In this symposium, findings of the Dutch/German NSF research programme LEMMA (Learning Environments, MultiMedia and Affordances) are presented. The main goal of this research programme is to gain insight on how representational codes influence learning processes and learning results. The ultimate objective of the programme is to produce a coherent framework of design rules and guidelines for the use of (multiple) representations in multimedia learning environments. From a theoretical perspective, the programme combines theories of ‘computational effectiveness’ (Larkin & Simon, 1987; Stenning & Oberlander, 1995), ‘dual coding’ (Paivio, 1990), ‘cognitive load’ (Sweller, 1999), and ‘multimedia design’ (Ainsworth, 1999).

In this symposium, a set of experimental studies on the effectiveness of different types of representations (pictorial, arithmetical, and textual, and their combinations) on learning processes and learning outcomes is presented. Four instructional approaches will be addressed: inquiry learning, observational learning, hypermedia learning, and example based learning. All approaches have used the same domain of probability theory, the same set of performance tests, and the same cognitive load measures for reasons of comparison. This comparison across instructional approaches will be presented in the final presentation.

 

Ainsworth, S.E., (1999) A functional taxonomy of multiple representations. Computers and Education, 33(2/3), 131-152.

Larkin, J.H., & Simon, H.A. (1987). Why a diagram is (sometimes) worth ten thousand words. Cognitive Science, 11, 65-99.

Paivio, A. (1990). Mental representations: A dual coding approach. New York: Oxford University Press.

Stenning, K., & Oberlander, J. (1995). A cognitive theory of graphical and linguistic reasoning: Logic and implementation. Cognitive Science, 19, 97-140

Sweller, J. (1999). Instructional design in technical areas. Camberwell: ACER Press.

 

Successful learning cannot be attained by processing subject matter only superficially. It requires a deeper level of processing, whereby the learner integrates new information into new or existing cognitive structures (Mayer, 2002, 2004; Novak, 2002). This is what Ausubel (1968) calls meaningful learning, as opposed to rote learning whereby the learner ends up with fragmented and unconnected pieces of knowledge. Mayer argues that meaningful learning may not simply be the result of behavioral activity per se. He suggests that only specific cognitive activities (e.g. selecting, organizing, and integrating knowledge) may promote meaningful learning (Mayer, 2002, 2004).

 

In some domains it is notoriously difficult for learners to gain conceptual understanding, for example in science and, in particular, in mathematics. The problem in mathematics may largely be explained by the formal, abstract way in which the subject matter is represented. This makes it hard for learners to relate the subject matter to every day life experiences and to existing cognitive structures. Many assignments have been designed to foster meaningful learning. All these assignments have in common that they require learners to externalize and express their knowledge, for example in the form of knowledge or concept maps, summaries, self-explanations, questioning, explanatory answers, drawings, notes, models, propositions, and learner-designed assignments. Learner-generated externalization of knowledge is expected to facilitate the construction of mathematical knowledge (Gelman & Greeno, 1989; Heritage & Niemi, 2006; Niemi, 1996). Because of the difficulties learners have with mathematical representations it needs to be established how learners understand and use representations, for example in expressing their domain knowledge.

 

Tarr (1997; cited in Tarr & Lannin, 2005) found that learners, in expressing probabilities, are likely to disregard standard mathematical notations, and use alternative, self-invented forms, ranging from textual statements to conventional numerical representations, or combinations of these forms. Although not entirely correct in a strictly formal sense, these self-invented forms were apparently more meaningful to the learners.

Following instruction, learners tended to use more conventional ways of representing the probability of events (i.e. using ratios or odds, or formal numerical probabilities), but others still used self-invented representations. We hypothesize that learners invent representations in order to be able to relate and integrate the newly acquired knowledge into existing cognitive structures.

 

The current study applies to probability and aims at addressing the following questions: (1) does the externalization of knowledge by the learners facilitate the construction of mathematical knowledge and (2) does the format in which the learners express their knowledge affect the quality of their expressed knowledge and learning outcomes? In order to find the answers to these questions, a pre-test-post-test design has been applied in which four conditions have been compared to each other: one control condition and three experimental conditions. Learners in the experimental conditions were asked to create a domain representation (i.e. express their domain knowledge) in a graphical, mathematical, or textual format. Learners in all conditions worked with simulation-based inquiry learning environments. All these environments were identical except for one feature: the so-called whiteboard, which learners could use to express their knowledge. In one experimental condition this whiteboard was pre-structured graphically, allowing learners to construct a concept map of the domain, in the second experimental condition the whiteboard was pre-structured mathematically, only allowing the input of variable names, numerical data, and mathematical operators, and in the third experimental condition the whiteboard resembled simple word processing software, allowing textual and numerical input.

