Another Forgotten Desirable Difficulty: Contextual Interference

As I wrote in a previous piece, Robert and Elizabeth Bjork once identified five desirable difficulties. Three of them have since become educational darlings: spaced practice, interleaving, and retrieval practice. They appear in books, professional development sessions, and policy documents. The remaining two have been far less fortunate. Contextual interference and reduced feedback rarely receive the same attention. Perhaps because they are harder to explain. Perhaps because they run counter to what feels like good teaching (whatever that is). My last piece was about reduced feedback. This one is about the other: contextual interference, a desirable difficulty that is quietly powerful and consistently overlooked.

What’s contextual interference?

Contextual interference refers to varying the context in which learning and practice take place. Crucially, this is not about changing the task itself (i.e., interleaving) but about changing the conditions surrounding the task. The content stays the same, but the setting, format, presentation, or circumstances shift. At first glance, this can seem trivial. Why should it matter whether students learn a concept in classroom A or B, from a textbook or from a situation outside of the classroom, on paper or on screen, in silence or with background noise? Yet decades of research suggest that this kind of variation makes the difference between knowledge that survives beyond the lesson and knowledge that only works, at best, in the class.

In classrooms with low contextual interference, we do our darndest to keep conditions stable. The same room, the same book, the same layout, the same types of exercises. This feels sensible because it reduces cognitive load, increases the likelihood of success, and creates a calm, orderly learning environment. And students make fewer errors, work more fluently, and feel more confident. For both teachers and students, this looks and feels like effective teaching and learning.

The problem is that this kind of learning is often fragile. What students ‘learn’ under highly stable conditions tends to become tied to those conditions. The knowledge is there, but it’s context-dependent. Change the format of the test, rephrase the question, alter the visual layout, or embed the problem in a slightly different situation, and suddenly the knowledge becomes inaccessible. Not because it was never learned, but because it was never disentangled from the context in which it was practised.

This is where contextual interference does its work. By varying the context during practice, students must rely less on environmental cues and more on memory. They can no longer think, “This is that type of problem from yesterday” or “This looks like that diagram in the textbook.” Instead, they have to ask themselves what actually stays the same when the surface features change. And that question is at the heart of transfer.

A particularly instructive educational example of contextual interference is what researchers sometimes call SSDD: same surface, different depth. In SSDD tasks, problems look superficially similar but require different underlying principles or solution strategies (see the figures above). The surface cues invite a familiar response, but that response is wrong unless the learner attends to the deeper structure. From the learner’s perspective, this can feel really uncomfortable. The usual contextual shortcuts don’t work. From a learning perspective, however, SSDD is powerful precisely because it disrupts ‘normal’ context-bound responding. Students are forced to discriminate, not to ask or think “What does this look like?” but “What kind of problem is this really?” In that sense, SSDD is contextual interference in cognitive clothing: the surface stays constant while the meaning shifts, forcing retrieval of underlying principles rather than reliance on surface appearance.

A textbook case from somewhere completely different

The Bjorks’ explanation fits neatly with what we know about memory. When context is stable, it acts as an additional retrieval cue. That makes recall easier, but also more superficial. When the context changes (e.g., physically, perceptually, conceptually) those cues disappear. Retrieval becomes more (desirably) difficult, less fluent, and more effortful, but also more productive. Students must reconstruct the knowledge itself rather than lean on the situation where and when it was learned. That extra effort leads to learning that’s more flexible and durable.

This idea didn’t originate in classrooms. Contextual interference actually has its beginnings in sport and motor learning, where its effects are striking and well documented. In classic experiments, athletes practised multiple motor skills either in blocked, predictable sequences or in varied and contextually changing conditions. The blocked practice groups, as expected, performed better during training. Their movements were smoother, their execution more accurate, and the result more consistent. Coaches would have preferred them. But when tested later in game-like or transfer situations, the advantage disappeared and blocked practice actually performed worse. Athletes who had trained under high contextual interference, despite poorer practice performance, showed superior retention and adaptability when it was needed. They could execute the skill under pressure, in new situations, and when conditions changed, that is, in a real game. The lesson was clear long before education caught on: practice that looks and feels good isn’t the same as practice that works.

An example from basketball

In a low contextual interference basketball practice, shooting or defensive skills are often trained in isolation: repeated shots from the same spot, the same defender defending in the same way, or the same pick-and-roll executed from the same side of the court. Players quickly improve during practice. Shots fall. Defensive rotations look crisp. Coaches feel progress is being made.

A high contextual interference approach keeps the skill constant—for example, shooting off the dribble or defensive decision-making—but varies the context. Shooters take shots from different locations, after different movements, with different defenders and time constraints. Defensive drills vary the offensive spacing, the angle of attack, or the number of attackers. Sometimes the action begins from transition from defence to offence after a basket, sometimes from a set play, sometimes after a steal or broken possession. The skill doesn’t change, but the cues do. Players can’t rely on the drill to tell them what’s coming next.

And the result? Accuracy decreases, reactions are slower, and errors increase. But players become better at executing skills under pressure and in unfamiliar situations. When the game environment shifts the skills hold up. Players trained with contextual interference are less dependent on rehearsed patterns and more adaptable when space, pressure, and tempo change in real games.

