The implementation of the program used in this study, TPIXIE, was predicated on the assumption that acquiring the skill to diagnose student errors was important for teacher trainees. The TPIXIE system (part of the PIXIE family of intelligent tutoring systems) was designed to present users with a series of task-student-answer pairs (where the student's answer was incorrect). The system presents several sets of tasks together with a student's responses, and then presents further tasks and asks the user to predict the responses. If the user is unable to do this, the cycle is repeated. The experiment reported compared the performance of two groups of teacher trainees, one using TPIXIE and the other a variant of PIXIE which merely asked the user to work algebraic tasks. The experiment suggested that the users benefited from interacting with TPIXIE, but transfer of skill was not demonstrated to analogous tasks. The paper concludes with suggestions for a further experiment. (TW)