The purpose of this study was to investigate the effects of both metacognition and hypertext structure on performance in a hypertext learning environment. Basic questions of this study were : Does learners' metacognition have significant effects on learner's performance in a hypertext-mediated instruction? Which of two types of hypertext structure, hierarchical and associative structure, is more effective for teaching novices an authoring language system? Are there any effects of interaction between metacognition and hypertext structure on performance? Subjects were 45 college students in Korea. They had no experience in HyperCard and Macintosh computers. The hypertext program used in this study was designed to teach both basic components and skills of HyperCard. This study defined metacognition as the combination of metacognitive knowledge and metacognitive regulation. Metacognitive knowledge was assessed with Metamemory in Adulthood(MIA). All experimental sessions were conducted on individual basis. Each subject was asked to think aloud while each hypertext instruction was implemented Videotaped trunk-aloud protocols were coded by two coders and numbers of verbalization about metacognitive regulation were counted. There were no significant effects of metacognitive knowledge on the recall test scores as well as the skill test scores, because some sub-components of metacognitive knowledge had negative effects on performance while the others had positive effects. In contrast, metacognitive regulation had significant relationships with two types of performance in a hypertext. The high metacognitive regulation group achieved the higher scores on the two types of performance tests regardless of the hypertext structure. The low metacognitive regulation group achieved the lower scores regardless of the hypertext structure. However, the low metacognitive regulation group attained higher scores on the skill test under the hierarchical hypertext than under the associative hypertext. The results of the regression analysis of the recall test scores indicated that about 40.6% of the variance was explained significantly by the set of sub-components of metacognitive knowledge, sub-components of metacognitive regulation, and hypertext structure. The effective predictors of the recall test scores were two sub-components of metacognitive knowledge (use of memory strategies and knowledge of memory tasks and processes) and two sub-components of metacognitive regulation (evaluation of cognition and remediation of cognition). The component of use of memory strategies, however, had negative effects on the recall test scores. The results of analysis showed that about 47.6% of the variance of the skill test scores were explained by the set of sub-components of metacognitive knowledge, sub-components of metacognitive regulation, and hypertext. The best predictors of the variance were perception of own memory capabilities, knowledge of memory tasks and processes, and monitoring of cognition. However, the sub-component of perception of memory capabilities had negative effects on the skill test scores.