 

It was found that the expression of domain knowledge by learners is related to enhanced levels of situational knowledge and overall knowledge. This difference could not be attributed to time-on-task. Second, the representational format in which learners expressed their knowledge (graphical, mathematical, or textual) does not influence either the quality of the domain representation or the learning outcomes. Third, the format does affect the perceived affordances of the representational format (graphical, mathematical, or textual). In case of a graphical or textual structure, about half of the learners create a domain representation. In case of a mathematically structured whiteboard, less than 20% of the learners create a domain representation.

 

 

Ausubel, D. P. (1968). Educational psychology: A cognitive view. New York: Holt, Rinehart, & Winston.

Gelman, R., & Greeno, J. G. (1989). On the nature of competence: Principles for understanding in a domain. In L. B. Resnick (Ed.), Knowing, learning, and instruction: Essays in honour of robert glaser (pp. 125-186). Hillsdale NJ: Lawrence Erlbaum Associates, Inc.

Heritage, M., & Niemi, D. (2006). Toward a framework for using student mathematical representations as formative assessments. Educational Assessment, 11, 265-282.

Mayer, R. (2002). Rote versus meaningful learning. Theory into Practice, 41, 226-232.

Mayer, R. (2004). Should there be a three-strikes rule against pure discovery learning? The case for guided methods of instruction. American Psychologist, 59, 14-19.

Niemi, D. (1996). Assessing conceptual understanding in mathematics: Representations, problem solutions, justifications, and explanations. Journal of Educational Research, 89, 351-363.

Novak, J. D. (2002). Meaningful learning: The essential factor for conceptual change in limited or inappropriate propositional hierarchies leading to empowerment of learners. Science Education, 86, 548-571.

Tarr, J. E. (1997). Using middle school students' thinking in conditional probability and independence to inform instruction. Doctoral dissertation. Illinois State University.

Tarr, J. E., & Lannin, J. K. (2005). How can teachers build notions of conditional probability and independence? In G. A. Jones (Ed.), Exploring probability in school: Challenges for teaching and learning (pp. 215-238). New York, NY: Springer.

In three studies we investigated how learning from animated expert models in the domain of probability calculation can be optimized. In animated expert models an expert solves a problem and explains how and why this is done in this particular way. Cognitive load theory contends that these models should be designed in such a way that learners prevent ineffective cognitive load and engage in relevant learning activities that impose effective cognitive load. Study 1 investigated guidelines to reduce ineffective cognitive load and revealed that learner-paced spoken continuous animated models led to better near transfer performance than spoken segmented animated models, whereas with written explanatory text learner-paced continuous animated models led to better far transfer performance than segmented animated models. Study 2 and 3 investigated guidelines that have learners engage in relevant learning activities. In Study 2 we argued that reflection prompts would stimulate learners to be cognitively active and integrate new information with their prior knowledge. The results showed that reflection prompts were effective with written explanatory texts, but not with spoken explanatory texts. Study 3 investigated whether alternating between observing models and practicing would ameliorate learning. We argued that observing and practicing yield different ways of processing information and that alternating between these two instructional techniques would yield enriched schemas. Data of this study will be available in February 2007. In the presentation the theoretical (e.g., what does this mean for the modality effect) and practical implications of these studies will be discussed.

In animated expert models an expert solves a problem and explains how and why this is done in this particular way. According to cognitive load theory, these models need to be designed, in such a way that the ineffective cognitive load is minimized and the effective cognitive load is maximized. We present three related studies in the domain of probability calculation in which guidelines were investigated to accomplish this. In all studies learners from secondary school first engaged in a prior knowledge test. Subsequently, in the experimental treatment the learners were confronted with eight problems. The number of eight problems was based on an analysis of the domain of probability calculation in which two characteristics are important: The order of drawing (not important vs. important) and replacement of drawing (without replacement vs. with replacement). This results in four different problem categories. In each problem category two problems were presented. The first problem in each category shared the same context, but the problems that had to be solved were different. Finally, the learners engaged in a knowledge test and solved 12 (8 near and 4 far) transfer problems.

