The theory of Speech Acts gives us a couple of dozen descriptive categories for the things people do with words and phrases. The theory of Dialog Acts gives us a couple of dozen descriptive categories focusing specifically on the things people do to a conversation with words and phrases. Rhetorical Structure Theory (RST) and its various competitors give us a couple of dozen descriptive categories for the ways people use relations between words and phrases in framing an argument or telling a story. There are several other descriptive systems for discourse structures, such as the one used by the Penn Discourse Treebank.
Discourse analysis using such categories, though often insightful, is rarely funny. But you can make people laugh by caricaturing a text or conversation through self-referential descriptions of discourse functions and relations, abstracted away from specific content. I can think of two specific examples of this, though I'm sure that I've seen others over the years.
What brought this to mind was Chris Clarke's recent blog post, "This is the title of a typical incendiary blog post", 1/24/2010. It starts this way:
This sentence contains a provocative statement that attracts the readers’ attention, but really only has very little to do with the topic of the blog post. This sentence claims to follow logically from the first sentence, though the connection is actually rather tenuous. This sentence claims that very few people are willing to admit the obvious inference of the last two sentences, with an implication that the reader is not one of those very few people. This sentence expresses the unwillingness of the writer to be silenced despite going against the popular wisdom. This sentence is a sort of drum roll, preparing the reader for the shocking truth to be contained in the next sentence.
This sentence contains the thesis of the blog post, a trite and obvious statement cast as a dazzling and controversial insight.
Each sentence gives an abstract description of its function in managing the interaction between reader and writer, and often also a description of its relationship to preceding or following material. These are the same sorts of things provided by theories of dialog acts and rhetorical structures and so on, except that Clarke's descriptions are much finer-grained, e.g. not "preparation", but "a … drum roll preparing the reader for the shocking truth to be contained in the next sentence"; not "statement" but "a trite and obvious statement cast as a dazzling and controversial insight".
An older instance of the same sort of thing, long a favorite of mine, is this sketch from the Neo-Futurist's show Too Much Light Makes the Baby go Blind:
The sketch starts like this:
He: Statement. xxxxxStatement. xxxxxStatement. xxxxxQuestion? She: Agreement. He: Reassuring statement. xxxxxConfident statement. xxxxxConfident statement. xxxxxOverconfident statement. She: Question? He: Elaborate defensive excuse. She: Half-hearted agreement.
Again, part of the fun is the elaboration of dialogic categories — a bit later in the script, we get items like "Self-assured agreement as denial", "Extremely exaggerated elucidation", and "Attempted condescending conclusive statement". But in this case, the comedic effect depends crucially on what the actors add in their performance. The script is not particularly funny to read, at least in my opinion, but the sketch is a hoot to listen to.
In John Cleese's doubletalk neuroscience lecture, embedded below, the discourse caricature is almost entirely embodied in the performance. It's clear that he's explaining something, and it all seems to fit together somehow, but I don't think that we're getting a clear picture of relations like exemplification, concession, generalization and so on. Still, Cleese convinces us that there is a structure there, somehow implicit in the discourse particles, prosody, gesture, and posture:
I suspect that linguistic theories of discourse and dialog would be better if they gave us a language for characterizing interpersonal and rhetorical functions at the level of Clarke's caricature of a blog post, or the neo-futurists' caricature of a failed flirtation. And Cleese's performative abstraction of professorial rhetoric may expose some similar opportunities for theories of prosodic interpretation.
[Update — in addition to the several other examples cited in the comments below, there's also Spamalot's "A Song Like This".]