The effects of learners' prior knowledge, self-regulation, and motivation on learning performance in complex multimedia learning environments
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
I INTRODUCTION 1
Research Questions 3
Significance of the Problem 4
Definition of Terms 6
II CONCEPTUAL FRAMEWORK 9
Prior Knowledge 10
III LITERATURE REVIEW 18
Prior Knowledge 18
Goal Orientation 33
Task Values 39
Summary of Literature Review 42
IV THE METHOD 47
Research Questions 47
Research Hypotheses 49
Recruitment and Background 49
vi Learning Materials 51
Measures: Predictors 52
Measures: Criteria 56
Data Analysis 58
V RESULTS 62
Preliminary Analysis 62
Model Testing 64
VI DISCUSSION AND CONCLUSIONS 74
Prior Knowledge 74
Summary of Discussion 83
Limitations of Study 85
Future Directions 86
A PRE KNOWLEDGE TEST 101
B SELF-REGULATION MEASURE IN COMPUTER- ASSISTED LEARNING (SRMC)
C SELF-EFFICACY MEASURE 108
D GOAL ORIENTATION QUESTIONNAIRE 109
E TASK VALUES QUESTIONNAIRE 112
F THE SCRIPT CONCORDANCE TEST FOR CAROTID ARTERY DISEASE
G MISSING VALUE ANALYSIS 120
H INTER-CORRELATIONS BETWEEN THE OBSERVED INDICATORS
LIST OF TABLES
1 The Number of Consented Students 50
2 Descriptive Analysis of Factors 62
3 Missing Pattern Analysis 63
4 Maximum Likelihood Parameter Estimates for the Full Measurement Model
5 The Fit Indices for the Original and Reduced Models 68
6 Unstandardized and Standardized Estimates of the Path Coefficients in the Reduced Model
LIST OF FIGURES
1 The Variation of Three Kinds of Cognitive Load 12
2 Outline of Conceptual Framework 17
3 Hypothesized Structural Model 46
4 WISE-MD Screenshot 52
5 Standardized Path Coefficients and Residual Variances of the Variables in the Hypothesized Structural Model
6 Standardized Path Coefficients and Residual Variances of the Variables in the Reduced Structural Model
As Whitehead declared, “knowledge keeps no better than fish” (1929, p.98) suggesting that too much knowledge exists and most knowledge quickly turns out to be inert or useless. Since Whitehead’s proclamation in the early 20th
century, advances in technology in the mid of the 20th
century have helped people easily acquire information regardless of the accuracy of the information source. Also, the rapidly changing knowledge demands that people keep searching for new knowledge. Therefore, students in the 21st century are required to have the ability and readiness to engage in lifelong learning, acquire and evaluate information, and collaborate with other students to solve complex learning problems (Law, Lee, & Chow, 2002). As one of the complex learning environments, which characterize 21st century learning, medical education also faces this knowledge dilemma. There is a vast amount of rapidly changing information, some proportion of which is becoming obsolete at any one time (Quirk, 2006). It is a myth that the regular curriculum in medical schools can teach students all the necessary medical knowledge they will need as practitioners (Smith, 1985). Medical students, who need to have various and essential clinical experiences, should be skilled learners who are able to gain knowledge not only from classroom didactics but also from practical experiences interacting with faculty members and
patients. However, studies have shown that not all medical students are skilled learners (Kalet, Coady, Hopkins, Hochberg, & Riles, 2007; Quirk). Medical knowledge is growing rapidly (Martin & Bunnett, 2004; Quirk, 2006). In order to fill the knowledge gap and provide standardized case-based resources to students, many medical schools have developed computer-based, multimedia learning environments (Harden & Hart, 2002). Multimedia presentations have great potential to support learning in that they can provide multiple types of information such as text, sound, and images (Mayer, 2005). In spite of the potential advantages, multimedia in education may impede effective learning if the characteristics of learners and tasks are not considered thoroughly in instructional design (Paas & Kester, 2006). The role of learner characteristics in learning has been investigated in aptitude- treatment interactions (ATI) studies (Cronbach & Snow, 1977), which indicate that the effects of educational treatments differ depending on personal characteristics such as a learner’s prior knowledge, learning styles, and personality (Jonassen & Grabowski, 1993). However, in traditional ATI research only prior knowledge has been consistently associated with learning (Boutwell & Barton, 1974; Federico, 1999 as cited in Kalyuga, 2007; Cronbach, 2002) and is critical to consider in instructional design since high prior knowledge students engaging with learning material designed for low prior knowledge students may demonstrate a temporary loss of learning. This phenomenon, the expertise reversal effect, described by Kalyuga and his colleagues takes into account a more comprehensive understanding of human cognitive processes and explains how the
success of instructional techniques depends on the level of a learner’s prior knowledge (Kalyuga, Ayres, Chandler, & Sweller, 2003). While traditional ATI studies consistently demonstrate the central importance of a learner’s cognitive processes on learning, cognitive processes are influenced by the learner’s level of self-regulation and motivation (Mayer, 1998; Short & Weissberg- Benchell, 1989). When a learner knows when and how previously stored knowledge is used in learning, and when he or she is motivated in learning tasks, learning becomes more effective and efficient. Therefore, it is critical to consider learners’ self-regulation and motivational processes in complex learning (van Merriënboer & Sweller, 2005). Even though many researchers agree that these three components - cognition, self- regulation, and motivation - are important to explain learning processes, few studies have investigated their combined effects. Therefore, this study aims to explore the relationship among medical clerkship students’ prior knowledge, self-regulation, and motivation on learning performance of complex tasks in multimedia learning environments.
