Interleaving refers to the benefits of sequencing learning tasks so that similar items – two examples of the same concept, say – are interspersed with different types of items rather than being consecutive. This results in a more variable and challenging task but is associated with benefits in terms of memory and transfer, which apply to concept learning as well as other domains (Kang, 2016).
Interleaving tends to be contrasted with classroom tasks that are scheduled in blocks of similar items, with the latter termed a ‘blocked’ arrangement. For example, in Figure 1, item ‘A’ is interleaved with items ‘B’ and ‘C’.
Interleaving versus spacing
Interleaving can also be distinguished from a much better known memory phenomenon: the spacing effect. The latter refers to the benefit of incorporating time delays between learning and practice, leading to improved performance over educationally relevant time periods (Cepeda et al., 2008), compared to ‘massed’ items, where practice sessions occur close together. An interleaved presentation of material is inevitably spaced to some extent, given that sequencing items so that they are non-consecutive leads to a time delay between one example and the next. For example, in Figure 1, above, there are larger gaps between the instances of item A in the interleaving sequence than in the blocked sequence. However, a key difference between the two effects is that when items are spaced, the gaps between learning and practice needn’t include related material.
Indeed, the benefits of interleaving seem to depend on the mixing of related items, such as examples from similar categories. This may be because such scheduling puts different items side by side, improving the perception of differences between them (Kornell and Bjork, 2008). This is known as the discriminative-contrast hypothesis, and is supported by research into interleaving of unrelated categories. One such study (Hausman and Kornell, 2014) interleaved anatomy terms with Indonesian vocabulary and found no benefit. Additionally, interleaving seems to be especially effective when differences between items are subtle (Carvalho and Goldstone, 2014).
To avoid confounding interleaving and spacing, a number of studies have used filler activities to increase the spacing of blocked items, meaning that spacing remains the same in both conditions (see Figure 2).
In one such study, Birnbaum et al. (Birnbaum et al., 2013) presented images of types of butterfly, either in groups with four different pictures of the same species (blocked) or four pictures of different species (interleaved). Participants were later tested on novel examples, and identified them more accurately following the interleaved study. However, this benefit disappeared if the two types of presentation (blocked versus interleaved) both had their temporal spacing increased, due to this interfering with discriminative contrast (see Figure 3).
Both interleaving and spacing are what Bjork and Bjork (Bjork and Bjork, 2011) term ‘desirable difficulties’, i.e. strategies that make learning more difficult, but in a way that is beneficial. They are both widely recommended among those who aim to apply cognitive psychology to education, appearing, for example, among the ‘Six Strategies for Effective Learning’ recommended by the popular Learning Scientists blog (The Learning Scientists, n.d.).
The two strategies differ in their primary benefits. The spacing effect boosts memory – practice or restudy of material is more effective if spaced out over time, with forgetting reduced – while interleaving boosts inductive category learning and later transfer. Recent demonstrations of this include the categorisation of chemicals into types (Eglington and Kang, 2017) and the conceptual learning of science categories or examples (e.g. (Rawson et al., 2015)).
However, there are drawbacks to their use. Smith and Scarf (Smith and Scarf, 2017) note that for spacing learning across days to be helpful, a minimum initial level of experience is required. Davis et al. (Davis et al., 2017) have found that frequent switching between studying and test questions can be detrimental, while Kang (Kang, 2016) reasons that a hybrid approach can be beneficial, with new learning occurring via blocked practice, and interleaving used in a practice or consolidation phase.
The application of spacing and interleaving to the school-based learning of new scientific concepts is open to question. It is unclear how they interact in realistic situations and what effect interleaving and spacing have when combined in authentic classroom learning of concepts. The aim of this study was to address these questions.
An opportunity sample of 31 school pupils was used. All were between 16 and 17 years of age. Data was gathered during an end-of-year taster session, during which pupils sampled several subjects. They were entirely new to the topic being learned.
Design and materials
In order to make the tasks as authentic as possible, all materials were based around a school psychology specification, featuring psychological theories of phobias. The experiment aimed to reproduce the range of activities in a typical school class, and so learners were taught both concepts (types of phobia) and relevant research evidence.
Tasks were delivered via an online protocol. The target learning material was presented in two main phases. In the first, a research study pertaining to phobias was shown on two screens, one with a description of the study and one with evaluation points, with the latter either presented immediately (massed condition) or after the second phase (spaced condition). The precise time delay for each participant therefore depended on their reading speed during the second phase (reading 353 words on screen); pilot testing had indicated a delay of two to three minutes.
During the second phase, types of phobia (specific phobia, agoraphobia and social anxiety) were defined, with key diagnostic information given; concepts and information were either presented together (blocked) or mixed with information about different types of phobia on the same screen (interleaved). For example, in the blocked condition, a participant would view three items relating to agoraphobia, while in the interleaved condition they would view a key feature of each of the three types. Definitions were based upon criteria in DSM-V (American Psychiatric Association, 2013).
