The Context-Dependence Problem in English Exams

Series ID: SETC-EN-FAIL-0310 · How My Students Fail their English Examinations Vol No.031

Some English skills work beautifully in the place where they were learned.

The same tutor asks the question. The same worksheet layout appears. The same topic family returns. The same paragraph structure is used. The same classroom language signals which method applies.

The student succeeds.

Then one contextual feature changes.

The passage is about an unfamiliar subject. The question is phrased differently. The paper looks different. The teacher is absent. The task appears in another component. The cue that normally announces the strategy is gone.

The skill seems to disappear.

This is the context-dependence problem.

A skill is not fully transferable if it depends on contextual cues the real examination is free to change.

Transfer research has long treated successful learning as more than performance in the original training situation. Research in English-for-academic-purposes contexts has documented transfer across varying distances, while newer 2026 work on variability shows that varied items can support generalisation when learners need to induce underlying rules rather than memorise surface patterns. The exact evidence comes from different domains, so we should be cautious about universal prescriptions. The useful principle is robust: variation can reveal what the learner has actually abstracted from the examples.

Maren Can Write About Technology but Not Ageing

Maren has practised argumentative writing through familiar technology topics.

She has examples, phrases and ready-made arguments.

Her essays improve.

Then she receives a question about ageing populations.

The planning system slows dramatically.

The writing structure may be stable.

The topic familiarity was doing more of the work than the score revealed.

Iona Knows the Question Type When the Heading Tells Her

Iona practises inference in a worksheet section labelled “Inference Questions”.

She performs well.

On a mixed paper, the label disappears.

She now has to recognise the reasoning demand from the question itself.

Part of the old success belonged to the heading.

Leonie Performs Better With Her Tutor Than Alone

Leonie’s tutor does not give obvious hints.

But the tutor’s presence still changes behaviour.

A raised eyebrow makes Leonie reconsider. Silence makes her elaborate. A familiar sequence of follow-up questions tells her what kind of answer is expected.

These are contextual cues.

The examination will not reproduce them.

Context Is Larger Than the Room

When we say “context”, we do not mean only physical location.

English learning can become tied to:

topic;
teacher;
worksheet layout;
question wording;
example style;
genre;
difficulty level;
sequence of tasks;
timing condition;
available tools;
peer discussion;
feedback timing;
emotional state;
physical environment.

The more context features remain fixed during training, the harder it is to know whether the learner acquired the underlying skill or a context-bound routine.

Context Failure No.1: Topic Familiarity Masquerades as Skill

Students perform better when they know the domain.

That is not a flaw. Background knowledge legitimately supports comprehension and writing.

The diagnostic mistake is assuming the skill itself is stable because it worked on a familiar topic.

Test the same process on unfamiliar but accessible content.

Context Failure No.2: The Worksheet Announces the Strategy

Blocked practice is efficient for initial learning.

If every exercise in the section uses the same method, the student does not need to select the strategy.

Later, mixed practice should remove the announcement.

The examination rarely writes, “Use your inference strategy now.”

Context Failure No.3: One Teacher’s Phrasing Becomes the Cue

A teacher may repeatedly say, “What exactly is the writer showing?”

The student learns the reasoning routine attached to that phrase.

Another paper asks, “What can you infer?”

The relationship is similar but the cue differs.

Vary wording while preserving the underlying operation.

Context Failure No.4: Model Formats Become Templates

A student learns from three model essays with the same paragraph architecture.

The structure becomes familiar.

Then a prompt demands a different balance of argument and example.

The student forces the old format onto the new task.

Context dependence can look like formula dependence.

Context Failure No.5: Support Is Embedded in the Routine

Prompting does not have to be explicit to be powerful.

The teacher always asks one follow-up. The worksheet always gives three lines. The model answer always appears immediately after the attempt.

When those features disappear, the student can feel as if the skill disappeared too.

Context Failure No.6: Practice Uses One Genre at a Time

A reasoning skill may appear stable in one genre and fail in another.

Cause-and-effect reasoning in an explanatory article may not automatically transfer to an argumentative essay or oral response.

Once the skill is learned, cross the genre boundary deliberately.

Context Failure No.7: Practice Uses One Difficulty Band

Students can become adapted to a narrow difficulty range.

When the examination gives one easy passage and one much harder passage, strategy flexibility becomes necessary.

Vary difficulty after the core method is stable.

Context Failure No.8: Practice Is Always Untimed

A skill learned only with unlimited time may be context-bound to low-pressure conditions.

That does not mean everything should be timed from the start.

It means timing must eventually become one of the varied contexts in which the process is tested.

Context Failure No.9: Practice Is Always Timed

The opposite can also limit learning.

If every attempt is timed, the learner may never slow down enough to understand why the method works.

Use both learning context and performance context.

Context Failure No.10: The Student Practises With the Answer Nearby

Notes, checklists and models can become contextual cues.

The student feels independent because no person is helping.

But the environment is still doing part of the retrieval.

Close the materials and retest after delay.

Context Failure No.11: Corrections Are Retested in the Same Surface Form

The student fixes a sentence, then sees the same sentence again.

Success may be tied to memory of the item.

Retest the same principle with different wording, topic or sentence structure.

Context Failure No.12: The Student Never Meets an Unfamiliar Examiner or Audience

Oral performance can become tied to a familiar listener.

The learner knows the tutor’s timing, facial expressions and follow-up style.

