Cognitive Load Theory and Instructional Design

Brian Chipperfield
Graduate Student
Educational Communications and Technology
University of Saskatchewan
April, 2004

Select a different paper
Download a copy of the entire paper

Cognitive Load Theory

Picture this. A grade nine math student has just been asked to open her textbook to page 54 and refer to section 1.3 subsection 1.3b and study the first example. This is what she sees.

The instructor then copies the example on the board and verbally describes the steps. As he is writing, the student tries to follow along with what he is doing at the board, with the example in the book and with what the teacher is saying. However, she is distracted by two things; the whispering of the two students behind her and the squeaky desk she is sitting in. She is trying not to move because she feels it is distracting to others. The teacher finishes the explanation and assigns ten questions due at the end of the session, which is thirty minutes away.

These are the things that may be going through our math student’s conscious mind at this moment.

  1. Because the page is all text it takes more time finding the first problem.
  2. Her eyes have to refer back to the example on the previous page.
  3. She is feeling worried because of the time limit on the assignment. She wants to go out with her friends tonight and does not want to have to do homework or stay late after school.
  4. She wishes that the two students in front of her would stop whispering and silently swears at the squeaky desk. She is sure her classmates are looking at her.

Regrettably, our student is suffering from cognitive load and if she and her teacher do not take steps to deal with it then she may not ever understand Algebra.

Computer Assisted Instruction has found a niche in modern day Education. The media savvy teacher has learned to augment his instructional bag of tricks by using computers to bring concreteness, more interest and reality to the instructional setting. Success of CAI depends upon the role the instructor has given the computer in assisting learners to grasp new skills and concepts. Success also depends on the role the instructor plays in the process. The instructor in this sense can fall into several categories:

  1. The teacher or parent who sits down one on one with the learner and shares the interaction.
  2. The teacher who guides the whole class through a computer mediated lesson, monitoring each user, helping them to troubleshoot through difficult concepts.
  3. The distance ed instructor who sets up the interface for online learners to use.
  4. The software developer who designs the interaction.

Good instructional design takes into account the role the on-screen presentation plays and the role the instructor plays in the experience for the learner. Good instructional design will help to avoid what our math student is experiencing as she tries to grapple with solving linear equations. Good instructional design therefore has to provide accessible skill attainment.

It is imperative that designers of computer based instructional materials keep learning theory in mind when one is setting up an interface to be used. Cognitive load theory is one area of learning theory research that has definite applicability to subjects that focus on problem solving skills such as Math, Physics, Chemistry and Computer Science

Cognitive Learning Theory would describes the learning process as follows:

  1. Our grade 9 math student takes the information (the written text, the auditory description from the teacher, the example from the boards) into her sensory memory.
  2. Her working memory (short-term memory) processes this information using two independent systems, the visual-spatial sketchpad and the phonological loop, both controlled by a central executive component. (Baddeley,1992).

Our math student doesn’t realize it but her working memory has a capacity limit of about seven units of information. (Miller, 1956). This limit is reduced if the bits of incoming information are dependant on each other and therefore must be retained in memory for the time it takes for understanding to occur. Our math student’s 3-step linear equation problem requires her to hold all information in her working memory until “she gets it”. This is because all the information is interrelated or as Baddeley puts it, has high interactivity. The limit of working memory can be extended if recoding or chunking were to occur. Chunking requires that she takes her prior knowledge of the solution steps, such as the four steps, isolating, grouping, simplifying and checking, and organize the incoming information into this schema.

This schema can then be processed in the working memory as one unit, freeing up space for the other information such as the two steps required to isolate the variable.

Figure 2 is an outline of the basic components of a schema we can call the basic linear equation solution schema. At the root of cognitive learning theory is schema formation. At one time our math student did not have this schema, this cognitive tool, stored in her long-term memory where she keeps all her other cognitive tools such as the one that enables her to operate her cellphone to access her email, the weather channel, the phone numbers of 150 friends and family. In fact this schema is made up of four sub schemas that she once had to learn and commit to long-term memory as single schema and then organize them into a more useful schema, the schema represented in figure 2. Learning requires that we build upon previous experiences (learned schema). As we experience new information it is up to our working memory to weave these new experiences into similar and related experiences we have had stored in long-term memory. This is what schema formation does for us. It is up to our working memory to form these schemas, commit them to long-term memory, retrieve them as single units to be used to process other information that has yet to be formed into schema.

