Subtleties in data collection: audio versus video

My colleagues and I typically perform data collection in the field using a variety of different methods. For instance, we often rely on diaries of various forms, system logs if we are deploying a new technology, and most importantly, recordings of our interviews and observations. These recordings can be both audio and video (e.g., from a camcorder), audio-only (e.g., a digital voice recorder), or handwritten or typed field notes.
I was recently listening to a Radiolab podcast where Ira Glass, Jad Abumrad, and Robert Krulwich were discussing the relative merits of radio and TV as media. Glass, speaking from his recent experiences producing a TV version of This American Life, made a really interesting quote:

People are able to open up on the radio too...the other producers, when they really wanted to get down with people, we would film part of the interview, and then they would just...tape part of the interview...just with a tape recorder...so we totally tricked it out to get the kinds of quotes that we wanted, because we weren't getting the quotes we wanted with the cameras rolling sometimes.

It made me curious as to whether there are measurable effects on the willingness of people to provide honest or substantive answers to questions depending on the type of device being used to record their answers. In my experience, people are clearly more self-conscious with a camera pointing at them, but how much less self-conscious are they with a tape recorder running, or with no recording equipment to speak of? Bernard's Research Methods in Anthropology devotes a few pages to video, but doesn't really discuss these effects, other than to note a few studies where subjects adjusted to having cameras pointed at them over time. Anyone out there have a few pointers to good papers on this topic?