1. What are the direct and/or indirect influences of prior knowledge on medical clerkship students’ comprehension and clinical reasoning in multimedia learning environments? 2. What are the direct and/or indirect influences of self-regulation on medical clerkship students’ comprehension and clinical reasoning in multimedia learning environments?
3. What are the direct and/or indirect influences of motivation (self-efficacy, goal orientation, and task values) on medical clerkship students’ comprehension and clinical reasoning in multimedia learning environments?
Significance of the Problem
Educational studies have identified that successful learners have balanced cognitive capacity, self-regulation, and motivational beliefs (Butler & Winne, 1995; Kauffman, 2004; Vermunt, 1996). Successful learners have more elaborated schemas, knowledge structures stored in long-term memory, compared to unsuccessful learners. These elaborated schemas help learners clarify the significance of new information and facilitate knowledge acquisition (Short & Weissberg-Benchell, 1989). While cognitive capacities can trigger better learning, this trigger is not automatic (Vermunt, 1996). In other words, a successful learner must be able to plan study activities, monitor his or her learning progress, and recognize and ameliorate weak points of learning. Furthermore, even though a learner’s cognition and self-regulation are important learner characteristics, they do not suffice for successful learning (Mayer, 1998). A learner must also be confident with his or her capability in a specific knowledge domain and put a high value on learning, in order to optimize the learning process. In summary, successful and sustained learning requires the interaction of multiple variables of learner characteristics. Previous research shows that learners in computer-based, multimedia learning environments need more self-regulation and motivational beliefs than those in traditional
learning environments because computer-based learning environments are usually structured as open-ended (Azevedo, Cromley, & Seibert, 2004; Greene & Land, 2000; Hadwin & Winne, 2001, Moos & Azevedo, 2008) rather than as structured tutorials. In the open-ended learning environments, learners are required to manage multiple resources available to them and be highly self-motivated in the learning processes. Therefore, it is paramount that a learner balance cognitive skills, self-regulation, and motivational beliefs (Kauffman, 2004; Ten Cate, Snell, Mann, & Vermunt, 2004) in computer-based, multimedia learning environments. Studies on self-regulation and motivation have attempted to show the importance of balancing cognitive, metacognitive, and motivational components (Schunk & Zimmerman, 2008; Short & Weissberg-Benchell, 1989). However there are few studies investigating the direct and indirect effects among these three learner characteristics, especially in multimedia learning environments. Furthermore, considering that medical education aims to educate medical students as experts who can have various and essential clinical knowledge and experience (Quirk, 2006), it is critical for these students to be cognitively, metacognitively, and motivationally prepared as skilled learners. Understanding the relationship among a medical student’s prior knowledge, self-regulation, and motivation on learning performance will help instructional designers to develop adaptive learning environments which will benefit most of learners in medical education.
