Beth Loring wrote a great post over at Farm on best practices in moderating usability tests.
Her webinar presentation below touches on bias in user research, how much to interact, and dealing with failure.
It's a solid, high-level resource on a topic that isn't as well-represented on the interweb as I'd like, and definitely worth checking out.
Also be sure to check out the follow-up podcast, where Beth answered some solid questions from the webinar attendees.
What do you think? Did she forget anything?