Study 1 focused on the transient nature of animated models which might make them incomprehensible for novices because they lack the prior knowledge to attend to the relevant parts. The modality, pacing, and segmentation guidelines were used to decrease this type of ineffective cognitive load. Two 2 x 2 experiments, investigated the relation between the pacing of the presentation (learner vs. system paced) and the segmentation of the model (step-by-step vs. continuous) with respectively spoken (Experiment 1a, n=60) and written explanatory text (Experiment 1b, n=78). Learners observed animated models (adapted to the conditions) showing an expert solving the aforementioned problems. In Experiment 1a the results showed that on transfer, spoken explanations with continuous models yielded best performance with learner pacing, but segmented models yielded best performance with computer pacing. In Experiment 1b no effects were found on transfer, however, an additional analysis on the near and far transfer items showed the same interaction pattern as in Experiment 1a for far transfer. The same additional analysis in Experiment 1a revealed that the observed interaction was attributable to near transfer. Some possible explanations for these results are discussed.

Study 2 focused on guidelines to further maximize effective cognitive load and enable learners to engage in relevant learning activities. We argued that reflection prompts would provoke learners to be cognitively active and integrate new information with their prior knowledge. In line with the findings of Study 1 we also contended that meaningful reflection is only possible when learners have constructed a coherent verbal mental representation of the explanations and that written explanations would facilitate such coherence. In a 2*2 study (n=97) the factors Reflection (reflection on model vs. no reflection on model) and Modality (written vs. spoken explanations) were investigated. Learners observed 8 learner paced continuous animated expert models in which an expert solved the aforementioned problems. An interaction between modality and reflection was hypothesized, indicating that learners with written explanations would profit more from reflection and yield higher performance on the knowledge and transfer tests, whereas learners with spoken explanations would profit not from reflection. The results confirmed the hypothesis for transfer, but not for the knowledge test. Possible explanations and implications for these results are discussed.

Study 3 (n=60) investigated whether engaging learners in relevant learning activities by alternating between observing and practicing would be effective. We argued that observing and practicing yield different ways of processing information. The schema constructed through observing can then be further refined by practicing the skill. By alternating between observation and practice, the schema will be enriched. In the first condition learners observed eight animated expert models in which the aforementioned problems were solved. In the second condition the participants had to solve the eight problems themselves. After each problem they received feedback on the correctness of their solution. In the third condition, the participants started each problem category with a problem which was solved by an animated expert model. The second problem in the problem category they had to solve on their own, but they received feedback on the correctness of their solution. Continuous, learner-paced animated expert models accompanied by written explanatory text were used. We hypothesized that learners in the alternation condition would outperform learners in the other conditions on both retention and transfer. Data will be available in February 2007.

During the presentation the theoretical (e.g., what does this mean for the modality effect) and practical implications will be discussed.

Hypermedia learning environments gain increasing influence within educational contexts. Such environments, contrary to rather system-controlled multimedia environments and other traditional forms of learning (like books or upfront lessons), offer a high amount of learner control which includes the option to select and combine different representational formats (e.g., text, graphs, static or dynamic pictures) on the one hand and to get access to information in a linear as well as in a nonlinear way on the other hand. Advantages of hypermedia environments thus include that they can foster active, constructive, self-regulated, and adaptive learning. However, learner control is not one-dimensional, but depends on the decisions to be made by learners, namely, pacing, sequencing, content control, representation control, and interactivity of dynamic representations. Until now, not much is known on how to design effective hypermedia environments, in particular with regard to combining different representational formats and providing an optimal level of learner control. Our own prior studies, for instance, investigated whether well-established multimedia design principles can be applied to hypermedia environments as well. These principles state to (a) use multiple representations (Multimedia principle, Mayer, 2001; MERs, Ainsworth, 1999), (b) to use spoken rather than written text for narrated animations (Modality principle, Mayer, 2001) and (c) to avoid presenting redundant information (Redundancy principle, Mayer, 2001). A hypermedia environment containing worked-out examples was used to teach high school students how to calculate complex-event probabilities. We compared experimental conditions that included different representational formats (i.e., combinations of written text, audio text, and interactive animations) as well as different levels of learner control (low versus high). Our findings were partly surprising in that, for instance, results revealed that conditions with single representations outperformed those with multiple representations, contradicting the multimedia principle. The modality principle could also not be confirmed as there were no differences between conditions containing spoken and those containing written text. However, we found evidence for the redundancy principle as learners with written or spoken text alone outperformed those with a combination of identical written and spoken text. As for learner control, we found that all learners, regardless of their prior knowledge, benefited more from environments with predefined representational formats. Additionally, it could be shown that learners did not make much use of their representational freedom.