Definition of Terms
Clinical reasoning: Using cognitive skills to efficiently and accurately generate hypotheses and strategically gather data in clinical settings (Charlin, Tardif, & Boshuizen, 2000). Cognition/ Cognitive: The function or description of a function of information processing which includes recall, comprehension, and application of knowledge (Guenther, 1998). Cognitive architecture: The manner in which cognitive structures such as working memory and long term memory are organized (Sweller, 2003). Cognitive load: The total amount of mental activity imposed on working memory at an instance in time (Cooper, 1998). Comparative fit index (CFI): Based on the non-centrality measure, it is computed as [d(Null Model) - d(Proposed Model)]/d(Null Model), where d = χ 2 - degree of freedom of the model (Kenny, 2009). Complex learning: The situation in which learners must integrate knowledge, skills, and attitudes through instructional materials and transfer what is learned to real work settings (van Merriënboer, Kirschner, & Kester, 2003; van Merriënboer & Sweller, 2005). Expert learner: The person who has ample, essential prior knowledge in a specific domain and well-honed self-regulation skills (Chi, Glaser, & Farr, 1988). Expertise Reversal Effect: The apparent loss of learning which occurs when expert learners use instructional materials that were designed with novice learners in mind (Kalyuga et al., 2003).
Mastery goal orientation: The individual preference for an absolute or intrapersonal standard in learning or achievement (Elliot, 1999). Metacognition/ Metacognitive: A student’s or a description of a student’s awareness of his or her own cognitive activity and of the methods employed to regulate his or her own cognitive processes (Brown, 1978). Model Chi-Square: the product (N - 1) F ML , where N - 1 are the overall degrees of the freedom in the sample and F ML is the value of the statistical criterion minimized in ML estimation (Kline, 2005). Motivation/ Motivational: The process or description of a process whereby goal- directed physical or mental activity is induced and maintained (Schunk, Pintrich, & Meece, 2008). Multimedia learning environments: Learning environments in which materials are presented in both verbal and pictorial forms (Mayer, 2001, 2005). Performance approach goal orientation: The preference for setting individual goals based on a positive normative standard in learning or achievement (Elliot, 1999). Performance avoidance goal orientation: The preference for setting individual goals based on a on a negative normative standard in learning or achievement (Elliot, 1999). Prior knowledge: A student’s declarative and procedural knowledge in a specific domain stored in his or her schema prior to learning (Dochy, Segers, & Buehl, 1999). Root Mean Square Error of Approximation (RMSEA): Based on the non- centrality parameter, it is computed as √[((χ 2 /df) - 1)/(N - 1)], where N is sample size and df is the degree of freedom of the model (Kenny, 2009).
Self-efficacy: The judgment of personal capability in a specific domain (Bandura, 1986). Self-regulation / Self-regulated learning: Self-generated thoughts, feelings, and actions for attaining learning goals (Zimmerman, 1998). Standardized Root Mean Square Residual (SRMR): the standardized difference between the observed covariance and predicted covariance (Kenny, 2009). Task values: A person’s subjective values of a learning task or activity (Wigfield, Hoa, & Klauda, 2008). The Tucker-Lewis index (TLI): Also called non-normed fit index (NNFI), its formula is [χ2/df(Null Model) - χ2/df(Proposed Model)]/[χ2/df(Null Model) - 1] (Kenny, 2009).
Complex learning challenges students to both integrate knowledge, skills, and attitudes through learning materials and then to transfer what is learned to real work settings (van Merriënboer et al., 2003; van Merriënboer & Sweller, 2005). In comparison simple learning targets one type of learning outcome such as knowledge, skills, or attitudes. Medical education, because it is aimed at preparing physicians for medical practice, is a prime example of a highly complex learning domain. Multimedia learning resources are rapidly being introduced into complex learning domains like medical education; however, it is not yet clear whether such resources are designed well enough to achieve their full learning potential or avoid common pitfalls (Chumley-Jones, Dobbie, & Alford, 2002; Kaelber, Bierer, & Carter, 2001; Martin & Bennett, 2004; McKimm, Jollie, & Cantillon, 2003). Many medical educators and instructional designers lack the theoretical and practical guidance needed to develop multimedia learning environments with state-of-the-art design features. And little is known about how individual learner characteristics of previously highly successful adult learners, such as medical students, influence learning in the complex multimedia learning environments.
In order to examine the effects of individual characteristics on learning performance in multimedia medical learning environment I reviewed the relevant literature on learners’ prior knowledge, self-regulation, and motivational domains.