This makes contextual interference a textbook example of a desirable difficulty. During practice, performance suffers. Students hesitate more, make more mistakes, and feel less confident. Over time, however, learning improves: retention increases and students are better able to apply what they know in new situations. As with other desirable difficulties, this is precisely why contextual interference is so rarely used. We judge teaching largely by what we see during the lesson. And contextual interference doesn’t look good.

Assessment

The relevance of contextual interference becomes especially clear when we look at assessment. Many so-called assessment failures aren’t learning failures as such, but transfer failures. Students are tested in contexts different from those in which they practiced: different wording, different layouts, different cues, different demands. From the teacher’s perspective, this feels unfair: “We covered this”. From the student’s perspective, it feels inexplicable: “But I knew this yesterday”. In reality, the assessment has stripped away the contextual supports that made performance possible during learning. What remains is the knowledge itself, and too often that knowledge has never been required to stand on its own. This is known as transfer appropriate processing (TAP): the idea that learning is most successful when the cognitive processes used during learning match the processes required at retrieval or transfer.

This also explains why teaching to the test can appear effective in the short term and yet fail so completely in the long term. When practice conditions closely mirror test conditions, students can rely on surface similarities and contextual cues to guide their responses. Performance improves, scores rise, and everyone is reassured. But this isn’t transfer; it’s alignment. The moment students are asked to apply the same knowledge in a different form, for example on a cumulative exam, in a subsequent course, or in a real-world situation, the apparent mastery collapses. Contextual interference deliberately prevents this illusion (maybe I should write a second book on instructional illusions 😊) by ensuring that knowledge is practised across varying conditions from the outset. It makes assessments feel harder and the results may initially be worse, but it also makes success on unfamiliar tasks far more likely.

What does contextual interference look like in a real lesson?

Across subjects, it means keeping the underlying idea constant while deliberately varying everything else. In mathematics, the same structure appears in different representations and wordings so students must decide what kind of problem this is rather than execute a rehearsed procedure. In languages, grammatical forms and vocabulary are practised across changing communicative situations, forcing retrieval of the rule instead of reliance on the exercise format. In science, core principles reappear in different phenomena, diagrams, and surface stories, requiring students to abstract the law from the example. In history and social studies, the same analytical concepts are applied to diverse cases and sources, preventing students from memorising context-bound answers. In all cases, practice becomes slower and messier, but knowledge becomes less tied to a particular worksheet, format, or moment. Students learn not what this looks like here, but what it is everywhere.

Here are two examples of using contextual interference, one in foreign language learning and one in geography.

In the first, we see contextual interference in verb conjugation where the deep structure is held constant, namely the correct conjugation of a specific tense (e.g. past tense, first person singular) but the context is varied, namely mode of use and communicative demand.

Here, the grammar rule is identical in all four quadrants. What changes is the retrieval context: written vs spoken, low vs high pressure. Learners can’t rely on the same cues (sentence frame, time to think, visual support). Retrieval becomes harder and less fluent, but the conjugation becomes usable beyond drills; that is in real communication.

In this second, we see contextual interference in geography (mountains and rivers), Here, the deep structure again is held constant, namely understanding and identifying mountain ranges and rivers (location, function, relationships). However, the context is varied, namely the representation format and task demand.

Here, students must retrieve the same geographic knowledge in all four quadrants. What changes is how that knowledge is cued: maps vs text, recognition vs explanation. Students can’t rely on “this looks like a map question” or “this is a labelling task.” They must retrieve the underlying spatial relationships and concepts regardless of representation.

A few closing words

First, it’s important not to confuse contextual interference with interleaving; while similar they aren’t the same. Interleaving varies the tasks or problem types; contextual interference varies the conditions under which the same task is practised. Note that while you can have one without the other, you also can combine them.

Why, then, do we avoid contextual interference? Stability feels safe. For students, it provides reassurance. For teachers, it offers control. For schools, it enables predictability. Variation, by contrast, can feel like noise. It’s something to eliminate rather than embrace. Yet that “noise” is exactly what students need to learn what really matters.

Second, as with all desirable difficulties contextual interference is only desirable when students have sufficient foundational knowledge. For true novices, too much variation is often overwhelming. If everything changes at once (e.g., task, context, format) there’s nothing stable to build on. When this happens, the difficulty becomes destructive rather than productive. This means that contextual stability is often necessary early on, but should never be the end point. As students become more competent, the context must become looser, more variable, and less predictable.

A useful test is to observe what students do when the context varies. If they slow down, compare, explain, and reflect, you’re likely on the right track. If they guess randomly or disengage, you might have gone too far. Contextual interference isn’t instructional chaos; it’s deliberate variation designed to detach knowledge from where and when and how it was acquired.

It also helps explain a familiar classroom complaint: “But I understood this yesterday.” Yesterday, what the student had wasn’t robust knowledge, but context-supported performance. Today, the context has changed and the support has vanished. That’s not a student failure; it’s a sign that learning was bound to the situation in which it occurred.

Yes, teaching with contextual interference feels less tidy. It produces more errors. It undermines the comforting sense of control. But if the goal of teaching and education is for students to use what they learn beyond the classroom, that discomfort is the price we must pay. Contextual interference forces learners to determine what remains when everything else changes. And that may be one of the most important things education can do.