We study the effectiveness of animated cartoons in an Intelligent Tutoring System (ITS) that teaches Fractions for 4th grade. By examining viewing patterns, and the relation between viewing and performance on related tasks, we seek to better understand how young students learn mathematical concepts with Cartoons. The analysis serves two purposes. One is optimizing the pedagogy of the specific ITS. Second is deriving general pedagogic principles that will guide the design of future ITSs. This study is part of a larger effort aimed at improving our understanding of how to make better digital learning environments using educational data mining. The study uses data from 650 4th grade students, who used the ITS for 4-8 weeks. The analysis is based on combining high-resolution clickstream data with meta-data that describes the content and its structure. To measure the relation between viewing and performance, we used a between-group quasi-experiment design and regression analysis.