A learner’s prior knowledge, his or her domain-specific declarative and procedural knowledge stored in schema prior to learning (Dochy et al., 1999), plays a primary role in learning achievement (Mayer, 1998; Kalyuga, 2007). Expertise studies (Chi, 2006; Chi et al., 1988; Ericsson, 2006; Sternberg & Frensch, 1991) have shown that learners perform best when they come to the learning situation with domain-specific knowledge and elaborated networks of knowledge. This effect has been attributed to the understanding that when faced with a learning task humans attempt to connect new information to existing schemas. When there is no relevant prior schema humans search for a solution randomly through trial-and-error (Kalyuga, 2007). This random search processes imposes unnecessary cognitive load and impedes learning (Sweller, 1988). Therefore sufficient domain-specific prior knowledge maximizes the effectiveness of any specific instance of learning.
Effect of Prior Knowledge on Cognitive Load
To understand the mechanism underlying the impact of prior knowledge on learning performance one must understand the limits of human working memory and the concept of cognitive load. Cognitive load is the term which describes “the total amount of mental activity imposed on working memory at an instance in time” (Cooper, 1998, p.
11). It is generally believed that within human cognitive architecture working memory is limited and long-term memory is unlimited (Sweller, 1999; Tuovinen & Sweller, 1999). It results both in limited short-term storage of information as well as limited ability to process information (Renkl & Atkinson, 2003; van Gog, Ericsson, Rikers, & Paas, 2005). Therefore, learning is reduced if the task requires mental activity beyond the available capacity. For instance, learning a novel task, one that has not yet been organized in long- term memory, is challenging because it requires many more mental resources to accomplish compared with a simple task. There are three distinct types of cognitive load, intrinsic, extraneous, and germane (Sweller, 1999). Intrinsic cognitive load, is the mental load inherent to a task, and is determined by the extent to which relevant task elements are interactive (Paas, Renkl, & Sweller, 2003; Renkl & Atkinson, 2003). That is, the task imposing high intrinsic cognitive load consists of highly interconnected elements (Kirschner, 2002). Complex learning tasks have high intrinsic load in that a learner must consider many cognitive elements simultaneously to achieve a successful learning outcome. Intrinsic cognitive load is a characteristic of the task which is difficult to reduce through instructional manipulation (Paas et al.; van Merriënboer & Sweller, 2005). Extraneous cognitive load is the unnecessary mental load attributed to poorly designed instructional material (Paas et al., 2003; Sweller, 1999). Because human beings have a limited working memory, the more extraneous cognitive load in a learning task, the less working memory capacity is available for cognitive activity intrinsic to the task and germane to the comprehension (Gerjets & Scheiter, 2003; Paas et al.; Renkl &
Atkinson, 2003). Extraneous cognitive load disturbs schema acquisition and automation (Sweller). Lastly, germane cognitive load is that mental effort required for learners’ comprehension (Paas et al., 2003; Renkl & Atkinson, 2003; Sweller, 1999). The more difficult a task is, the more germane cognitive load is imposed on a learner. However, this kind of cognitive load facilitates learning because it occurs as a consequence of the learner actively conducting elaborations, abstractions, comparisons, and inferences in an attempt to master the learning task (Gerjets & Scheiter, 2003). Both extraneous and germane cognitive load can be manipulated through instructional design (Brünken, Plass, & Leutner, 2003; Paas et al.).
Figure 1. The variation of three kinds of cognitive load (Adapted from Cooper, 1998, p.14) Figure 1 compares two instructional approaches (case 1 with more extraneous cognitive load than case 2) to the same task in the same learner (equivalent intrinsic and germane cognitive load and working memory). In case 1, fewer resources are available for intrinsic and germane cognitive load since considerable mental resources are already invested in extraneous cognitive load. As extraneous cognitive load is decreased in case
2, through better instructional design, more mental resources are freed up for intrinsic and germane cognitive load. All else being equal, theoretically, learning outcomes would be expected to be better in case 1. Cognitive load researchers believe that knowledge structures stored in long-term memory, also known as prior knowledge, influence the intrinsic cognitive load of an instructional event (Sweller, 2003; van Merriënboer & Sweller, 2005). Learners with more prior knowledge will experience less intrinsic cognitive load on a learning task than learners with less prior knowledge since they already have relevant schemas stored in long-term memory. Consequently they can use the freed-up working memory capacity to take on the germane and extraneous cognitive load. This leads to differing impact of the same instructional procedure depending on learners’ prior knowledge.
