I have to chime in as well. If anything I've found overuse of NOLOCK by people who don't really understand why blocking exists.

Blocking is good.

It ensures that there are not conflicting versions of data at any given time. In some instances that's unimportant and that's when NOLOCK can come in handy. But I've found unfortunately that people use locking hints as a way to artificially increase performance on already poorly written queries. NOLOCK can be good, but it should be used with a lot of forethought, IMO.

This article does a great job of showing how NOLOCK works and how you can realize performance benefits from using it. I think the article would have benefited from a good warning about the reasons for not using NOLOCK, however.