal expressions — and these cause similar operations to occur inside Jack. As a result, Jack builds a structure similar to q. To be able to do that, Mary must have learned at least one expressive technique that corresponds to each frequently used mental operation. And Jack must have learned to recognize those expressive techniques — we'll call them grammar-tactics — and to use them to activate some corresponding isonomes and polynemes.
To build her new version of p, Mary could employ a goal-achieving scheme: she keeps comparing p with the latest version of q, and whenever she senses a significant difference, she applies some operation to q that removes or reduces the difference. For example, if Mary notices that p has an Origin pronome where q lacks one, her memory-control system will focus on p's Origin. In this case, if p itself is a motion frame, the usual speech-tactic is to use the word from. Next she must describe the substructure attached to p's Origin pronome. If this were a simple polyneme like Boston, Mary's speech-agency could simply pronounce the corresponding word. But if that pronome is assigned to some more complicated structure, such as an entire frame, Mary's language-agency must interrupt itself to copy that. This is expressed, as we have seen, by using words like who or which. In any case, Mary continues this difference- duplication process until she senses no significant discrepancies between q and p. Of course, what Mary finds significant depends on what she wants to say.
This re-duplication theory of speech describes only the first stages of how we use language. In later stages, the mental operations we use to construct q are not always immediately applied to pronouncing words. Instead, we learn techniques for storing sequences of grammar-tactics temporarily; this makes it possible to modify and rearrange our words and sentences before we say them. Learning these arts takes a long time: most children need a decade or more to complete their language-systems and many keep learning, throughout their lives, to sense new sorts of discrepancies and discover ways to express them.