Original paper says no better then random... (Score 2, Informative)

You should definately read the original paper "When what you type isn't what they read: The perseverance of stereotypes and expectancies over e-mail", it has a lot of interesting stuff in it.

If you read it you'll find a mistake that showed up in the Wired piece. People in their experiments didn't have the a 50/50 chance of detecting emotional tone -- instead, the chance of picking correctly the intent was no better then random chance. A much more interesting interpretation than 50/50.

There is a long history of academic research substantiating Eply/Kruger thesis that we don't interpret the emotional content (or as they call it, para-linguistic content) of text very well. The first academic paper that I've found that deals with this topic goes back to:

http://citeseer.ist.psu.edu/context/1589611/0 Sproull, L. and Kiesler, S. 1988. Reducing Social Context Clues: Electronic Mail in Organizational Communication. Readings in Computer-Supported Cooperative Work, 684--712. Los Altos, California: Morgan Kaufmann.

I've written more about this topic and other sources for the cycle of flames in my blog at Flames: Emotional Amplification of Text.