There are several possible explanations for these findings. On the one hand, it might simply be that animations and spoken text did not reveal the expected benefits because they were not designed in an optimal way. On the other hand, learners might have lacked metacognitive knowledge on the benefits of different representational formats which could have prevented them from effective learning.

The current study aims at overcoming these flaws by investigating whether improved animations and different forms of instructional support foster effective learning. We are again working with a hypermedia environment that aims at conveying basic principles of probability theory by means of worked-out examples. The level of learner control in our environment is high and involves pacing, sequencing, content control, representational choice and interactivity of representations. Learners can work through the environment in a nonlinear fashion by using hyperlinks. For each worked-out example, they can choose whether they want to retrieve the solutions in a purely mathematical format or enriched with written text, spoken text, animations or any combination of these representational formats. Before and after the learning phase, participants have to solve different types of transfer problems which are meant to assess prior knowledge on the one hand and performance as well as knowledge gains on the other hand.

Instructional support is implemented by using a 2*2 design. Firstly, we are varying the factor metacognitive support. Participants in the experimental condition with metacognitive support receive a modelling video prior to the learning phase. This video shows them an exemplary “good” learner who navigates through the environment and chooses worked-out examples and representational formats according to his/her needs and learning progress. For example, it is demonstrated that the learner who does not understand an example, should carefully look at it a second time to rehearse the important aspects.

Secondly, we are varying the factor representational prompting. As being mentioned before, for each worked-out example, participants can choose which representational format or combination of representations they want to use for retrieving the solution steps of the example. In the experimental condition with representational prompting, participants who want to choose a certain format receive a screenshot of the respective format accompanied by a spoken explanation about the corresponding advantages and disadvantages.

These variations result in the following four experimental conditions: (1) No metacognitive support / no representational prompts, (2) Metacognitive support / no representational prompts, (3) No metacognitive support / representational prompts, and (4) Metacognitive support / representational prompts. We expect that learners who receive both metacognitive support and representational prompting, will display more elaborated learning strategies and accordingly reach higher performance scores compared to learners who only receive one form of instructional support. Those learners, in turn, should outperform learners who neither receive metacognitive support nor representational prompting.

The study is currently being conducted within German high schools. We expect final results by the end of January.

 

 

Ainsworth, S. E. (1999). A functional taxonomy of multiple representations. Computers and Education, 33, 131-152.

Mayer, R. E. (2001). Multimedia learning. Cambridge, MA: Cambridge University Press.

Multiple representations in learning materials provide unique benefits when learners are to gain a deep understanding. However, they often do not lead to the ex­pected results, especially because the learners do not integrate the different representations. Due to such problems, it seems wise to instructionally support the integration and understanding of multiple representations. In the present experiment, we employed worked-out examples and tested the effects of multi- vs. mono-representational solutions, an integration aid in form of a flashing-colour-coding procedure, and scaffolding self-explanation prompts (‘fill-in-the-blank‘ explanations; cf. Figure 1) on learning processes (i.e., self-explanations) and learning outcomes (i.e., conceptual and procedural knowledge).

            170 school students learned about probability theory under four mono-representa­tional and four multi-representational conditions: (1) ’Pictorial/ no prompts‘, (2) ’pictorial/ prompts‘, (3) ’arithmetical/ no prompts‘, (4) ’arithmetical/ prompts‘, (5) ’pictorial and arithmetical/ no prompts/ no integration aid‘, (6) ’pictorial and arithmetical/ prompts/ no integration aid‘, (7) ’pictorial and arithmetical/ no prompts/ integration aid‘, (8) ’pictorial and arithmetical/ prompts/ integration aid.’

