A thought about how empirical research on lawyer-client meetings might be conducted, given that permitting a third party to be part of such a meeting risks stripping it of the lawyer-client privilege: What if these meetings were video recorded, but with no sound?
It might seem that studying a meeting without knowing what the participants are saying would be meaningless, but that isn't right. One can tell quite a lot about what's going on in an interaction without knowing any of the words being uttered. It's possible to follow the participants' body language and facial expressions; it's possible to see who interrupts whom and with what frequency; it's possible to see who speaks, and for how long at a time -- and probably more. Malcolm Gladwell, in Blink: The Power of Thinking Without Thinking (2005), reports a study finding that "[a] person watching a silent two-second video clip of a teacher he or she has never met will reach conclusions about how good that teacher is that are very similar to those of a student who has sat in the teacher's class for an entire semester." (13)
So these studies would be meaningful. Would turning the sound off do anything to preserve the attorney-client privilege? I'm not sure, but I think it would be worth looking into. It is true that a blown-up shot from the video might reveal legible words in documents the lawyer and client are looking at (as an indiscreet British spymaster recently found, at the cost of his job). It's also true that a lip-reader could probably make out what the lawyer and client are saying to each other. But every communication is subject to being intercepted, and in general we don't require lawyers and clients to take extraordinary steps against this danger in order to preserve the legal privilege protecting their conversation. So it seems to me that an argument could be made that this kind of recording (especially if done without an actual camera operator able to listen in) is consistent with lawyer-client privilege.
Probably most lawyer-client conversations, however, are not really all that sensitive; the client is not revealing dark secrets, nor is the adversary going to go to court to extract them. For the bulk of conversations, the most important effect of proposing to record with the sound off might be simply to reassure the client that his or her privacy really is being respected. That could make it easier for them to consent to the recording. So for reasons of both objective legal risk and subjective perception of privacy, the no-sound recording might be a relatively unobjectionable method of study.
Whether the reason lawyer-client meetings have been so rarely subjected to empirical study is either the problem of legal privilege, or clients' emotional sensitivity, isn't clear. Perhaps lawyers and law professors, even clinicians, just aren't very much disposed to undertake this sort of research -- for reasons that would no doubt be worth discussing in themselves. But if the reason isn't the researchers' disinclination but the reluctance of those being researched, then the no-sound recording might be a partial solution to the problem.