Self-diagnosis tasks aim at fostering diagnostic behavior by explicitly requiring students to present diagnosis as part of the activity of reviewing their problem solutions. The recitation classes in an introductory physics class (~200 students) were split into a control group and three experimental groups in which different levels of guidance was provided for performing the self-diagnosis activities. We have been a) investigating how students in each group performed on subsequent near and far transfer questions given as part of the exams; and b) comparing student's initial scores on their quizzes with their performance on the exams, as well as comparing student's self-diagnosis scores with their performance on the exams.  We discuss some hypotheses about the students' ability to self-diagnose with different levels of scaffolding support and emphasize the importance of teaching students how to diagnosis their own mistakes.  Our findings suggest that struggling with minimal support during in-class self-diagnosis can trigger out-of-class self-diagnosis.  Students therefore may be motivated to make sense of the problem they may have not been able to self diagnose, whether independently or in a collaborative effort.