Planned contrasts showed that multi-representational solutions and an integration aid fostered conceptual knowledge but did not influence the acquisition of procedural knowledge. Furthermore, scaffolding self-explanation prompts elicited rationale-based self-explanations (i.e., giving reasons why the principle is as it is) and principle-based explanations (i.e., assigning meaning to a solution step by identifying the underlying domain principles) and thereby also fostered conceptual knowledge (scaffolding self-explanation effect). However, they also evoked incorrect explanations (e.g., misconcepts, confusion of two principles, etc.) that impaired the acquisition of procedural knowledge (paradox self-explanation prompt effect). Evidently, multi-representational solutions and corresponding instructional procedures differ in their advantage for learning specific knowledge types. Thus, it is only for certain learning goals that learners should master the demanding task of translating between two representations.

Multiple representations in learning materials provide unique benefits when learners are to gain a deep understanding. However, they often do not lead to the ex­pected results, especially because the learners do not integrate the different representations. Due to such problems, it seems wise to instructionally support the integration and understanding of multiple representations. One support procedure is to design the learning materials in a way that helps the learners to figure out which elements in different representations correspond to each other. Additionally, self-explanations are suited above all to foster learning when different information formats have to be integrated. The present study aims to analyse the effects of multiple representations, an integration aid in form of a flashing-colour-coding procedure, and scaffolding self-explanation prompts (‘fill-in-the-blank‘ explanations; cf. Figure 1) on learning processes (i.e., self-explanations) and learning outcomes (i.e., conceptual knowledge and procedural knowledge).

In the present experiment, we employed worked-out examples from the domain of probability theory and addressed the following hypotheses: (a) Multi-representational examples foster conceptual and procedural knowledge. (b) An integration aid fosters conceptual and procedural knowledge. (c) Scaffolding self-explanation prompts foster conceptual and procedural knowledge. (d) Scaffolding self-explanation prompts foster high-quality self-explanations.

The participants of this study were 87 female and 83 male students from grades 10 and 11 of German Gymnasiums (i.e., highest track in the German three-track system). The mean age was 16.21 years (SD = .91). The participants were randomly assigned to four mono-representa­tional (pictorial or arithmetical solutions) and four multi-representational (pictorial and arithmetical solutions) conditions: (1) ’Pictorial solutions/ no self-explanation prompts‘, (2) ’pictorial solutions/ self-explanation prompts‘, (3) ’arithmetical solutions/ no self-explanation prompts‘, (4) ’arithmetical solutions / self-explanation prompts‘, (5) ’pictorial and arithmetical solutions / no self-explanation prompts/ no integration aid ‘, (6) ’pictorial and arithmetical solutions/ self-explanation prompts/ no integration aid‘, (7) ’pictorial and arithmetical solutions/ no self-explanation prompts/ integration aid‘, (8) ’pictorial and arithmetical solutions/ self-explanation prompts/ integration aid’.

First, the learners worked on a pretest[1]. Then they entered the computer-based learning environment and studied four pairs of isomorphic worked-out examples. One focus of our learning environment was on understanding the multiplication rule. This rule can be understood by integrating the multiplication sign of the arithmetical code with the ramification in the tree diagram (for the numerator in Figure 1, there is twice one branch; for the denominator, there are five times four branches).

In the mono-representational conditions, a pictorial tree diagram or an arithmetical equation was presented. In the multi-representational conditions both a pictorial tree diagram and an arithmetical equation were provided (see Figure 1). In two of the multi-representational conditions the learners were supported in integrating the arithmetical information and the information from the tree diagram. Participants of the conditions with scaffolding self-explanation prompts received questions that should elicit self-explanations. Finally, the participants completed a post-test on conceptual and procedural knowledge. Conceptual knowledge referred to knowledge about the rationale of a solution procedure, that is, why the solution procedures are as they are. Procedural knowledge referred to problem-solving performance.