Indirect Effect of Prior Knowledge through Self-regulation and Motivation
Even though learners’ prior knowledge plays a primary role in learning achievement, cognitive factors do not fully explain or predict learning performance. Expert learners, defined as persons who have more relevant schema in long-term memory compared to others, also have been shown to have relatively strong non-cognitive characteristics such as self-regulation and motivation (Ericsson, 2006; Glaser & Chi, 1988), which predict learning performance independent of prior knowledge (Mayer, 1998). Therefore, beyond the direct effect of prior knowledge, this study examined the indirect effects of prior knowledge through self-regulation and motivation to provide a
more comprehensive understanding of the role of learners’ prior knowledge in complex multimedia learning environments.
Successful learners manage their own learning process cognitively, metacognitively, and motivationally (Zimmerman, 1986, 1989). These self-regulated learners use appropriate cognitive strategies such as surface or deep cognitive processing. In addition, self-regulated learners can plan, self-monitor, and self-evaluate their learning process. In terms of motivation, self-regulated learners perceive themselves as confident learners.
Self-regulated Learning as a Characteristic of Expert Learners
According to reviews on expertise studies, an expert has plenty of essential prior knowledge in a specific domain and appropriate self-regulation skills (Chi et al., 1988). With more prior knowledge, an expert can detect and recognize features, generate the best solution, and perform faster and more accurately with minimal mental effort (Chi, 2006; Ericsson, 2006). This automation of schemas also helps a learner save limited working memory capacity for other cognitive activities (Paas et al., 2003). Experts use a larger number of cognitive and self-regulated strategies than novices (Chi, 2006; Plass, Kalyuga, & Leutner, 2010). They tend to set productive goals and plan before performance, use appropriate strategies during performance, and self- evaluate after performance (Zimmerman, 1998). For instance, medical experts diagnose based on patients’ data whereas novice clinicians tend to form hypotheses prematurely
(Patel & Kaufman, 1995). In addition, experts in physics can judge the difficulty of a problem more accurately than novices (Chi, Glaser, & Rees, 1982). In summary, even though learners’ domain-specific knowledge is a key component of expertise, it is not sufficient to explain the superior performance of an expert. Experts have not only more domain-specific knowledge but also more self- regulated strategies compared to novices.
Self-regulated Learning as an Aptitude
Self-regulated learning has been viewed as both an aptitude and an event (Muis, Winne, & Jamieson-Noel, 2007; Snow, 1996; Winne, 1997; Winne & Perry, 2000). When viewed as an aptitude self-regulated learning is defined as “a relatively enduring attribute of a person that predicts future behavior” (Winne & Perry, p. 534) or a disposition of the learner. When viewed as an event or “a transient state embedded in a larger, longer series of states unfolding over time” (Winne & Perry, p. 534) it is dynamic and situational and expressed as products produced through self-regulated learning. Individual learners will bring different levels of self-regulatory skills to any learning task (Kanfer, Ackerman, & Heggestad, 1996), and although self-regulatory skills can be taught (Boekaerts, 1997; Zimmerman, Greenberg, & Weinstein, 1994) in the absence of self-regulation training, baseline individual differences (treated as an aptitude) in self-regulatory skills are important modifiers of learning achievement. Therefore, in this study I examined the effects of initial levels of learners’ self-regulation on learning.
Indirect Effect of Self-regulation through Prior Knowledge or Motivation
A learner may well have significant knowledge, but if he or she does not have self-regulated strategies to control and monitor cognitive processes, or in other words if he or she does not know when or how the knowledge should be used, the domain-specific knowledge will not be appropriately used in learning processes. In terms of the relation between prior knowledge and self-regulation, cognitive load theory, emphasizing that human working memory is very limited, would predict that when a learning task is complex and demanding, the learner may not have sufficient mental capacity for self-regulation (Plass et al., 2010). On this basis, Winne (2001) suggested that “reducing demands of the task, schematizing and automating information, and off-loading information from memory to the environment” (p.170) would promote self-regulated learning. In addition, self-regulated learning is strongly related with learners’ motivation (Schunk & Zimmerman, 2008). Therefore, this study examined not only the direct effect of self-regulation but also its indirect effect through motivation on learning performance.
