Cognitive load theory has been designed to provide guidelines intended to assist in the presentation of information in a manner that encourages learner activities that optimize intellectual performance. The theory assumes a limited capacity working memory that includes partially independent subcomponents to deal with auditory/verbal material and visual/2- or 3-dimensional information as well as an effectively unlimited long-term memory, holding schemas that vary in their degree of automation. These structures and functions of human cognitive architecture have been used to design a variety of novel instructional procedures based on the assumption that working memory load should be reduced and schema construction encouraged. This paper reviews the theory and the instructional designs generated by it. KEY WORDS: cognition; instructional design; learning; problem solving.

Worked examples are instructional devices that provide an expert's problem solution for a learner to study. Worked-examples research is a cognitive-experimental program that has relevance to classroom in-struction and the broader educational research community. A frame-work for organizing the findings of this research is proposed, leading to instructional design principles. For instance, one instructional de-sign principle suggests that effective examples have highly integrated components. They employ multiple modalities in presentation and em-phasize conceptual structure by labeling or segmenting. At the lesson level, effective instruction employs multiple examples for each concep-tual problem type, varies example formats within problem type, and employs surface features to signal deep structure. Also, examples should be presented in close proximity to matched practice problems. More-over, learners can be encouraged through direct training or by the structure of the worked example to actively self:explain examples. Worked examples are associated with early stages of skill develop-ment, but the design principles are relevant to constructivist research and teaching. The Historical Context In recent years, learning from "worked examples " has received a consider-able amount of attention from researchers (e.g., Chi, Bassok, Lewis, Reimann, & Glaser, 1989; Ward & Sweller, 1990), particularly in such fields as mathematics, physics, and computer programming. Although there is no precise definition, worked examples share certain family resemblance (Wittgenstein, 1953). As instructional devices, they typically include a problem statement and a proce-dure for solving the problem; together, these are meant to show how other similar problems might be solved. In a sense, they provide an expert's problem-

Previous studies have demonstrated the learning benefit of personalized language and worked examples. However, previous investigators have primarily been interested in how these interventions support students as they problem solve with no other cognitive support. We hypothesized that personalized language added to a web-based intelligent tutor and worked examples provided as complements to the tutor would improve student (e- )learning. However, in a 2 x 2 factorial study, we found that personalization and worked examples had no significant effects on learning. On the other hand, there was a significant difference between the pretest and posttest across all conditions, suggesting that the online intelligent tutor present in all conditions did make a difference in learning. We conjecture why personalization and, especially, the worked examples did not have the hypothesized effect in this preliminary experiment, and discuss a new study we have begun to further investigate these effects.

This article provides an overview description of the four-component instructional design system (4C/ID-model) developed originally by van Merriënboer and others in the early 1990s (van Merriënboer, Jelsma, & Paas, 1992) for the design of training programs for complex skills. It discusses the structure of training blueprints for complex learning and associated instructional methods. The basic claim is that four interrelated components are essential in blueprints for complex learning: (a) learning tasks, (b) supportive information, (c) just-in-time (JIT) information, and (d) part-task practice. Instructional methods for each component are coupled to the basic learning processes involved in complex learning and a fully worked-out example of a training blueprint for “searching for literature ” is provided. Readers who benefit from a structured advance organizer should consider reading the appendix at the end of this article before reading the entire article. The instructional design enterprise is a bit like an ocean liner—huge, slow, ponderous, and requiring large amounts of energy and a great deal of time to move it even one degree off its current path. Recent discussions and developments in the field concern rapid technological and societal changes and the resulting need for very complex knowledge at work (Berryman, 1993; Cascio, 1995); new constructivist design theories for problem solving (Jonassen, 1994;

Research has shown that it is effective to combine example study and problem solving in the initial acquisition of cognitive skills. Present methods for combining these learning modes are, however, static and do not support a transition from example study in early stages of skill acquisition to later problem solving. Against this background, we propose a successive integration of problemsolving elements into example study until the learners solve problems on their own (i.e., complete example increasingly more incomplete examples problem to-besolved) . We tested the effectiveness of such a fading procedure against the traditional method of employing exampleproblem pairs. In a field experiment and in a more controlled lab experiment, we found that the fading procedure fosters learning, at least when near transfer performance is considered. Moreover, this effect is mediated by a lower number of errors under the fading condition as compared to the example-problem condition.