 

--- Figure 1 about here ---

 

 

Figure 1: Screenshot of the Learning Environment

 

For testing the hypotheses or addressing the research question, we performed F tests comparing the (aggregated) multi-representational groups with the mono-representational groups, the groups with an integration aid with the groups without an integration aid, and the self-explanation prompt groups with the no self-explanation prompt groups. We found that multi-representational solutions and an integration aid fostered conceptual knowledge but did not influence the acquisition of procedural knowledge. Furthermore, scaffolding self-explanation prompts elicited rationale-based self-explanations (i.e., giving reasons why the principle is as it is) and principle-based explanations (i.e., assigning meaning to a solution step by identifying the underlying domain principles) and thereby also fostered conceptual knowledge (scaffolding self-explanation effect). However, they also evoked incorrect explanations (e.g., misconcepts, confusion of two principles, or incorrect elaboration of a principle) that impaired the acquisition of procedural knowledge (paradox self-explanation prompt effect).

In a nutshell, our results showed that multi-representational solutions and corresponding instructional procedures differ in their advantage for learning specific knowledge types: Multi-representational solutions, an integration aid, and scaffolding self-explanation prompts fostered conceptual knowledge but did not influence or even hindered the acquisition of procedural knowledge. Evidently, it is only for certain functions and associated learning goals (i.e., conceptual knowledge) that learners should master the demanding cognitive task of translating between two representations. Apparently, it was not necessary to integrate the different representations in order to reach the goal of acquiring procedural knowledge. In contrast, it might have been sufficient – and even more parsimonious – to simply concentrate on the arithmetical equations for later problem solving. Thus, only if the intended learning goals require multiple representations, they should be provided. The scaffolding self-explanation prompts elicited high-quality self-explanations and fostered conceptual understanding. However, these benefits did not come for free. We conclude that the learners reached the upper limit of their working memory capacity by interpreting the representations with respect to conceptual knowledge so that correct essential processing with respect to procedural aspects was hindered.

---

In this study, the influence of instructional approach on learning outcomes has been investigated. Many studies often use a single instructional approach which is claimed to lead to good learning results. However, the question which instructional approach leads to the best results is rarely addressed. In this paper we will answer this question by comparing four instructional approaches: (a) inquiry learning, (b) observational learning, (c) hypermedia learning, and (d) example based learning. Besides instructional approach, representational code has been taken into account. In the learning environments, single and multiple representations were used.

Results were taken from the experiments of the projects in the LEMMA (Learning Environments, MultiMedia and Affordances) cooperation. To ensure that the four projects could be compared, common elements were developed. This resulted in all projects using the same concepts in the domain of probability theory, the same pre- and posttests measuring conceptual, procedural, intuitive and situational knowledge, and the same cognitive load measures. Results show that instructional approach influences all types of knowledge. For conceptual knowledge, example based learning outperformed all other instructional approaches, and inquiry learning scored significantly higher than observational learning. For procedural knowledge, again example based learning scored higher than the other three instructional approaches. And this time, inquiry learning outperformed hypermedia learning. For intuitive and situational knowledge, example based learning scored significantly higher than observational learning as well as hypermedia learning. The analyses show that in this comparison example based learning is the best instructional approach followed by inquiry learning. In the symposium, we will discuss these results and their theoretical implications. Furthermore, within instructional approaches differences were found between representational codes. However, in the comparisons across instructional approaches, the type of instructional approach appeared to be of more influence than the representational code.

Since the introduction of the computer into the classroom, a whole range of new computer-based learning environments has been developed. The subject matter in these environments can be presented to the learners in a wide range of different ways. It is an invitation for instructional designers to present the subject matter in such a way that learners are facilitated in their learning processes. A factor that plays an important role in learning and understanding a new domain is the instructional approach that is chosen in the learning environment. Many studies have been performed in order to evaluate the effectiveness of learning environments. Most studies, however, only address a single instructional approach which is claimed to lead to good learning results. However, the question which instructional approach would lead to the best results is rarely addressed. In this paper we will answer this question by comparing four instructional approaches. The first instructional approach concerns observational learning in which learners learn from observing others performing a task or solving a problem. This approach is supposed to be effective as the emphasis of observational learning lies not only on the procedure itself but also on the rationale behind this procedure (Collins, 1991; van Gog, Paas, & van Merriënboer, 2004). The second instructional approach is example-based learning in which learners learn from multiple examples consisting of a problem formulation, solution steps and the final solution. This approach is assumed to be effective, as learners are relieved from finding the solution on their own (Paas, Renkl, & Sweller, 2003). Instead, they can use their cognitive resources to concentrate on understanding the solution and the underlying principles. The third instructional approach concerns hypermedia learning. A hypermedia learning environment is a nonlinear computer environment in which information represented by different media is stored in nodes that are interconnected by hyperlinks. Hypermedia learning environments can be characterised by a high degree of learner control, which is supposed to lead to deeper understanding as they allow for adaptive information retrieval, thereby enabling active, flexible, and constructive learning (Spiro & Jehng, 1990). The final instructional approach concerns inquiry learning in which the learner must deduce concepts and variables from an underlying model by performing experiments and inferring information from the collected data. This type of learning is a constructivistic way of learning in which active and meaningful learning are supposed to lead to good learning results (de Jong & van Joolingen, 1998).

