This study aimed to investigate the influence of science textbook graphical design based on principles of cognitive process on graphical-textual learning performance of different academic achievement fifth graders students. A quasi-experimental design mainly and auxiliary to half-structure questionnaire and interview, fifty-seven elementary students participated in this research. The experiment treatment was students in different classes used different graphical science textbook (one was based on cognitive principles, and another was published by civil). The influences of learning performance compared to published science textbook were assessed using self-developed tests and the qualitative data were analyzed using independent sample t-test. Administration of a student graph perceptive questionnaire and then nine students from high and low achievement groups in each class were interviewed; qualitative data analyzed to infer the mental process. The results showed:(1) High academic achievement students were significantly enhanced leaning performances on knowledge dimension and cognitive process dimension and had strong effect size, the experimental group was better than the contol group. As to low academic achievement students, there was no significant different between these two groups. (2) Schema organizational graph enhance high academic achievement students on conceptual knowledge and assist low academic achievement students recognize material outline; experimental transformational graph enhance high academic achievement students on procedural knowledge and assist low academic achievement students participate in the experiment.