In summary, our math student has room for seven units of information in her working memory. The linear equation problem is made up of highly interactive units of information so therefore she has room for maybe two or three units. The four-step process (isolating,

grouping etc) can be processed as one since from prior learning she has a schema ready for this job. The other two units (figure 3) can be processed easily and she can also use some of her working memory to build upon her present schema to include the two steps it takes to isolate the variable. (Figure 4)

“The primary role of an instructor is to transform the novice into an expert within a given subject area” (Cooper, 1990). The difference between a novice and an expert lies in the expert’s ability to categorize problems using schemas stored in long-term memory. The only two distinguishing features of expertise are:

  1. The expansive schemas (information networks) that experts hold, and
  2. The high level of automation (ability to perform tasks without concentrating) that experts exhibit.

Earlier it was noted that what makes good instructional design is a design that makes learning accessible. Accessibility then could be equated with ease of schema formation.

Cognitive load theory serves to describe those variables that hinder schema development. Cognitive load theory is based on the following tenets of cognitive learning.

  1. That short-term memory (working memory) is limited in capacity to about seven informational units.
  2. Long term memory is unlimited in capacity and is where all information and knowledge is stored.
  3. Knowledge is stored in long-term memory as schemas or schemata.
  4. Schemas, no matter how large or how complex, are treated as a single entity in working memory.
  5. Schemas can become automated.

Cognitive load can be of three distinct types, intrinsic cognitive load, extraneous cognitive load, and germane cognitive load. Kirschner (2002) Using our math student as an example we can picture her conscious memory mental effort in the following way:

  1. Intrinsic load is the load on memory required by the thinking task at hand. It serves to quantify how much of the working memory is used by the interactivity of the units of information being processed. The intrinsic load for our math student would be substantial. In the process of solving the linear equation all steps towards the solution are highly dependant upon each other.
  2. Learning, changing a novice to an expert and the formation of new schema adds to the load on working memory. Working memory must process this new information into advanced and more complex schema. Our math student moves closer to expert status as she integrates the two-step isolation process into the already complex solution schema. The load this process places on working memory is called germane cognitive load.
  3. The teacher in his presentation, the textbook in its format and the external distractions, the internal emotional concerns all represent extraneous cognitive load. These variables take up valuable working memory resources away from the learning task. Extraneous cognitive load does not contribute to learning.

Taken together the three load variables must stay within the limits of mental effort, the total cognitive load on working memory. This relationship is represented in figure 5.

For any given problem, task or learning behavior, I cannot be changed. However G + E can vary and are inversely proportional to each other. The more extraneous load the less room for germane load. Hence it is the instructional designers job to limit the amount of extraneous load and to build into their presentations activities that foster germane load or schema formation. Research has focused on what factors or variables can be modified to facilitate the increase in germane load and the decrease in extraneous load.

Research into Reducing Cognitive Load: Instructional Design Strategies

Instructional design and the The Goal Free Effect

Traditional methods of solving algebraic problems rely on means-end analysis. Sweller(1985). For example:

The novice learner must work backwards from the goal state (the value of a) to the initial problem state a = b – 4. In doing this he must keep all the highly interactive units of information in his conscious working memory producing high intrinsic load. The same problem written as goal-free would read as follows:

The novice learner in this instance only has to work from c=5 to b = c + 3 to find the value of c. The problem poses less intrinsic load on his working memory thus freeing up more mental effort for germane load to work on schema formation and transfer to long-term memory.

Instructional Design and the Use of Worked Out Examples

Much research has been done on the use of worked out examples to reduce cognitive load especially in the areas of Math, Physics and computer Science. Essentially it takes the familiar method of teaching a new concept, figure 8,

….. and applying a new twist. Research has shown that if you simply coordinate the worked out example with a similar, isomorphic problem to solve then less cognitive load is required. Figure 9.

Using the structurally similar example on the left as a model to solve the problem on the right requires that the learner only have to attend to each step at once. Repeated practice of this type would eventually led to schema formation and subsequent transfer to long-term memory. Knowing what the outcome will look like means he can attend to the process involved in isolating the required variable. Less intrinsic load is required.

Consider the area of mechanics. The relationship between velocity, distance and time can be expressed by the formula v=d/t where velocity is equal to distance divided by time. Physics students must use this formula to solve velocity, distance, and time problems. If time is the unknown to be solved for then students must have the schema to rearrange the formula for the purpose specified in the problem. Rearranging the formula requires that the student multiply both sides by t and divide both sides by v revealing the formula for time, t=d/v. This process goes beyond the schema needed to solve standard linear equations of one variable. Cooper uses this as an example where studying worked examples works best to facilitate schema formation.