In addition to cognition and self-regulation, it is critical to consider motivation in complex learning (Mayer, 1998). Many studies have shown the role of motivational variables such as self-efficacy, goal-orientation, and task values as the driving force of cognitive and metacognitive factors and a determinant of successful learning (Schunk & Zimmerman, 2008). In particular, if learners are self-efficacious, they tend to use more cognitive and self-regulated strategies and show better performance (Bandura, 1986;
Pajares, 2008, Zimmerman, 2000). In addition, learners who seek to master learning materials or desirable outcomes in learning are more adaptive than learners who seek to avoid undesirable outcomes in learning (Ames & Archer, 1988; Fryer & Elliot, 2008). When learners consider a learning activity important to them, enjoy the activity itself, and think it useful to them, they tend to show better learning achievement (Eccles (Parsons), Adler, Futterman, Goff, Kaczala, Meece et al., 1983; Wigfield et al., 2008). Because these motivational constructs are essential factors of effective learning the direct effects of self-efficacy, goal orientation, and task value on learning performance were examined in this study. The indirect effects of these motivational factors through prior knowledge and self-regulation were investigated as well. In summary, in order to develop a comprehensive understanding of learner characteristics in complex multimedia medical learning environments, the current study examined the direct and indirect effects of learners’ prior knowledge, self-regulation, and motivation on learning performance. The conceptual framework is illustrated in Figure 2.
In order to promote an understanding of the effects of cognitive, metacognitive, and motivational learner characteristics in multimedia learning environments, this literature review focuses on learners’ prior knowledge as cognitive learner variables, learners’ self-regulated strategies as metacognitive learner variables, and learners’ self- efficacy, goal orientation, and task values as motivational learner variables, and describes studies that show how these factors have impacted on learning.
The effects of instructional intervention are different depending on a learner’s individual characteristics (Jonassen & Grabowski, 1993), in particular, a learner’s prior knowledge in the specific domain examined. Prior knowledge can be defined as a person’s declarative and procedural knowledge in a specific domain stored in his or her schemas before learning (Dochy et al., 1999). Thomson and Zamboanga (2004) examined whether a learner’s prior knowledge could be used as a predictor of his or her learning achievement even after controlling his or her general aptitude measured by the ACT scores. In order to measure a learner’s prior knowledge in psychology, they administered two different pretests to 353 undergraduate students. The first pretest was a 25 multiple-choice test with 5 alternatives. The second
pretest asked students to identify whether a series of statements about psychological ideas was accurate by using a 4-point Likert scale (ranging from 1 = Very sure it’s false to 4 = Very sure it’s true). According to the results, the two pretests were significantly correlated, and each test was also positively correlated with exam performance. In particular, the pretest on prior psychological knowledge was more highly associated with the performance than one on accuracy in psychological statements. However, this study showed that prior coursework in psychology was not related to achievement. Hierarchical linear regression analyses revealed that the pretest of psychological knowledge was a significant predictor of learning achievement even after controlling for a learner’s ACT scores, and course participation and involvement. Keig and Rubba (1993) examined students’ ability to make translations between three representations of the structure of matter, and its relationship to specific knowledge on the representations, reasoning ability, spatial reasoning ability, and gender in high school chemistry classes. Forty-two high school students participated in this study. According to the results of regression analyses, the most significant predictor of learning performance in translation of representation variable was domain specific prior knowledge. In addition to prior knowledge, learners’ reasoning ability measured by the Group Assessment of Logical Reasoning (GALT) was a significant predictor. These two significant variables accounted for 28% of the variance in the learning performance. However, both spatial reasoning ability and gender were not significant. Therefore, Keig and Rubba (1993)
suggested that students’ performance could be predicted by the level of their prior knowledge. Kalyuga, Chandler, Tuovinen, and Sweller (2001) investigated the role of a learner’s prior knowledge in multimedia-based learning. According to previous cognitive load studies, problem solving may impose unnecessary working memory load on learners because a mean-end strategy, which is used to figure out the goal state, the current problem state, and the difference between two states, requires significant conscious effort from a learner (Sweller, van Merriënboer, & Paas, 1998). On the other hand, when worked examples including solution examples about the problem are provided, learners may experience less working memory load and better learning performance. However, Kalyuga and his colleagues assumed that the worked example effect could be different depending on the level of a learner’s prior knowledge. Whereas inexperienced learners tend to benefit from worked examples because additional information can help learners’ cognitive processes, experienced learners may consider more guided information as redundant. In order to examine the effect of learners’ prior knowledge on the worked example effect, 24 trade apprentices were randomly assigned to either the worked examples group or the problem solving group using multimedia-based learning materials. This longitudinal study was conducted in two stages with different level of complexity. Kalyuga et al. (2001) assumed that learners could acquire schema through stages of continuous training sessions. The results indicated that the worked examples group was superior to the problem solving group in an early stage but the effect was reversed in a later stage. These