A new review paper has been published buy Chen, Castro-Alonso, Paas and Sweller in the journal Frontiers in Psychology: Educational Psychology. The beauty of this journal is that it is open access and so you can read the whole thing without a subscription.
And I recommend reading it. It is perhaps a little technical, but it does contain an excellent summary of some key principles of cognitive load theory that I don’t think I’ve seen before in an open access article. However, this is an aside.
The aim of the paper is to try to understand the sometimes contradictory findings of research into ‘desirable difficulties’. These are strategies that temporarily make learning a little harder but that supposedly lead to better retention and transfer of learning. Desirable difficulties include the generation effect, the testing effect (or retrieval practice) and varying the conditions of practice. You are probably familiar with some of these strategies because they tend to be the ones promoted by evidence-based education blogs such as The Learning Scientists.
Unfortunately, we don’t always find a positive effect for introducing desirable difficulties. Chen et al., draw on evidence to suggest that this is because we need to take element interactivity into account.
Element interactivity is a controversial idea that has been introduced into the framework of cognitive load theory. In essence, the element interactivity of a task is the number of elements it requires a student to process in parallel in working memory. This will be affected by the complexity of the task itself, as well as the level of expertise of the student. A student who can draw on schema held in long-term memory to process elements will need to process fewer in working memory.
Chen et al. use the example of learning the word for ‘cat’ in a foreign language. Although this may be a difficult thing to do and retain, the element interactivity is low. There is just one item – the word for ‘cat’ – to process. Learning a list of such words would involve processing one item at a time in this way and so it would be a low element interactivity task.
In contrast, a novice trying to solve 2x + 5 = 3 for the first time must process the numbers and operators in parallel. Doing something to the 5, for instance, has implications for the rest of the equation. In this case, we would say element interactivity is quite high. It would not be high for a mathematical expert, however, because she can quickly and easily apply solution methods held in long-term memory.
Chen et al., draw on experimental evidence to suggest that we obtain desirable difficulty effects when element interactivity is low and we see the reverse effect when it is high. For instance, generation effect studies might ask students to either study pairs of opposite words e.g. “inside/outside” or generate the second word for themselves e.g. “inside/o_____”. This is a low element interactivity task and so there is working memory capacity available to do some extra work. In this case, this extra work probably helps by making meaningful (semantic) connections between the pairs of words and that’s why we see a gain for the students who generate the second word for themselves.
However, when element interactivity is high, such as when novices learn to solve algebra problems or to use a bus timetable, it is better to study completed worked examples than to try to generate solutions.
The implications for teachers are that we should be careful about how we introduce desirable difficulties. It is probably a good strategy to make use of the generation effect for learning labels and definitions from the very start of learning. However, if we want students to learn how to solve particular problems or implement procedures involving a number of interdependent steps, we should probably ensure that we have embedded this learning with full, explicit instruction, before we seek to make the process a little harder.