Orthographic learning is the process that supports children in becoming skilled word readers. How orthographic learning occurs has been one of the central questions in the scientific studies of reading. The present systematic review focuses on experimental studies of orthographic learning via self-teaching. It explains and discusses the universality and specificity of the Self-Teaching Hypothesis concerning written word learning among young children. To advance the Self-Teaching Hypothesis with existing empirical evidence, methodologies and critical findings on the roles of phonological recoding, context, and word property are reviewed and analyzed systematically across target languages and types of learners. Sixty-two experimental studies from 45 articles (1995-2022) were included in the current analysis. Ample discrepancies in methodologies exist across studies. Across writing systems, language-specific word properties affect children's self-teaching outcomes, yet evidence consistently suggests that the presence of phonological recoding supports self-teaching. Enhanced phonological recoding improves orthographic learning, whereas reduced phonological recoding hinders orthographic learning. In contrast, findings on the effect of meaningful context are mixed. Our findings advance the original Self-Teaching Hypothesis, reveal the gaps in the self-teaching research, and point out new directions for future work. Our findings also inform educational practices for enhancing written word learning.