Something interesting, and which is in the same direction as our working hypothesis. This study analyzed some gender-specific characteristics of computer-mediated communication, as a function of the visual aspect of the avatars ... Well, clearly we are going in that direction too (just see some of our recent papers about multimodal integration). It is in a linguistic journal, so it is longer than the papers we are used to read in some other fields, but there are some very interesting concepts ...
Palomares NA, Lee EJ (2010) Virtual gender-identity: the linguistic assimilation to gendered avatars computer-mediated communication. Journal of Language and Social Psychology, 29L5-23.
And just to promote ourselves, some of our papers which are related:
Lortie CL, Guitton MJ (2011) Social organization in virtual settings depends on proximity to human visual aspect. Computers in Human Behavior, Accepted.
Guitton MJ (2010) Cross-modal compensation between name and visual aspect in socially active avatars. Computers in Human Behavior, 26:1772-1776.