Despite the huge production of all sorts of multimedia instructions, educational research has yielded surprisingly few general design principles for instructions in which verbal and visual information are combined. Instructional designers seem to base their design choices more on intuitive ideas than on sound research results. There are however some interesting theories that give guidelines for instructional designers. Mayer (1997) describes in his Generative Theory of Multimedia Learning how the learner builds mental representations of multimedia material. One important step in this process is the integration of both visual and verbal information. Multimedia instruction in its most elementary form consists of a picture with an explanatory text. Because picture and text cannot be perceived simultaneously, the learner is forced to switch back and forth between the two and integrate them mentally. According to Cognitive Load Theory (Sweller, 1988; Sweller, van Merriënboer & Paas, 1998) this integration process is cognitively demanding and at the expense of mental resources that could otherwise be allocated to the learning process. Sweller et al. call the

Our work explores the assistance dilemma: when should instruction provide or withhold assistance? In three separate but very similar studies, we have investigated whether worked examples, a high-assistance approach, studied in conjunction with tutored problems to be solved, a mid-level assistance approach, can lead to better learning. Contrary to prior results with untutored problem solving, a low-assistance approach, we found that worked examples alternating with isomorphic tutored problems did not produce more learning gains than tutored problems alone. On the other hand, the examples group across the three studies learned more efficiently than the tutored-alone group; the students spent 21 % less time learning the same amount of material. Practically, if these results were to scale across a 20-week course, students could save 4 weeks of time – yet learn just as much. Scientifically, we provide an analysis of a key dimension of assistance: when and how often should problem solutions be given to students versus elicited from them? Our studies, in conjunction with past studies, suggest that on this exampleproblem dimension mid-level assistance may lead to better learning than either lower or higher level assistance. While representing a step toward resolving the assistance dilemma for this dimension, more studies are required to confirm that mid-level assistance is best and further analysis is needed to develop predictive theory for what combinations of assistance yield the most effective and efficient learning.

To foster more robust student learning, when should instruction provide information and assistance to students and when should it request students to generate information, ideas, and solutions? In different forms, this dilemma for instructors has been a part of debates on education since Plato. However, it is fair to say that we remain far from a precise and sound scientific response. We believe this “Assistance Dilemma ” is one of the fundamental unsolved problems in the cognitive and learning sciences. To address this dilemma, we suggest a four step strategy for more clearly articulating the problem and tackling it with computational models that can be used to make precise, replicable, and testable predictions about when instructional assistance should be given vs. withheld. We illustrate these steps on two different dimensions of instructional assistance. On the “problem spacing ” dimension, we present a computational model that generates precise predictions of the kind we call for. On the more complex “example-problem ” dimension, we illustrate how the field is at a point where such a precise computational model may be possible.

recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. Keywords: computational modeling, cognitive modeling, instructional theory, machine learning, learning science, second language learning, mathematics learning, science learning, robust learning, learning theory, knowledge componentsExecutive Summary The volume of research on learning and instruction is enormous. Yet progress in improving educational outcomes has been slow at best. Many learning science results have not been translated into general practice and it appears that most that have been fielded have not yielded significant results in randomized control trials. Addressing the chasm between learning science and educational practice will require massive efforts from many constituencies, but one of these efforts is to develop a theoretical framework that permits a more systematic accumulation of the relevant research base. A key piece in such a theoretical framework is the development of levels of analyses that are fine enough to be supported by cognitive science and cognitive neuroscience, but also at levels appropriate to guide the design of effective educational practices. An ideal scientific solution would be a small set of universal instructional principles that can be applied to produce efficient

Performance and preference measures are commonly used in the assessment of visualization techniques. This is important and useful in understanding differences in effectiveness between different treatments. However, these measures do not answer how and why the differences are caused. And sometimes, performance measures alone may not be sensitive enough to detect differences. In this paper, we introduce a cognitive approach for visualization effectiveness and efficiency assessment. A model of user performance, mental effort and cognitive load (memory demand) is proposed and further mental effort and visualization efficiency measures are incorporated into our analysis. It is argued that 1) combining cognitive measures with traditional methods provides us new insights and practical guidance in visualization assessment. 2) analyzing human cognitive process not only helps to understand how viewers interact with visualizations, but also helps to predict user performance in initial stage. 3) keeping cognitive load induced by a visualization low allows more memory resources to be available for high level complex cognitive activities. A case study conducted supports our arguments.