Methods and approaches used at the U.S. Army Research Institute to investigate intelligent computer-assisted foreign language instruction are discussed. The research described explores the use of hypertext and natural language processing for improving language training by articulating student knowledge structures and for providing, incidentally, a new basis for aptitude testing. The techniques being implemented model the cognitive skills underlying foreign language competence by using computational linguistic models and semantic networks built into hypertext systems. It is proposed that these techniques could be useful for assessing basic language competence. Aspects of the research and related technology are examined: intelligent tutoring systems, the expert model and second language learning, student (trainee) models and language aptitude testing, the pedagogical model, the knowledge base, the interface, hypertext systems and command menus, smart computer-assisted language learning environments, grammars and grammar representations, parsing strategies, using grammars for computer-assisted language learning, use and adaptation of specific software packages, immersion techniques, and semantic mapping. (MSE)