Peer Instruction, a pedagogy utilizing handheld classroom response technology to promote student discussion, is one of the most popular research-based instructional practices in STEM education. Yet, few studies have shed theoretical light on how and why Peer Instruction is effective. In this article, J. Bryan Henderson explores the Peer Instruction technique through a controlled methodology where theory--in this case the Interactive-Constructive-Active-Passive (ICAP) framework for differentiating various modes of cognitive engagement--drives pedagogical adaptations that serve as the differing experimental conditions. He finds that among the four high school physics classes he studied which employed Peer Instruction, the students achieved learning gains that, when normalizing for pretest performance, on average were more than 10 percent greater than those of college students not exposed to the ICAP-driven methodology when learning introductory physics. This article serves as an example to the educational research community of how the ICAP framework can help illuminate theoretical mechanisms behind instructional techniques in ways the more general use of the term active learning cannot.