Shapiro and Niederhauser (2004) explain that learning from hypertext is more complicated than learning from traditional text. While many of the elements are the same--character decoding, word recognition, sentence comprehension, etc.--research of hypertext in education focuses on the unique features that add complexity. First, hypertext is non-linear. For this reason, information within a hypertext may be consumed by each user in a unique sequence. This places greater metacognitive demands on the reader because he or she must monitor comprehension, determine what information will need to be found to close information gaps, and make decisions about where best to look or that information in the text. Second, user traits (goals, motivation, and prior knowledge) are also factors that interact with the characteristics of hypertext and influence learning outcomes. Shapiro and Niederhauser "identify the variables that affect HAL most strongly and the mechanisms through which this occurs" (pg. 605). Shapiro and Niederhauser claim that the two theories that have had the greatest impact on research and our understanding of the process are the construction-integration model (CIM; Kintsch, 1998) and cognitive flexibility theory (CFT; Spiro, Coulson, Feltovitch, & Anderson, 1988; Spiro, Feltovitch, Jacobson, & Coulson, 1992).
The CIM of text processing (Kintsch, 1988) is a three-stage process of text comprehension:
According to Shapiro and Niederhauser, the third stage of the process--creation of the situation model--is significant to our understanding of learning from hypertext. It is also noted that hypertext promotes active learning because the learner must choose which links to click on to interact with the learning. Construction Integration model has become many hypertext researchers' standard framework for understanding hypertext-assisted learning. Specifically, user behaviors such as link choice, navigation patterns, and metacognitive practices.
Cognitive Flexibility Theory "is based on the supposition that real-world cases are each unique and multifaceted, thus requiring the learner to consider a variety of dimensions at once" (pg. 606). In other words, the prior knowledge necessary to understand new knowledge is derived from aspects of a variety of combined prior experiences and applied to the new situation. The implication of this model then is that advanced learning takes place as a consequence active learning, the use of prior knowledge, and as a consequence of constructing new knowledge for each new problem. CFT is relevant to hypertext because a learner can access a single document from multiple other sites. In doing so, the he or she will come to that document with multiple perspectives. In turn the mental representations resulting from repeated exposure to ill-structured hypertext will be multifaceted and therefore one's ability to use that knowledge should be more flexible. Shapiro and Niederhauser claim that CFT offers an explanation of meaningful learning on the part of advanced learners.
Numerous cognitive factors associated with reading and learning from hypertext show that there are distinct differences between reading traditional text and reading hypertext. Factors include--
Shapiro and Niederhauser summarize that the "nature of hypertext renders HAL a more cognitively demanding mode of learning" (pg. 608). For this reason, the use of metacognitive strategies is all the more important in this context. However, several studies have shown that minimal training and/or automated prompts may be used to promote metacognitive strategies and influence learning outcomes with some degree of success.
Interest around learning with hypertext stems from the belief that notion hypertext information structures may mirror the semantic structures of human memory. There is little evidence, though, that the simple act of working with a hypertext designed to mirror an expert's conceptual understanding of a topic can lead to a direct transfer of expert like mental representations to the reader. Research shows conflicting results about the effect of system structure (e.g. organization of links on pages, maps, overviews, and indexes) on learning. While some studies have shown advantages to using a highly organized system structure such as a hierarchy, others have actually found advantages of working with ill-structured hypertexts. Yet, other existing studies show the pitfalls of an ill-structured system design. Despite these contradictions, two general conclusions have been drawn from the literature and are said to explain the ways in which these variables interact to impact learning. First, "well-structured hypertexts may offer low-knowledge learners an introduction to the ways in which topics relate to one another and an easy-to-follow introduction to a domain" (pg. 611). And second, ill-structured hypertexts are beneficial to advanced learning for active, engaged learners. To conclude, research on organizing tools and system structure has shown that well-defined structures like hierarchies are helpful if the goal is to achieve basic, factual knowledge. Ill-structured systems are often beneficial for deep learning. This is especially true for advanced learners.
Researchers have also attempted to identify learning variables like individual knowledge and engagement, reading patterns, and learning goals. In regards to individual knowledge and engagement, those with limited prior knowledge are unable to establish information needs in advance. Shapiro and Niederhauser explain that individual differences in learning style are often important to the learning outcomes because they interact with other factors such as system structure. As for reading patterns, researchers have sought to identify patterns of reader navigation as they read hypertext. They found that learner interest and domain knowledge had a notable impact on readers' navigational strategies. They also noted that knowledge seekers tend to learn more from the text than feature explorers.
Hypertext navigation is not always systematic and purposeful though. A great deal of research has attempted to address what is called the keyhole phenomenon. In short, the keyhole phenomenon examines the effect of different types o user interfaces on user disorientation. Shapiro and Niederhauser summarize that the need to navigate through a hypertext is a defining characteristic that differentiates reading and learning with traditional text. Therefore, navigation strategies may influence what the reader learns from the text. This in turn may be influenced by the conceptual difficulty associated with the content and the learning task (pg. 614). The literature also shows that with consistency, learning with hypertext is greatly enhanced when the learning goal is specific.
According to Shapiro and Niederhauser, the bulk of related literature surrounds techniques in user modeling. User modeling refers to any methods used to gather information about users' knowledge, skills, motivation, or background. Characteristics o users are used to alter system features like links and document content. These studies suggest a need for further investigation into the educational effectiveness of adaptive systems to determine what characteristics are most effectively used in user modeling, and what system characteristics are most important to adapt.
Several theoretical issues surround HAL research as well as methodological issues. Methodological issue stem from a difficulty in comparing and reviewing hypertext research because of the absence of a unified, coherent framework for studying hypertext. Shapiro and Niederhauser argue that this creates two problems when trying to understand the hypertext literature. First, the text based reading research foundation is compromised when extensive graphics and audio and video components are included in the hypertext. Second, issues comparing research studies emerge when our language about the field is lacking in precision. It should also be noted that methodological flaws have been widely reported in the literature. While a great deal of excitement surrounds hypertext as an educational tool, Shapiro and Niederhauser conclude with a reminder that there is very little research published on the technology that is related to education and learning. They call for future research in this area to generate a well-grounded understanding of the processes underlying HAL and a standardization of terminology and methodology to be developed.
Shapiro, A., & Niederhauser, D. (2004). Learning from hypertext: Research issues and findings. In D. H. Jonassen (Ed), Handbook of Research for Educational Communications and Technology (pp. 605-620). New York: Macmillan.