This thesis investigated how worked examples could be used to fade assistance in the domain of algebra. The method of fading assistance was novel. It used similarity as the mode of assistance, with similar problem-example pairs providing high assistance and dissimilar pairs providing low assistance. Learning, performance, and gaze behaviours captured by an eye tracker were analyzed across three conditions – a faded assistance condition, a constant assistance condition, and a reverse faded assistance condition. Participant's personality traits and attitudes and behaviours regarding math were also collected and correlated with eye gaze sequences. We found that, contrary to our hypothesis, the reverse faded assistance condition resulted in the greatest learning gains. We analyzed the gaze behaviours to shed light on this finding and found that participants in this condition focused significantly more on the problem solution, suggesting more cognitive processing during problem solving than in the other conditions.