Cognitive Load Theory is Wrong?
“[Rethinking the boundaries of CLT] may allow reconciling seemingly contradictory results from studies of the effectiveness of worked examples … and studies within the frameworks of productive failure and invention learning that have reportedly demonstrated that minimally guided tasks provided prior to explicit instruction might benefit novice learners.” -Kalyuga & Singh
For those of you who know me and follow my blog, you know that I am heavily influenced by Cognitive Load Theory because I believe it is an excellent framework to help me think about teaching and learning. However, you may have noticed that there are certain aspects of Cognitive Load Theory that have been bothering me as of late. For example, in my last post, I reflected a bit on the expertise-reversal effect and how Cognitive Load Theory tends to view most learners as novices. To be frank, I find this rather uninspiring. Similar to the ending of The Force Awakens – it leaves me wanting more.
Over the past couple weeks I have had the opportunity to discuss some really interesting articles with a couple of awesome educators. One article that fell into my lap was “Rethinking the Boundaries of Cognitive Load Theory in Complex Learning” by Kalyuga & Singh. As an aside, it may be important to note here that Kalyuga was one of the authors contributing to this important text I commonly refer to as The Bible – so I utterly respect and value this fellow’s opinion. Let me tell you, I was not disappointed.
The reconceptualization of cognitive load theory proposed in this paper was triggered by the attempts to reconcile some empirical evidence that seemingly contradicts established findings of this theory.
So what empirical evidence is Kalyuga & Singh speaking of? And how does it contradict Cognitive Load Theory? Perhaps a good place to start are with the assumptions of Cognitive Load Theory.
The typical way Cognitive Load Theory is framed is based on the assumption that acquisition of schemas in the long-term memory is the end-goal of an instructional task. As such, the varying types of cognitive load are defined with this assumption in mind, which leads to particular instructional techniques being selected to decrease working memory load in certain situations. For example, it is suggested for novice learners that explicit instruction should be favoured to minimally guided techniques, since minimally guided techniques may involve using search methods (such as means-end analysis) that overload working memory space, impeding schema acquisition. The situation is different for expert learners, since experts already have well-developed domain-specific knowledge. Explicit instruction tends to be ineffective for experts, since too much attention is spent trying to integrate the pre-existing knowledge with the externally provided support, decreasing working memory space.
A growing body of knowledge in the areas of productive failure and preparation for future learning seem to contradict Cognitive Load Theory. That is, certain well-designed minimally guided tasks given to novice learners before explicit instruction tend to be beneficial when compared to beginning with explicit instruction. (For a taste, one of Kapur’s articles is here.)
It Gets More Complex
In lieu of the above empirical evidence, Kalyuga & Singh suggest that we need to rethink the boundaries of Cognitive Load Theory in the context of complex learning. In this article, complex learning refers to tasks involving “multiple learner activities that may have different goals.”
Some of such goals may indeed differ from the acquisition of domain-specific schemas and therefore require corresponding learner activities and instructional methods for their achievement that are different from the activities and methods that are best suitable for learning domain-specific schemas.
Now this is interesting! If we relax the assumption that our end-goal is acquisition of domain-specific schemas, then we can allow the empirical evidence that Kapur and Schwartz (and others) have given us into the realm of teaching and learning. In fact, Kalyuga & Singh give a possible simplified version of instructional goals in complex learning. First, one might consider low-level goals, or goals related to creating necessary prerequisites for schema acquisition. Items that may belong to this level include motivation to learn, activation of prior knowledge, engagement with the task, searching for deep patterns (as opposed to surface characteristics), or making students aware of gaps in knowledge. From this level, we move to mid-level goals, which include any goals that may be related to acquisition of domain-specific procedures and concepts. Here, we can turn to Cognitive Load Theory as a well-developed and tested framework. Finally, Kalyuga & Singh suggest a third tier of high-level goals, which may include generalization of procedures or flexibility in performance. [NOTE: this is simply a suggested model. For a more tangible model, see this article by Kirschner & van Merriënboer.]
Borrowing and Reorganizing
As Cognitive Load Theory began to emerge, Sweller suggested connections between natural information processing systems and the architecture of human cognition. One premise that came out of this evolutionary perspective was the borrowing and reorganizing principle. This principle states that natural information systems will borrow information from other stores (in human cognition, this information is borrowed from other people) rather than create new information. For example, during sexual reproduction the information held in your genome contained borrowed information from both our mother and father that was reorganized into a unique genome.
The borrowing and reorganizing principle underpins the worked example effect (that is we borrow the solution procedure from the long-term memory of a knowledgeable individual). Much research around the worked example effect has shown us that if acquisition of domain-specific schemas are the end-goal of an instructional activity, then worked examples are an efficient way to reach this goal. However, might it be that learning is not as simple as natural information processing systems? This is exactly the argument of Kalyuga & Singh:
The processes of learning biologically secondary information patterns (schemas) in human cognition may require additional learning activities that are not present or even required in other natural information processing systems.
This means the borrowing and reorganizing principle, and its associated consequences, may not extend to these particular activities.
In closing, I want to make note that this post is not meant to downplay the significance of Cognitive Load Theory. I like to think of it more as an attempt to nest Cognitive Load Theory into a larger schema of learning. Currently, this means we may need to be considerate of borrowing aspects from other well-developed learning theories and reorganizing them into a bigger picture of best-practices for our students.