The validity of this method of teaching has been under scrutiny. One of the questions being asked is why does this method facilitate schema formation better than the more traditional problem solving method. Chi et al, as sited in Trafton (1993), claim that studying examples works because students explain the examples to themselves. “Successful learners explain examples to themselves more fully than for less successful learners”(Trafton, 1993,pp6). And further, “successful learners attempt to explain to themselves how each line in a problem… is derived from the previous line.” Traftons research found that participants in his study who were given a set of worked out examples interleaved with problems to solve performed faster and more efficiently than participants who were given a set of examples to study first and then a set of problems to solve after. Van lehr et al, as sited in Trafton(1993), further argues that studying worked out examples is not enough. Students perform far more skillfully if they self-explain while studying worked examples. Chi et al (1989) as sited in Renkl (2002 ) coined this the self-explanation effect. Renkl’s research revealed that successful learners, in studying an example, assigned meaning to operators in the example problem, identified sub goals of operators in the example problem and anticipated the next step in the example problem.

Overall the research has shown that the use of worked out examples proves to have a greater potential for skill acquisition, especially among novice learners, than the standard problem solving method of learning problem solving skills. Studying worked out examples allows for more working memory resources to be assigned to schema formation particularly if they involve self-explanation to reinforce the rules and principles inherent in the type of problem to be solved.

Implications for instructional designers can be summed up as follows. Visual presentation of material would include a worked example alongside an example to be solved. The worked example would be shown on the screen first with perhaps audio clues to foster self-explanation of operators and anticipation clues. The problem to be solved would be near identical to the example. Interaction with the interface could include cues to have the user key on a self-explanation of the strategy he/she is using.

Instructional Design and the Management of Intrinsic Cognitive Load

Intrinsic cognitive load is under the control of the learner’s cognitive architecture and is determined by the amount of interactivity of information elements. Elements being “information that can be processed as a single unit of working memory.” (Pollack, 2002) Interactivity is discussed as how much elements depend on each other to be understood and therefore processed. Elements that are low in interactivity would be items that can be learned independent of each other. Pollack and his associates refer to learning new language vocabulary as an example. Conversely, elements of high interactivity would be the elements that make up the syntax of the language the learner is trying to master.

What if it proved to be beneficial to teach concepts isolated from their meaning? For learning highly interactive instructional material elements have to be incorporated into schema to limit the load on working memory. As Pollack notes highly interactive elements have to be processed simultaneously for schema to be produced. This presents a paradox. “Material can be understood once a schema has been constructed allowing all the elements to be processed in working memory simultaneously but until that point has been reached, the elements cannot be processed simultaneously in working memory and so cannot be understood.” (Pollack,2002,pp.64). Remember when your high school English teacher forced you to commit to memory Hamlet’s famous soliloquoy, “to be or not to be”. Perhaps he was using good instructional design. We memorized it word for word because we had to. It was a sequential act learning to recite the passage from memory, independent of its meaning. Once committed to memory however understanding it took less working memory resources. Forming schema relevant to its meaning was easier when the words and language were committed to memory.

Pollack hypothesizes that “initially learning some elements of information, even if comprehension is not possible, may ultimately increase a students understanding of a topic.” (Pollack,2002,pp.64). Pollack and his colleagues proposed that if highly interactive elements were presented as isolated elements first, to initialize the formation of prior schemas, and then interactive elements presented after schema formation then cognitive load would be reduced and learning would be optimized. The results of their research demonstrated that novice learners benefited from the isolated interactive element presentation. Students with prior knowledge of concepts showed little benefit. Pollack further attributes this to the presence of schemas in the expert group.

What can we generalize from this? Prior knowledge has to be considered when designing instructional media dealing with highly interactive information such as Algebra. Introducing the subtraction principal of equality in solving linear equations might require our learner to rediscover what equality means and what subtraction means in a concrete sense such that previously learned schema may be brought to the forefront and used in processing the concept. Perhaps an assessment of prior knowledge has to be built into instructional design. The use of rote learning techniques for assigning math symbol associations with math vocabulary could be used as a stepping stone to actually teaching the concepts.

Use of Multimedia in Instructional Design

Mayer and Moreno(2002) propose that multimedia learning involves three cognitive processes

  1. Selecting: where verbal information is processed as a text base and visual information is processed as an image base.
  2. Organizing: where the verbal base and the image base are applied to the yet to be learned concept.
  3. Integrating: where the learner builds connections between the two.

