Prompting students to self-explain during problem solving has proven to be an effective instructional strategy across many domains. However, despite being called “domain general”, very little work has been done in areas outside of math and science. In this dissertation, I investigate whether the self-explanation effect holds when applied in an inherently different type of domain, second language grammar learning. Through a series of in vivo experiments, I tested the effects of using prompted self-explanation to help adult English language learners acquire the English article system (e.g., teaching students the difference between “I saw a dog” versus “I was the dog”). In the pilot study, I explored different modalities of self-explanation (free-form versus menu-based), and in Study 1, I looked at transfer effects between practice and self-explanation. In the studies that followed, I added an additional deep processing manipulation (Study 2: analogical comparisons) and a strategy designed to increase the rate of practice and information processing (Study 3: worked example study). Finally, in Study 4, I built and evaluated an adaptive self-explanation tutor that prompted students to self-explain only when estimates of prior knowledge were low. Across all studies, results show that self-explanation is an effective instructional strategy in that it leads to significant pre- to post-test learning gains, but it is inefficient compared to tutored practice. In addition to learning gains, I compared learning process data and found that both self-explanation and practice lead to similar patterns of learning and there was no evidence in support of individual differences. This work makes contributions to learning sciences, second language acquisition (SLA), and tutoring system communities. It contributes to learning sciences by demonstrating boundary conditions of the self-explanation effect and cautioning against broad generalizations for instructional strategies, suggesting instead that strategies should be aligned to target knowledge. This work contributes to second language acquisition theory by demonstrating the effectiveness of computer-based tutoring systems for second language grammar learning and providing data that supports the benefits of explicit instruction. Furthermore, this work demonstrates the relative effectiveness of a broad spectrum of explicit learning conditions. Finally, this work makes contributions to tutoring systems research by demonstrating a process for data-driven and experiment-driven tutor design that has lead to significant learning gains and consistent adoption in real classrooms.