Investigating the influence of transitory information and motivation during instructional animations

This study investigated the cognitive load theory prediction that the inconsistent findings concerning the effectiveness of instructional animations are exacerbated by their transitory nature. Three groups were compared who received different but equivalent forms of instruction in learning a topic in economics. One group received an animation presentation with integrated text and diagrams, a second group received a static diagram presentation with integrated text and diagrams, and the third group received a static diagram presentation with non-integrated text and diagrams in a classical split-attention design. Results indicated that the animated design was superior to the static integrated design only on test questions that closely resembled the presented information. No other significant group differences were identified. Furthermore a battery of self-rating measures of cognitive load and motivational items indicated that test performance was predicted by motivation and concentration levels, rather than by mental effort or task difficulty. This study had two main aims. The first aim was to advance the research on instructional animations, and secondly, to explore the use of multiple self-rating measures of cognitive load. The paper begins by briefly describing some of the main principles of cognitive load theory as it used as the theoretical framework for the study. In the last decade cognitive load theory has developed into a well-known theory of learning and instruction. From its early development in the 1980s, where it focused on explaining why problem solving was an ineffective method of learning, it has more recently become a theory, which considers cognition as a natural information processing system similar to evolution by natural selection (Sweller, 2003, 2004). At the centre of CLT is the interaction between long-term memory (LTM) and working memory (WM). Whereas WM is very limited in both capacity and duration (see Miller, 1956; Peterson & Peterson, 1959), the capacity of LTM is extremely large, and stored information can last a lifetime. Furthermore it is argued by CLT theorists that information is stored in LTM in the form of schemas (Chi, Glaser & Rees, 1982). By recalling schematic knowledge from LTM into WM, multiple bits of connected information can be treated as single elements, thus relatively increasing the capacity of WM. Hence humans are able to learn complex material and solve difficult problems in spite of the limitations of WM. However, when learning novel information, where schemas are not available, the limited capacity of WM can be a severe impediment to learning. Much of the research into CLT has investigated the various conditions under which WM resources become overstretched and how these conditions can be alleviated by instructional designs that take into account the human cognitive architecture (van Merriënboer & Sweller, 2005). CLT identifies three categories of cognitive load: intrinsic, extraneous and germane (see Sweller, van Merriënboer & Paas, 1998). Intrinsic load is the load caused by the complexity (element interactivity, see Sweller & Chandler, 1991, 1994) of the materials to be learnt, extraneous load is the load caused by the instructional procedures, and germane load is the load directly invested in schema acquisition. Whereas intrinsic load is considered fixed, dependent upon only the prior knowledge of the learner (se Kalyuga, Ayres, Chandler & Sweller, 2003), extraneous load is under the control of the instructional designer. An effective design lowers extraneous load and induces germane load. Over two decades CLT research has identified a number of strategies, such as worked examples and the modality effect, to optimize learning environments (see Sweller, 1999; van Merriënboer & Ayres, 2005). In the last few years, researchers have extended the research on CLT to outline a theory to explain the conditions under which instructional animations may be designed more effectively. The results of studies that have used dynamic representations (animations) as a learning tool are somewhat inconclusive. A review by Tversky, Morrison and Betrancourt (2002) found that animation was no more effective than static representations, although it was postulated that animation might be best used in environments that “convey real time changes and reorientations in time and space”. (p. 257). Mayer, Hegarty, Mayer and Campbell (2005) found static diagrams to be more effective than animations when learning about mechanical systems. In contrast, a meta-analysis by Höffler and Leutner (2007) found evidence that a number of studies showed animation to be more effective than static pictures, particularly when the animation was highly realistic and/or procedural motor knowledge was involved. As well as research that has compared static diagrams with animations, a number of studies have looked to improve the animated designs themselves. For example, researchers have shown that animations can be improved by more directed extraction of information (Lowe, 1999, 2003, 2004) user interactivity (Hasler, Kersten & Sweller, 2007), segmenting the presentation into smaller chunks (Mayer & Chandler, 2001; Moreno, 2007), and by signaling information (De Koning, Tabbers, Rikers & Paas, 2007; Lusk & Atkinson, 2007; Moreno, 2007). Whereas a number of these compensatory tactics have been successfully employed and a number of explanations tendered, the research field has lacked a comprehensive theory to explain why animations are sometime effective and other times not. However, CLT has provided some further insights. CLT theorists argue that animations can be highly transitory in nature (see Ayres, Kalyuga, Marcus & Sweller, 2005; Ayres & Paas, 2007a, 2007b). Animations that involve information disappearing from the screen as the animation progresses, forces learners to process current information while trying to remember previous information, leading to heavy demands on working memory load. In such situations, animation creates an extraneous cognitive load. This argument gives a plausible explanation as to why stopping an animation (user-interactivity) or segmenting it into smaller parts is beneficialboth strategies lower working memory load by negating the effects of transitory information. The study reported here continues the research into the transitory effect. We argue that another way to overcome transitory information it to ensure that no information disappears from the screen. To test this prediction the following study was completed. Main hypotheses of the study The study consisted of three treatment groups and participants were required to learn some economics content, a domain rarely used in CLT research. One group (Animated) was required to learn from an animated presentation. To overcome the transitory effect the computer-based presentation was designed so that no information disappeared from the screen. To eliminate other potential sources of extraneous cognitive load, namely the split-attention effect (see Ayes & Sweller, 2005; Chandler & Sweller, 1992), all text and diagrams were integrated together. A second group (Static integrated) received an equivalent presentation to the Animated group except that a set of static diagrams replaced the animation. Again all text and diagrams were integrated to avoid a split-attention scenario. The third group (Static non-integrated) was identical to the Static integrated group, except that the text and diagrams were deliberately kept separate in order to induce a splitattention effect. By inducing such an effect it was expected that the least learning would occur and therefore this group served as a control condition. The following two hypotheses were tested: • Hypothesis 1: Learners will benefit more from an animated instructional format than a static format • Hypothesis 2: Learners will benefit more from an integrated format than a non-integrated format. Cognitive load measures In much of cognitive load theory research a global self-rating measure of cognitive load has been used based on the original Paas (1992) 9-point scale. This scale was worded as follows for the lowest mental effort rating (1): “In solving or studying the preceding problem I invested very, very low mental effort”, and was administered after students had taken the test questions. However, Van Gog and Paas (2008) have pointed out that many researchers have used variations of the original scale and changed the wording of the scale considerably. Often using difficulty instead of mental effort. Further, most researchers administer the scale following instruction rather than the test phase. Accordingly, Van Gog and Paas have argued that these variations may pose a threat to the validity of the instrument. In more recent times researchers have become interested in measuring intrinsic, extraneous and germane load individually. For example, Ayres (2006) has measured changes in intrinsic load, while others have investigated methods to measure all three loads (see Cierniak, Scheiter, & Gerjets, 2007; Scheiter, Gerjets, & Opfermann, 2007). An interesting finding in the Cierniak et al. study was that a concentration measure was found to mediate the split attention effect. Furthermore, it has been argued by Moreno (2007) that motivational factors mediate learning and influence cognitive engagement. Whereas research into motivation is a huge field in its own right, it is rarely used in CLT. As a result of these new directions and findings, a number of subjective cognitive load and motivational measures were introduced into this study. The broad aim of the subjective measures, which was exploratory in nature, was to collect information on: • Potential measures of intrinsic, extraneous and germane load • Concentration and motivation measures • Predictors of performance