Occasionally change the listener, recording format or question sequence.

Context Failure No.13: The Student Learns Examples but Not the Rule

Three examples can create familiarity without abstraction.

Ask:

What stays the same across these examples even though the surface changes?

That invariant is what must transfer.

Context Failure No.14: Variation Arrives Before the Learner Has Anything Stable to Vary

Variation is not always beneficial immediately.

If the learner has not yet understood the core process, constant surface changes can create overload.

Stabilise the basic relationship first, then vary the context to test abstraction.

This connects to Vol No.030 — The Challenge-Mismatch Problem.

The Context-Stripping Test

Take a skill that appears stable and remove one contextual support at a time.

remove the label;
change the topic;
change the wording;
change the layout;
remove the checklist;
change the teacher or listener;
delay the task;
add realistic time.

The point at which performance falls reveals which context feature is still carrying part of the skill.

The Variability Ladder

same process, similar examples;
same process, different surface details;
same process, different topic;
same process, different genre;
mixed processes requiring selection;
unseen task under realistic conditions.

Newer 2026 evidence on variability and transfer suggests that varied items can improve generalisation by helping learners induce the underlying rule rather than bind it to one example set.

The Stable-Invariant Rule

When changing context, keep one thing stable enough to learn from the variation.

For comprehension: the evidence relationship.

For writing: the paragraph job.

For grammar: the underlying condition.

For oral: the development process.

Variation teaches what changes and what does not.

The Context-Dependence Audit

Where was the skill learned?
Which topics dominate practice?
Which teacher cues are always present?
Is the task usually labelled?
Is the layout always familiar?
Are notes or models usually visible?
Has the skill been tested in another genre?
Has it been tested after delay?
Has it been tested with an unfamiliar surface?
Which contextual change causes the largest drop?

Twenty Signs English Is Context-Bound

Performance is strong only with familiar topics.
Question labels are needed.
Teacher phrasing cues the strategy.
Worksheet layout affects accuracy.
Model formats are copied rigidly.
One tutor produces much better performance than another.
Oral fluency falls with unfamiliar listeners.
Skills work in one genre only.
Corrections succeed only on repeated items.
Untimed work is much stronger than timed work.
Or timed drilling is stronger than thoughtful untimed work.
Notes need to remain visible.
Prompt dependence rises when wording changes.
Background topic change causes large performance loss.
Examples are remembered more easily than principles.
Mixed practice is much weaker than blocked practice.
The student says “I know how when it looks like the worksheet”.
Support is hidden inside familiar routines.
Transfer drops as contextual distance increases.
The skill belongs partly to the environment rather than fully to the learner.

Twenty Signs Context Independence Is Growing

Topic changes produce smaller score drops.
Question wording changes are manageable.
Labels can disappear.
Mixed tasks improve.
Different teachers get similar performance.
Unfamiliar listeners are less disruptive.
Skills cross genre boundaries.
Corrections transfer to changed examples.
Notes and checklists can fade.
Delayed performance remains stable.
Timed and untimed performance converge appropriately.
New layouts do not derail strategy selection.
The learner can explain the invariant across examples.
Models are adapted rather than copied.
Background knowledge still matters but no longer carries the whole task.
Support is recognised and deliberately removed.
Variability becomes informative rather than confusing.
Unseen tasks feel new but not structureless.
The student can retrieve the method from task features rather than classroom cues.
The skill travels.

The Parent Rule: Ask Whether the Skill Works Somewhere Else

If a child is doing well in tuition, that is useful evidence.

Then ask whether the same process works on school material, unseen material and independent work.

The goal is not to distrust tuition success.

It is to prove transfer.

The Tutor Rule: Vary Surface, Preserve Principle

Once a process is understood, change the topic, wording, genre, sequence or support condition.

Keep asking what stays invariant.

The Student Rule: Do Not Learn the Worksheet

Learn the relationship the worksheet was trying to show you.

Then prove you can recognise it when the surface changes.

What Happened to Maren, Iona and Leonie

Maren wrote the same argumentative structure across technology, ageing, sport and public-space topics. Her idea generation slowed at first, then the writing architecture became less topic-bound.

Iona moved from labelled question blocks to mixed unseen sets. Accuracy fell temporarily, revealing that question recognition had not been independent. It then recovered.

Leonie recorded oral responses for unfamiliar listeners and changed question order. The invisible cues from her tutor became less necessary.

They did not train for every possible context.

They trained the underlying process across enough variation that the context stopped owning the skill.

Where This Volume Fits in SETC

This volume owns the broad failure landscape where English performance depends too heavily on familiar topics, prompts, teachers, layouts, genres or environmental cues and therefore weakens when the context changes.

For false progress, use Vol No.002. For preparation fidelity, use Vol No.008. For language transfer generally, use SETC’s How English Learning Transfers.

The Context Rule

Once a skill works in training, change the surface until you know the learner has acquired the principle rather than merely learned the environment.

The final volume in this batch goes deeper: sometimes the context is not the problem at all. The advanced task keeps collapsing because a lower prerequisite in the stack never became secure.

Research and Further Reading

Series: Vol No.030 — Challenge-Mismatch Problem · Vol No.029 — Tool-Dependence Gap

Continue with English examination failure guides or browse The Library.

How My Students Fail their English Examinations Vol No.031 | The Context-Dependence Problem — When English Works Only With Familiar Topics, Teachers, Formats or Cues