Learning from hypertext depends on metacognition

Hypertext is a user controlled medium in which the user interacts with the text by navigating through a collection of text nodes using links. The present study attempted to form and test a framework of how metacognition interacts with hypertext to affect learning. It is theorized that learning from hypertext particularly depends (more so than linear-text) on the use of metacognitive skills, specifically monitoring and control. Two experiments were conducted, each using a different approach to have some converging evidence on the relative merits of the theoretical conceptions made here. Experiment 1 used an approach based on individual differences in metacognitive ability, and Experiment 2 used an experimental manipulation of metacognitive monitoring. In the first experiment, measures of individual differences in metacognitive abilities which included two objective measures (judgment of learning of paired-associate words, and retrospective confidence judgments of a general knowledge test) and two subjective measures (the Metacognitive Awareness Inventory or MAI, and the Inaccurate/Inactive Metacognition Inventory or IMI) were collected from 116 participants who either read hypertexts or linear-texts and were subsequently tested on their learning. Findings revealed that metacognitive ability measures predicted learning from hypertext but not from linear-text. The second experiment included the manipulation of one group of participants' process of comprehension monitoring by embedding adjunct questions within a passage, which has been shown to increase the accuracy of metacognitive monitoring. The purpose of the embedded-questions was to cause the participants' to self-test themselves or monitor their comprehension by actively retrieving information from memory about the passage while reading. A control group was not given embedded-questions. Half of each group received a hypertext version and the other half received a linear-text version, and were subsequently tested on their learning. The results showed an interaction of metacognitive activation by text medium on learning in that learning of the passage was best for the group that had their metacognition activated and received the hypertext version. The results of both experiments provided converging evidence that learning from hypertext depends on metacognition. Theoretical implications, opportunities for future research, and implications for system design are discussed throughout the paper.