Self-diagnosis activities require students to self-diagnose their solutions to problems they solved on their own by detecting and explaining their errors. Worked examples, a step-by-step demonstration of how to solve a problem, are often used to support students in self-diagnosis activities. However, studies indicate that students often fail to exploit worked examples in traditional self-diagnosis activities when simply required to self-diagnose their solutions. This study analyzes a new self-diagnosis activity developed by the first author to prompt students to effectively use worked examples when self-diagnosing: the written worked examples only constitute one part of a scoring rubric and the students are required to both self-diagnose their solutions and then self-score them. This activity was hypothesized to encourage students to exploit the worked examples more thoroughly, and by extension detect and learn from their errors to a greater extent than students administered the traditional self-diagnosis activities. Six 10th grade advanced physics classes completed a pre-test/intervention/post-test after finishing a unit in geometric optics. Students in each class were randomly assigned to the new self-diagnosis activity (83 students) or the traditional self-diagnosing activity (79 students). Students assigned the new activity detected and learned more from their errors than students administered the traditional activity. It is argued that more in-depth error detection contributed overall to students' learning by triggering a series of implicit steps that prompted them to self-regulate their cognitions in a way that provided opportunities to self-repair their naïve concepts.