I created a couple of sociograms on an IRC-channel I frequently visit (#knageroe on irc.knageroe.nl).
Who speeks with whom
This one graphs who speeks with whom most frequently. This is measured by counting the lines that start with 'nick: ...'. This graph is an heavy filtered version of the graph. It only shows the most frequent chatters. This graph is the same but with less filtering. Unfortunately creating a graph without any filtering is not possible due to time (I use Graphviz to draw these pictures and the big graph already takes half an hour to render) and resource (rendering uses loads of memory) constrains.
In the #knageroe irc-channel one can give users points indicating how popular they are. E.g. user++ means the user is popular, user-- is the opposite. This graph shows who likes who.
The #knageroe channel has an IRC bot which understands all kinds of commands. One of them is "summon". If a user has registered his e-mail address and/or mobile phone number, an other user can then request his presence using the summon-command. This graph shows who summons who.