The focus of this dissertation is a web-based, second language (L2) instructional resource called VVT (Virtual Vocabulary Trainer) designed to teach integrated vocabulary depth of knowledge and dictionary referencing skills to tertiary-level learners of English as a Second Language (ESL). In addition to evaluating the potential of online resources to address long-standing challenges in the field of L2 strategy instruction, the project also targeted research objectives in other related areas. The work is described in four separate articles aimed at peer-reviewed journals in the fields of computer-assisted language learning (CALL), applied linguistics, and teaching English as a Second Language. The first paper discusses the theoretical underpinnings of the VVT course, the procedures followed in developing it, and the materials themselves, while also suggesting general design principles for L2 strategy instruction based on frameworks derived from cognitive theory. The second paper represents the evaluation component of the project and reports how the feasibility of automated, online strategy instruction was investigated by studying the resource's effectiveness, both actual and perceived, in an experiment involving 64 ESL composition students. In the third paper, the VVT course serves as the platform for a study critiquing a recently proposed, structural model of L2 vocabulary learning, while at the same time adding to the literature on the acquisition of L2 vocabulary depth-of-knowledge features. The final paper uses discrepant cases from the evaluation study, i.e. individual participants whose performances diverged significantly from group norms, to investigate the metacognitive strategy of task definition, the stage of learning in which students develop internal, and potentially idiosyncratic, representations of instructional tasks. The findings provide evidence of the feasibility of automated, online strategy instruction for complementing teacher-led forms, while also shedding light on the challenges many L2 learners face in self-directed learning of vocabulary depth of knowledge. Finally, it demonstrates the potential of an integrative, multicomponential model of self-regulation for researching and theorizing about L2 learning.