Mitch Ratcliffe, over at BuzzLogic, posts about a number of aspects of online influence. Part of BuzzLogic's approach is to do with the changes in influence over time. One thing that I don't believe is captured in Mitch's post is the distinction between atopical influencers and topical influencers. In other words, there are people who have high standing in the blogosphere, and when they post on a topic that isn't even in their domain, they have huge influence. A canonical example, of course, is Jeff Jarvis and the whole Dell Hell saga. Part of what I believe is going on there is that certain broad topics in the blogosphere are connected to the rest of us as a whole in such a way as to promote the influence of the bloggers. I've posted a number of times some experiments that suggest this type of thing (here is an example).
The notion of topical influence captures the idea that within a particular domain, that person is influential. Engadget might be a good example of this.
These are just two aspects of influence - there are many more.
It is always pertinent, when discussing influence - and influence based analysis, that one should maintain two models of the blogosphere. The first is the networked, influence rich model (which BuzzLogic is promoting). The second is the data driven, signal based model. In this second model, rather than Jeff Jarvis being the sole type of threat to a brand, 1, 000 LiveJournalers who each express independently some dissatisfaction with a product, a movie, etc. within the same timeframe are equally important. There is a signal in the data as a whole that is of vital importance to the brand being monitored.
An illustration of something of this view is the discussion around The Illusionist. In this case, there is no clear online influencer out there shaping the reaction to the movie - what we are seeing is a reflection of offline activity manifesting as a record of individuals' experiences and their propensity to recommend the movie. Monitoring this type of thing requires comprehensive data coverage.
Finally, I would also comment that it is sometimes the story or message that has influence - this is something closer to a true meme - an idea that due to some intrinsic quality will make itself spread between nodes in the network.