Every discourse works on several scales. Each word you hear can change your state in a way that depends upon all the structures you have built while listening to the words that came before. Most of those structures are themselves mere transient things, which persist for only a few moments before you rearrange some of their parts and perhaps discard the rest entirely. Thus, a car might first appear as the subject of a sentence, then become a mere vehicle or instrument in the next sentence; finally, the whole scenario might be used merely to modify a personal trait of some actor in a larger scene. As a discourse proceeds, details on each scale become absorbed into larger-scale representation networks whose outlines become increasingly remote from the individual words that were used to construct them.
It would be wonderful to have a compact, self-contained theory that explains all our language-forms. But that ideal cannot be realized because words are merely the external signs of very complex processes, and there is no clear boundary between language and all the rest of what we call thinking. To be sure, the boundaries of words themselves are relatively clear, and when they have multiple meanings, our grammar-tactics can often help us to assign the proper senses to various terminals and other structures. These tactics include all sorts of inflections, prepositions, word orderings, and signals that indicate how to include one phrase inside another. We also combine words into larger expressions that range in vagueness of boundaries from compact clichés like hot dog to diffuse signals that are scarcely linked to specific words at all; these include our hard-to-describe nuances of phrasing, rhythm, intonation, and shifts of style and flow.
We're normally quite unaware of how our grammar-tactics constrain us in our choices of words. We're often somewhat more aware of other language-tactics we use to guide our listeners' minds — to change the focus from one theme to another, to adjust the levels of detail, to shift between foreground and setting. We learn to use phrases like by the way to change the topic of concern, to say for example to shift to a finer level of detail, to say but to modify an expectation or to interrupt the usual flow, or to say in any case or in spite of that to indicate the end of an interruption or elaboration.
But even all this is only a small part of language. To understand what people say, we also exploit our vast stores of common knowledge, not only about how specific words are related to the subjects of concern, but also about how to express and discuss those subjects. Every human community evolves a great array of discourse-forms to shape its stories, explanations, conversations, discussions, and styles of argument. Just as we learn grammar-forms for fitting words to sentence-frames, we also build up stocks of plots to organize our story-tales, and standard personalities to fill the roles of their protagonists — and every child must learn these forms.