Just saw a truly wonderful talk by Karrie Karahalios from UIUC about the relatively-understudied role of audio in social computing. She makes a compelling case for using social computing methods for visualizing and incorporating audio within the various social software systems that people are building.
Which was interesting because last night I was playing WoW in the same room as some of my guildies and also the leader of another guild. The most remarkable thing, for me, was the amount of Teamspeak voice chatter that the other guild engaged in. Our guild only ever uses Teamspeak for instances and raids, and so to hear the other guild just chatting was intriguing.
I don't know if this is correlated with expectations about roleplay--I know that Richard has a view about roleplay and voice--although I suspect that it's not necessarily related. Our guild is utterly uninterested in roleplay, but we just don't use TeamSpeak for non-instance play. I have no idea why this is so.
I also happen to know that Dmitri has done some studies about the role of voice in various metrics of play, and I don't want to steal his thunder. No doubt he'll tell us all about it in due course. But I've never thought about it before, and I think that I've been missing something significant.