“It can be argued that there is no conceivable central executive, apart from long-term memory, in the human cognitive system. Nevertheless, central executives are postulated in cognitive theories, with Baddeley’s (1992) working memory theory providing the best example. We argue that an independent central executive dissociated from knowledge held in long-term memory results in an infinite regress of central executives. If knowledge held in long-term memory does not have the attributes normally attributed to a central executive, we are immediately faced with the question of how a knowledge-independent central executive determines its actions. We require another central executive governing the first central executive that in turn requires a third central executive, etc. The problem is eliminated by assuming that knowledge in long-term memory acts as a de facto central executive…”
John Sweller, Paul Ayres and Slava Kalyuga (2011). Cognitive Load Theory
In any discussion of cognitive science, it is important to remember that it is the experiments that are real and any mental structures or processes that we suggest to account for the results of these experiments are just models, ripe for replacement as new evidence becomes available. It is particularly important to remember that the models we propose will colour the way we think about the evidence.
A good example of this issue is the difference between Baddeley’s model of working memory, which contains something called a ‘central executive’ which performs such functions as directing and inhibiting attention, and research into ‘executive functions‘ which typically include working memory, control of attention and behaviour, and cognitive flexibility. Both of these models cannot be true at the same time. Working memory cannot logically be both part of something and also be a construct that includes that same something. Nevertheless, both of these models assume that there are ‘domain general cognitive skills that exert top-down control over attention and behaviour‘. It is this assumption that, to some extent at least, cognitive load theory challenges.
For instance, if you start believing that executive functions are general purpose then the possibility arises that if you train these executive functions then this will lead to general gains in performance. This is the thinking behind working memory training. Typically, subjects are given a task to complete that draws heavily on working memory – which is essentially the ability to manipulate new information. The more tasks they complete, the better the performance, leading some to suggest that working memory has been improved. Unfortunately, the improved performance tends to be limited to the task used in training or very similar ones and does not lead to a general boost.
An alternative explanation is that working memory is not trainable and that what is happening during working memory training is that subjects are developing schemas in long-term memory that relate to the specific training task. It is the development of these schemas that boost performance, but they are task-specific and that is why no general improvement occurs.
We can be drawn into similar arguments about the other executive functions such as the ability to direct or inhibit attention. Perhaps some of this is innate in a way that is similar to working memory capacity. Indeed, perhaps the two are linked. However, just like working memory, we should not assume that simply because we model these as general capacities, training will deliver general improvements. Again, it is possible that training leads to the development of task-specific schemas that improve performance on the training task but do not deliver a boost on unrelated tasks that supposedly depend upon the same executive functions.
A new study by British researchers sheds some light on this question. Unusually for an education study, it is a preregistered randomised controlled trial. Preregistration is a bonus because it means that researchers have to declare what they will consider as positive or negative evidence in advance, removing the possibility of mining the data once they have the results to see if they can find a way of analysing it that supports their hypotheses.
The researchers trained three- and four-year-olds on a variety of working memory and ‘inhibitory control’ tasks. The latter required them to ignore a stimulus that pointed in the wrong direction and so is relevant to the control of attention. They then investigated whether this training would lead to improved maths performance and improvements in other executive functions.
They found that executive functions did correlate with maths performance and could be a cause of children from lower socio-economic background performing less well in maths. However, they found that while the children in the experimental group improved on the training tasks, there was no improvement in other measures of executive function and no improvement in mathematics. It seems that if executive functions are indeed general, they cannot be generally improved by training.
Cognitive load theory takes an unusual stance on executive functions. It gives working memory capacity – our raw processing power – primacy and even doubts the existence of a central executive coordinating the other executive functions, suggesting instead that they are a function of schema-building in long-term memory. You may not agree with this unusual view, but at least the model appears to be broadly consistent with the data.