Their research into the effectiveness has yielded the following major principles of Multimedia design.

  1. Multiple Representation Principle – It is better to present an explanation in words and pictures than solely in words.
  2. Contiguity Principle – words and pictures are to be presented simultaneously rather than separate.
  3. Split Attention principle – present words as auditory narration rather than as visual on screen text.
  4. Coherence Principle – use few rather than many extraneous words and pictures.
  5. Modality Principle – students learn more effectively from animation and narration than from animation and text.
  6. Personalization Principle – students learn more effectively from animation and narration when the narration is conversational rather than formal style.
  7. Redundancy Principle – students learn more effectively from animation and narration than from animation, narration and text.

Mayer and Moreno (2002) promote the idea that instructional designers work from a cognitive theory of learning rather than from an information delivery viewpoint. A crucial design goal is to promote schema construction in the learner. Their research into the effects of the Multimedia design principles discussed above have shown that students are able to successfully generalize problem solving sets to post test problems. They summarize that to develop useful multimedia learning tools designer should construct these tools on the basis of three areas of research:

  1. Dual coding theory – which states that working memory can process visual and auditory information simultaneously with no adverse affect on cognitive load.
  2. Cognitive load theory.
  3. Constructivist learning theory - which implies that students should be guided to an awareness of being responsible for their own knowledge formation.

This research and more research along this vein will only help to narrow the gap between the useless and the useful use of computers in our schools. The use of multimedia, based on these and other design principles can have the effect of bringing the world into the classroom. Quicktime video, Flash animations, interactive Director movies created to teach simple isolated cause and effect concepts in science and in math could be effective learning aids and a viable alternative to the text and blackboard presentation. These multimedia tools can be used to:

      1. invoke the self-explanation effect.
      2. to present elements of a concept in isolation from each other.
      3. to present worked out examples alongside concrete images of the concept

Summary

Cognitive load is real. It is the reason I had to write this essay in long hand before I entered it into the computer. Attention to my keyboarding took up too much of my working memory leaving little room for creative thinking.

What will ensure the success of these endeavors is dependant on how much the creators andthe teachers themselves, take the learner and learning styles into consideration in preparing an interface. Perhaps instructional technologists could take more of a leadership role when the teacher becomes the subject matter specialist as well as the designer. Cognitive load theory, split attention theory, dual processing theory and the principles of multimedia design are not the entire solution to the poor design problem, but they do offer simple ways in which multimedia can achieve learning goals more effectively.

In closing, on the following page are three screens from a very highly touted and a very expensive piece of articulation software. The problems are obvious but I have summarized a few of them. These examples are more the norm through the program rather than the exception. It is obvious that learning theory was not taken into consideration when it was put together.

The format of the software however is very good. A speech-language pathologist can use this to help students with articulation problems learn sounds in all positions, in words, phrases and sentences. It has a recording function that allows the user to hear how close he/she is to the sound approximation. A more appropriate use of this format would be to have the ability to choose words and pictures for the target sounds that are more considerate of the learners’ vocabulary.

The problems shown here are very indicative of what is supposed to be well designed learning software. The structure is good, the interface is fast and appropriate but the content can be highly irrelevant to the learner and this creates a barrier to accessibility.

This is a ship or a boat. How many students in grade 1 would associate this with a bow? It is a poor choice of words for the goal of the lesson. It is obvious that the designers were not thinking of the end user.
This is a picture of three red flowers, many green leaves and a bee. Cognitive load theory would suggest that the user would key in on the other elements of the picture before finding the bee. A misuse of working memory relative to the goal of the lesson.
This is a tractor. Again the user would be trying to associate the word with the tractor. Perhaps if he is from a farming community he would eventually get it but why give him the task when he needs his working memory for the task at hand, learning how to make the hard “c” sound.

References

Baddeley, A. D. (1992). Working memory. Science, 255, 556–559.

