1. Tidwell LC, Walther JB. Computer-Mediated Communication Effects on Disclosure, Impressions, and Interpersonal Evaluations: Getting to Know One Another a Bit at a Time. Human Communication Research. 2002;28(3):317-348. Available at: http://doi.wiley.com/10.1111/j.1468-2958.2002.tb00811.x.
Useful paper describing 3 different theories of how people interact and an experiment to try to find out which might be right!
Social information processing (SIP) theory (Walther 1992) suggests that people adapt to a lack of nonverbal cues to the remaining cues available such as content and linguistic strategies. Also related to a hyperpersonal perspective of computer-mediated communication, which argues that in the absence of nonverbal cues people form deeper (not broader) impressions of each other, suggesting that detailed information is shared on a smaller range of areas.
Social identity and deindividaution (SIDE) theory (Lea & Spears, 1992) suggest that rather than a hyperpersonal perspective, the absence of nonverbal cues means that the person is reduced to forming impressions based on social categories of the communicators. This reduces the individual aspect, and increases the group image.
Uncertainty Reduction Theory (URT) (Berger & Calabrese, 1975) sounds pretty interesting. This paper suggests that it hasn't really been evaluated for online (computer-mediated) stuff, but this paper was published in 2002 so worth following up on that. 7 strategies people employ to reduce uncertainty about the people you deal with. I think it's worth following up on this.
The methodology seems really pretty labour-intensive and hardcore, but could be useful to bear in mind. They use a panel of three judges to encode the text they get back - I think that would be outside the scope of my work but useful to note that the lack of separate judges may introduce bias. Also the difficulty of trying to get the same number of phrases in face-to-face and computer-mediated situations. They allowed 4 times as much time for the email conversation as for the face to face one, and still the number of utterances in the face-to-face experiments was twice that of the online conversation. I guess length of utterance would be worth looking at, plus they were using an email system which may have had a little lag. Also the interface may have encouraged longer messages, rather than the short ones people type in Skype (for example).
They seem to find in favour of the hyperpersonal model for CMC - people asked more direct questions over email and found out more about each other, but on a comparatively limited range of subjects. It would be interesting to see if this finding has been ported to MMOGs yet, with their avatars and guilds and 3-d representations, or not.