"Self-diagnosis tasks" aim at fostering students' learning in an examination context by requiring students to present diagnoses of their solutions to quiz problems. We examined the relationship between students' learning from self-diagnosis and the typicality of the problem situation. Four recitation groups in an introductory physics class ([approximately equal to] 200 students) were divided into a control group and three intervention groups in which different levels of guidance were provided to aid students in their performance of self-diagnosis activities. The self-diagnosis task was administered twice, first in an atypical problem situation and then in a typical one. In a companion paper we reported our findings in the context of an atypical problem situation. Here we report our findings in the context of a typical problem situation and discuss the effect of problem typicality on students' self-diagnosis performance and subsequent success in solving transfer problems. We show that the self-diagnosis score was correlated with subsequent problem-solving performance only in the context of a typical problem situation, and only when textbooks and notebooks were the sole means of guidance available to the students for assisting them with diagnosis. (Contains 10 tables and 6 figures.)