Another factor that influences learning is the representational codes that are used in the learning environment. For instance, studies have shown that pictures are processed easier than words (e.g., Larkin & Simon, 1987), although this effect seems not to be consistent over domains (e.g., Cheng, Lowe & Scaife, 2001) and depends on the amount of information in the representation and the effort it costs to infer this information (Palmer, 1978; Larkin & Simon, 1987). In the learning environments of the four projects, single and multiple representations were used including arithmetical representations, pictorial representations, textual representations, audio representations, and combinations of these. It was expected that learning outcomes are affected by a combination of instructional approach and type of representation.

Results were taken from the experiments of the projects in the LEMMA (Learning Environments, MultiMedia and Affordances) cooperation. To ensure that the four projects could be compared, common elements were developed. This resulted in all projects using the same concepts in the domain of probability theory, the same pre- and posttests measuring conceptual, procedural, intuitive and situational knowledge, and the same cognitive load measures. In total, 619 German and Dutch subjects (mean age 16.3) participated. It appeared that prior knowledge differed significantly between groups, so this variable was taken into account by using it as a covariate in the analyses. Results show that instructional approach influences all types of knowledge. For conceptual knowledge, learners in the example based learning environment outperformed learners in all the other instructional approaches, and learners in the inquiry learning environment scored significantly higher than the learners in the observational learning environment. For procedural knowledge, again example based learning scored higher than the other three instructional approaches. And this time, learners in the inquiry learning environment outperformed the learners in the hypermedia learning environment. For intuitive and situational knowledge, the results were similar. Learners in the example based learning environment scored significantly higher than the learners in the observational learning environment as well as learners in the hypermedia learning environment. The analyses show that in this comparison example based learning is the best instructional approach followed by inquiry learning. In the symposium, we will discuss these results and their theoretical implications. Furthermore, within instructional approaches differences were found between representational codes. However, in the comparisons across instructional approaches, the type of instructional approach appeared to be of more influence than the representational code.

 

Cheng, P.C.H., Lowe, R.K., & Scaife, M. (2001). Cognitive science approaches to understanding diagrammatic representations. Artificial Intelligence Review, 15, 79-94.

Collins, A. (1991). Cognitive apprenticeship and instructional technology. In L. Idol & B.F. Jones (Eds.), Educational values and cognitive instruction: Implications for reform (pp. 121–138). Hillsdale, NJ: Lawrence Erlbaum.

van Gog, T., Paas, F., & van Merriënboer, J.J.G. (2004). Process-oriented worked examples: Improving transfer performance through enhanced understanding. Instructional Science, 32, 83-98.

de Jong, T., & van Joolingen, W.R. (1998). Scientific discovery learning with computer simulations of conceptual domains. Review of Educational Research, 68, 179-202.

Larkin, J.H., & Simon, H.A. (1987). Why a diagram is (sometimes) worth ten thousand words. Cognitive Science, 11, 65-99.

Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive load theory and instructional design: Recent developments. Educational Psychologist, 38, 1-4.

Palmer, S.E. (1978). Fundamental aspects of cognitive representation. In E. Rosch & B.B. Lloyd (Eds.), Cognition and categorization (pp. 259-303). Hillsdale, NJ: Lawrence Erlbaum.

Spiro, R.J., & Jehng, J.-C. (1990). Cognitive flexibility and hypertext: Theory and technology for the nonlinear and multidimensional traversal of complex subject matter. In D. Nix & R.J. Spiro (Eds.), Cognition, education, and multimedia (pp. 163-205). Hillsdale, NJ: Erlbaum.