Bannert, M. (2002) Managing cognitive load – recent trends in cognitive load theory, Learning and Instruction , 12 139-146

Cooper, G. (1990) cognitive load theory as and aid for instructional design, Australian Journal of Educational Technology, 6(2), 108-113

Gellevij, M., Van der Meij, H., De Jong, T., Pieters, J. (2002) Multimodal versus Unimodal Instruction in a complex learning context, The Journal of Experimental Education, 70(3), 215-239

Kalyuga,S., Chandler,P., & Sweller,J. (1999). Levels of Expertise and Instructional Design, Human Factors, 40(1)

Kirschner, P. (2002), Cognitive load theory: implications of cognitive load theory on the design of learning, Learning and Instruction, 12, 1-10

Mayer, R., Moreno, R. A Cognitive Theory of Multimedia Learning: Implications for Design Principles. Available at www.eng.auburn.edu/csse/research/research_groups/vi3rg/ws/mayer.rtf

Mayer, R., & Moreno, R. (2002). Aids to computer-based multimedia learning. Learning and Instruction, 12, 107–119.

Mayer, R., (2003), The promise of multimedia learning: using the same instructional design methods across different media, Learning and Instruction 13 125–139

Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review, 63,81–97.

Pollock, E., Chandler, P., Sweller, J. (2002) Assimilating complex information. Learning and Instruction 12 61–86

Renkl, A. (2002). Worked-out examples: instructional explanations support learning by self-explanations, Learning and Instruction 12, 529-556

Stark, R., Mandl, H., Gruber, H., & Renkl, A. (2002). Conditions and effects of example elaboration. Learning and Instruction, 12,39–60.

Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12, 257-285.

Sweller, J. (1994). Cognitive load theory, learning difficulty and instructional design. Learning and Instruction, 4, 295–312.

Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59-89.

Tabbers, H., Martens, R., & van Merrienboer, J. (submitted). Multimedia learning and cognitive load theory: effects of modality and cueing.

Trafton, J. (1993) Studying examples and solving problems: Contributions to skill acquisition. Available at http://citeseer.nj.nec.com/93611.html


Valcke, M., (2002) cognitive laod; updating the theory? Learning and Instruction 12, 1470154

Van Bruggen, J.M., Kirschner, P.A., Jochems, W. (2002) External representation of argumentation in CSCL and the management of cognitive load. Learning and Instruction 12, 121-138

Appendix A

Using the principles of good instructional Design – More software Assessments
This is a Cree language instruction program developed by OBI systems of Saskatoon. The software is useful as an audio visual aid for a presentation to a class. It can also be used individually at a workstation. This page borders on being too busy and would increase extraneous cognitive load. Using the cree alphabet for the months is a hindrance to schema formation and would further contribute to load. Students do not have much prior knowledge of the cree alphabet. Clicking on the month brings up the month in English, a visual and textual description of what the cree phrase stands for and the cree phrase as sound. Split attention theory would suggest it is better to just have the cree phrase coincide with the picture, the phrase in English (flying up moon) and the English name for the month.
This is a followup exercise to reading the story, the Three Little Pigs. Ideally an adult would be present with the child when working with this software. The goal of this exercise is to put the items in the right order. Audio narration explains the tasks and there is a help screen if needed. From a cognitive load viewpoint this screen presents a very difficult task for the user. Each of the items is taken from quite different parts of the story. To do this exercise the child and the adult helper would have to have the entire story in working memory. We could put these items in order but in the end we would not have a story. It would be better to sequence three or four sequential parts. Then you would have a complete segment that has meaning to the child.
The goal here is initial sound discrimination. Directions are given by the dog in a strong new Orleans accent. Users are to choose the object that has the different initial sound. One of the problems here is the choice of objects. Clown, crown and ghost would most probably be in the vocabulary schema of a child at this age but coast may not. Care should be taken to use vocabulary that is in the realm of the age group the software is directed at. The picture of “coast” is not clear. This may confuse the user as he is listening to the cards being read. This is a cognitive load issue as is the background. The wheel and scoreboard on the right are distractors and serve no purpose. A user might be curious about them. Curiosity requires use of working memory detracting form the task at hand.
The goal with this program is shape and colour discrimination. The interface is very simple in design and quite easy for the user to use. The child has to drag the correct colour or shape or both to the bears mouth. The bear cues the user by saying red or red square. It is well sequenced and would be very easy for a young child to use. Narration is clear and simple. There are no cognitive load issues here.
This is a Canadian History software package. The screens consist of maps such as this alongside scrollable text. The images are very poor and the text in the images is pixelated. Scrollable text presents a problem for working memory. The learner has to keep the entire passage in mind to fully understand its relevance to the diagram. Split attention theory would suggest that there is a high cognitive load present here. It would be more appropriate to have small snippets of text in a larger font alongside clearer diagrams.
This is a more accessible example of a history tutor. Text is large and the graphics are clear. Popup quicktime videos with narration and animation add interest to the learners experience.