The online tasks also featured a test, comprising multiple choice questions about the research studies and the categorisation of novel examples of each type of phobia.
Participants sat at individual PCs, and a teacher oversaw the session. After a general briefing, each completed an on-screen consent form, followed by viewing the material presented in an order that depended on allocation to experimental conditions, which was decided via random numbers. As soon as participants had completed the task, the software automatically initiated the test.
Ethics approval followed the school’s framework; as a research-focused school, it had set up its own in-school ethics board, with an academic panel providing oversight.
The mean percentages of correct answers on the end-of-task test for the interleaved and blocked conditions are shown in Figure 4. A between-subjects ANOVA was carried out. This analysis revealed a significant main effect of spacing (performance in the spaced condition being worse than the massed condition, with mean scores of 12.25 vs 9.45, p = .002), while interleaving did not have a significant main effect.
Importantly, there was also a significant (p = .009) interaction between the two variables (spacing vs interleaving), indicating that interleaving had a mediating or protective effect against the difficulties caused by spacing (see Figure 5).
Figure 4: Table of mean scores
Figure 5: Graph showing interaction between interleaving and spacing
The findings demonstrated that spacing had a harmful effect on the immediate test, while the main effect of interleaving was neutral. The results fit with the idea that these are ‘desirable difficulties’, with the potential to impede learning in the short term. Soderstrom and Bjork (Soderstrom and Bjork, 2015) describe how such strategies often lead to performance being slower and more error-prone, but improve learning over longer intervals.
Nevertheless, increased errors within a short learning session could suggest inefficiency in the learning process, and raise questions about the use of spacing, in particular with new concept learning. In a related finding, Donovan and Radosevich (Donovan JJ and Radosevich, 1999) found that spacing was not beneficial for complex tasks; complexity interacts with learner experience and, when learning a new concept, complexity for a learner can be high. Such an explanation suggests that desirable difficulties interact with learner skill, as proposed by McDaniel and Butler (McDaniel and Butler, 2011).
One way to get around short-term difficulties is to utilise a hybrid schedule, with interleaved or spaced practice being utilised subsequent to an initial learning phase. This fits with the recommendation of Rawson and Dunlosky (Rawson and Dunlosky, 2011) that learners should first automise recall of concepts, and later retrieve items three times at widely spaced intervals. Interestingly, Yan et al. (Yan et al., 2017) found that a blocked-to-interleaved schedule was not superior to pure interleaving, which fits the current findings that interleaving alone did not cause short-term difficulties.
The findings of this experiment support the hypothesis that spacing and interleaving can interact within a classroom situation as, surprisingly, interleaving one phase of the task appeared to attenuate the short-term difficulties caused by spacing the other. On the face of it, this result does not fit well with Birnbaum et al.’s (Birnbaum et al., 2013) finding of a negative interaction between spacing and interleaving, but it is important to note that, in the present experiment, the two interventions were used in different task phases, and spacing therefore didn’t prevent discriminative contrast.
Why did adding two desirable difficulties make the task easier? It is conceivable that the difficulty of one task focused attention on the other – a variation of the attenuation of processing theory (see (Dellarosa and Bourne, 1985)), which suggests that spacing results in more attention being paid to a task. However, it is unclear why increased attention would not also impact on spacing in isolation.
Another possibility is that the learners benefited from forming conceptual links between the two task phases, with interleaving prompting learners to seek outside meaningful connections beyond what was on a single screen in a way that did not happen when the concept learning phase was presented in blocks. Such connections could be harder to recall or perceive if the first learning phase had been massed and appeared complete. This explanation could be examined in future by comparing the learning of sets of information with varying levels of conceptual similarity, on the basis that meaningful links could not be formed if items were unconnected (similarly to the findings of Hausman and Kornell (Hausman and Kornell, 2014), whereby material from different domains couldn’t be productively contrasted).
The present study was limited in terms of its scope and sample. A small number of pupils were tested and, as teenagers in the latter stages of education, their levels of motivation or general knowledge could differ from adults or younger learners, potentially affecting the findings. Any discussion of the effects of spacing and interleaving must take account of individual learner differences, and future studies should increase sample size and diversity.
As the learners in this study were new to the material being studied, a control group was not deemed necessary, given that prior concept knowledge was assumed to be absent. Future studies could confirm this assumption by presenting the test phase alone to a comparable group of pupils.
The present study focused on a relatively brief set of tasks within a single lesson, and future work could follow learners over a longer period to see how desirable difficulties play out across the learning of a topic. It is also essential to establish the extent to which the negative short-term classroom effects of interleaving and spacing would be counteracted by improved long-term ability to remember and transfer learning, and whether the interaction found in the present study would apply over longer timescales.
This classroom-based study found that while spacing caused short-term harm to learner performance, interleaving counteracted this problem. As such, it is suggested that combining interleaving and spacing may prompt learners to seek meaningful connections during concept learning.