Reviewers use different roles or points of view when reviewing. For example, reviewing as a tester, reviewing as a designer, etc. Each role has scenarios that include questions and activities that tell the reader how to review.
I think this has also been experimentally shown to be more effective than DefectBasedReading.