Where did germane cognitive load go?
The evolution of a theory
During the development of cognitive load theory, three types of cognitive load have been proposed. The first is intrinsic load. This is the load that is unavoidably imposed by a learning task. Many people assume that this is a fixed constraint, it is not and it depends on the learner and can be manipulated by, say, breaking larger tasks down into smaller components.
The second type of load is extraneous load. This is load imposed by items that are not essential to task completion. For instance, a question may contain irrelevant information that a learner needs to consider and reject. However, this also provides an illustration of the interplay between these two types of load. If the learning objective is to be able to answer questions that contain distracting information — a legitimate goal as learners gain more expertise — then this information becomes intrinsic.
The third type of load that has been proposed is germane cognitive load. This was initially thought of as an additional kind of load that related to the learning process itself. You could think of it as the load imposed by transferring knowledge from working memory to long-term memory. However, this formulation of germane cognitive load causes problems.
The first problem with germane cognitive load is a philosophical one. The second is an empirical one.
The philosophical issue is that germane cognitive load makes cognitive load theory unfalsifiable. If a teaching method reduces cognitive load and is effective, we may conclude that it has reduced extraneous load. If a teaching method increases cognitive load and is effective, we may conclude it has increased germane load. Continuing this line of thought suggests all possible experimental results can be explained. No matter what happens in an experiment, cognitive load theory can explain it.
This may initially seem like a strength but it is a fatal flaw. Unless a theory can make predictions that can be tested as right or wrong, it is an unscientific theory. This is a key flaw in popular sociological theories about structural oppression or ‘whiteness’ and often why trained scientists find such theories implausible. The original formulation of germane load therefore renders cognitive load theory unscientific and this is a bad outcome.
Nevertheless, some teaching methods really do increase cognitive load and are effective , while others really do reduce cognitive load and are effective. So, we still need an explanation.
The practical problem with germane load is one of load switching. If we take the view that reducing extraneous load will lead to the switching of resources from to germane load then when we reduce extraneous load, the overall load should stay the same. This is not what is found experimentally.
The somewhat unsatisfactory solution to the switching problem is to identify germane load as a function of intrinsic load. To maximise learning, we therefore want to maximise intrinsic load and minimise extraneous load. This explains the experimental results but feels unsatisfactory because it doesn’t provide much of an explanation for how transfer from working to long-term memory happens.
So, how can we explain the potentially conflicting experimental results where both raising and lowering cognitive load can lead to better outcomes under different circumstances?
This is where the concept of element interactivity comes in. Element interactivity is a property of a learner completing a task and it depends as much on the learner as it does on the task. For example, a learner who knows no algebra and who is introduced to the equation 5x=45 will have four separate elements to process — 5, x, = and 45. Moreover, these elements are in a relationship with each other — if we divide the left-hand side by five we have to do the same to the right-hand side. In other words, these elements interact with each other, with these interactions representing additional elements to process. Such a task could therefore well exceed the three-to-four element limit of working memory.
However, if we give this equation to anyone fluent in arithmetic and algebra, they are likely to involuntarily solve it and determine that x = 9. This is because they have activated the relevant schemas in long-term memory — perhaps by momentarily drawing them into working memory but we are not sure exactly how this works. To a relative expert, 5x=45 is therefore low in element interactivity, at least in terms of the elements and the interactions the expert will need to process independently in working memory.
Therefore, if we want to optimise intrinsic load — and hence germane load, although this term is a bit redundant now — to about three items, it would depend on learner expertise. For a novice we would have to consider ways to reduce load and for an expert, we would have to consider ways to increase it. Again, this could involve shifting conceptions of what we consider to be ‘intrinsic’ and this would depend on what we think are educationally relevant goals.
We must also note that some tasks are inherently low in element interactivity. For instance, imagine I want to teach you that the capital of Australia is Canberra. That is just one item and no interactions. In this case, it may always be preferrable to try to increase cognitive load. For instance, I could given you the stem C________ and ask you to guess before telling you the answer.
Using element interactivity in this way makes testable predictions. When load is high we should try to reduce it. When load is low, either inherently or because we are dealing with relative experts, we should try to increase it.
A teaching method that increases load for high element interactivity learning tasks, and is found to be more effective than one that stabilises or reduces it, would refute cognitive load theory.