The term cognitive load has been around for a long time but definitions have tended to shift and often depend on theoretical positions. Early researchers rarely attempted to define cognitive load at all. In a 1966 paper, Levine describes cognitive load as the amount of information that the observer is required to store in memory and later retrieve. Levine’s definition is quite different from that put forward by Palinko et al. (2010): the relationship between cognitive demands placed on the user by a task and the user’s cognitive resources.
Pierre Barrouillet approaches cognitive load through the lens of the Time-Based Resource Sharing model (TBRS) of working memory and defines it as the proportion of time during which processing occupies the bottleneck and impedes the memory task (more on this later). Sweller, alternatively, views cognitive load from a Cognitive Load Theory (CLT) position, where load is sub-divided into intrinsic, extraneous and germane. So popular has Sweller’s CLT become that it’s often conflated with cognitive load more generally. Indeed, the Wikipedia entry for cognitive load spends very little time defining and describing cognitive load and the majority of the article explaining CLT.
What many of these later definitions and models have in common is that they view mental resources as limited and related to working memory. Working memory is thought to have a capacity of between 5 and 9 (Miller, 1956) or 4 (Cowan, 2000) pieces of information. However, strategies such as chunking can increase this. The duration of working memory is thought to be between 18 and 30 seconds before it decays (Peterson and Peterson, 1959), but see this post on decay for a more critical evaluation. This means that if we try and carry out a task that exceeds these limits, performance is going to suffer. Similarly, the multicomponent model of working memory proposed by Baddeley and Hitch, would indicate that individual components of working memory can become overloaded. Say, for example, I ask you to read a passage from a book silently to yourself but also ask you to repeat the word ‘the’ over and over again in your head (a technique known as articulatory suppression). This articulatory suppression has increased the load because you are attempting to carry out two sound-based tasks simultaneously, resulting in a drop in comprehension of the passage you’re reading. In the language of the multicomponent model of working memory the phonological loop (the component dealing with sound-based information) is finding it hard to manage the resources at hand, leading to increased load and, thus, a reduction in performance.
Many things can increase load, including the difficulty or novelty of the task, emotional states such as anxiety and other factors including sleep quality. Perhaps you’re chatting with a friend and something comes on the radio that sounds interesting or important; you can either listen to the radio or carry on with the conversation, but very rarely can you do both. Similarly, attempting to complete a complex mathematical equation in a noisy environment is going to force us to lose focus because there is simply too much stimuli to process. This means that cognitive load is related to attention, being able to maintain our focus on the task in hand. Skills such as learning to drive place immense pressure on cognitive resources, which is why learner and new drivers need to concentrate much harder than experienced ones.
Not only do we have to process new information, we also have to store it and this, in itself, is a resource heavy activity. The role of attention was crucial in the formulation of early models of memory and has witnessed a resurgence of interest over the past decade or so as a way to investigate many of the unanswered questions around working memory. Pierre Barrouillet of the University of Geneva has attempted to answer some of these questions through the Time-Based Resource Sharing model (TBRS) of working memory.
Barrouillet proposes that when people carry out a task, the cognitive system rapidly switches between processing the information and storing it, so our focus of attention continually flips between carrying out the task and refreshing the memory trace. Storage and processing can’t take place at the same time and this results in a bottleneck where both functions are forced to compete for resources.
When the focus of attention is on processing, memory suffers; when the focus of attention is on storage, processing suffers. Crucially, when cognitive load is low, there is more time to refresh the memory trace, but when cognitive load is high, more resources are directed towards processing and retention is negatively affected; working memory span decreases as cognitive load increases. With any given task, there is a processing/storage trade-off.
Barrouillet, therefore, defines cognitive load as the proportion of time during which processing occupies the bottleneck and impedes the memory task. Crucially, this indicates that recall performance isn’t dependent on the duration of the task or the number of digits to be recalled but, rather, on the ratio between the amount of work to be done and the time allowed to do it, that is, successful recall depends on cognitive load.
Whichever definition we adopt and through whichever lens we choose to view it, cognitive load in concerned with the relative demands of a task and the mental resources we have available to meet those demands. Models of instructional design, such as Cognitive Load Theory and Load Reduction Instruction provide ways in which instructors can help ease this load. However, such models assume that human memory systems are modular and that memory decays from the short-term module if there is insufficient rehearsal. Yet even the notion of separate short-term and long-term stores isn’t a universally agreed proposal and contemporary investigations might suggest that older models (such as Atkinson and Shiffrin’s Multi-store model and Baddeley and Hitch’s Working Memory model) are due for retirement.
Different perspectives, therefore, lead to minor differences in emphasis. Some definitions and models lean towards the multi-component view of memory (the separation between short-term and long-term memory), while others may adopt an increasingly unitary approach (no separate stores or ‘types’ of memory) with an emphasis on attentional processes. Both Cowan’s Embedded Processes model and Barrouillet’s TBRS model are two examples that highlight the current debate.