High school content teachers — let’s take math teachers for example — aside from teaching math, part of their job is to socialize the students into the academic discourse of math. Think “talking math” along the lines of Jay Lemke’s “talking science.”
The most obvious way to do that is by modeling, that is, performing “math talk” aloud for the class, but if you do that then you’re not giving students any chance to respond. I’m not even talking about progressive pedagogy or participatory or exploratory instruction — even simple things like the nods and uh-huhs that show uptake in conversation are not available to students during large-group instruction.
So teachers can’t do “math talk” on their own, but they can’t do a real collaborative math dialogue with students either, because the students don’t know the language yet. So they compromise: they ask specific, pointed questions to which they already know the answer, and expect students to chime in with the appropriate bit of information. The content (“ideational”) function of known-information questions is to test whether students know that fact, but the interaction (“interpersonal”) function is to get students involved in the co-construction of “math talk,” as far as they are able.
If this is true, then you should be able to take a recording of a large-group interaction, including known-information questions, and show that the whole combined product of teacher and student talk is linguistically cohesive. And you should be able to look at the known-information questions in particular and identify the specific ways that they provide cohesion.