This study investigates how the illustrations in a science textbook, with their design modified according to cognitive process principles, affected students' learning performance. The quasi-experimental design recruited two Grade 5 groups (N?=?58) as the research participants. The treatment group (n?=?30) used the modified version of the textbook, and the comparison group (n?=?28) used the standard textbook published in Taiwan. Scores from a researcher-developed performance assessment for the two groups were examined with a one-way ANCOVA, using the first unit examination as the covariate. Results showed that the modified textbook with a cognitive principles-driven design enhanced the participants' learning performance in terms of conceptual knowledge, as well as transfer and retention, compared with that of the group using the standard textbook. However, such improvement did not occur in procedural knowledge. Thus, the learning benefits of cognitive principles-driven illustrations warrant further investigation.