HSW-0254 · How Studying Works
A learner studies a diagram and is told to track colour.
On the next trial, colour no longer matters. Orientation matters.
The learner still sees the right colours. The right orientations are also present. Yet when asked which feature belonged to which location, the answer becomes more vulnerable to a particular kind of mistake: the learner may report a real feature value that belonged to the wrong object.
That is not the same as memory simply becoming blurry.
Feature-switch binding cost is the loss that can occur when the cognitive system has to change which feature dimension is currently relevant and the previous priority is not fully cleared. The result can be a binding error: the remembered feature exists, but it is attached to the wrong location or object.
This is a narrow mechanism from visual short-term-memory research. It should not be inflated into the slogan “switching subjects is bad.” The useful educational lesson is more precise: when the rule for what must be attended to changes, the learner may need to reconfigure the selection system before the next representation is encoded accurately.
Auditory Memory Feature Asymmetry retains ownership of how retrospective attention can improve one remembered sound feature more than another. Spatiotemporal Working-Memory Addressing owns how temporary memories can be found through when-and-where cues. How Cognitive Flexibility Works owns the broader skill of changing rules or strategies. Here the question is narrower: what happens to temporary feature–location binding when the relevant dimension itself switches?
What the 2026 Experiments Add
Stuart B. Moore and James A. Grange studied this problem in The Impact of Dimension Switching on Visual Short-Term Memory, first published online on 28 November 2025 and appearing in the 2026 volume of Quarterly Journal of Experimental Psychology. See Moore and Grange.
Across change-detection and delayed-estimation experiments, participants had to remember either colour or orientation. Sometimes the relevant dimension repeated from the previous trial; sometimes it switched.
Performance was generally worse on switch trials. The delayed-estimation experiment was especially informative because the researchers could model different sources of error. The switch cost was not well explained by a broad loss of memory precision or by more random guessing. Instead, switching selectively increased non-target responses: participants were more likely to report a feature value belonging to another item.
The authors interpret this as evidence that dimension switching can disturb feature–location binding, perhaps because the previous dimension remains active or the new task set has not fully reconfigured. They are appropriately cautious: some parts of the proposed unified mechanism remain theoretical and need further modelling.
The Important Distinction: Fuzzy Memory vs Misbound Memory
Imagine four coloured arrows.
If memory becomes imprecise, you might remember the top-left arrow as slightly more blue than it really was.
If binding fails, you might report the exact red from the bottom-right arrow as though it belonged in the top-left location.
Those are different errors.
The second error is especially dangerous in learning because every ingredient can look familiar and correct. The failure is in the relationship.
Why the Previous Rule Can Leak Into the Next One
A task set tells the mind what matters now.
When the rule is stable, the system can keep prioritising the same feature. When the rule changes, attention and memory must be reconfigured. The previous configuration does not necessarily vanish instantaneously.
In the Moore and Grange paradigm, the irrelevant feature also changed. That made lingering attention to the previously relevant dimension consequential. A stale priority could compete with the new one.
For education, the analogous question is not “Did the learner switch subjects?” It is “Did the learner switch the criterion that determines which information must be encoded and related?”
A Classroom Example: Same Diagram, New Question
Consider a graph used for three consecutive questions.
- Question 1 asks for the highest value.
- Question 2 asks for the fastest rate of change.
- Question 3 asks for an anomalous point.
The visual object has not changed. The relevant feature has.
A student who remains configured for “find the largest number” can carry the wrong priority into “find the steepest interval.”
This is not evidence that the exact laboratory binding mechanism caused the classroom error. It is an illustrative application: when task relevance changes, stale priorities can plausibly create relational mistakes even when the underlying information is visible.
Mathematics: The Variable Stayed the Same; Its Role Changed
Students often move between questions that reuse the same symbols but change their roles.
In one question, x is the unknown to solve. In another, x is an input varied across a function. In another, it is a coordinate used to locate a point.
The symbol remains familiar. The relevant relation changes.
A useful reset is to require one sentence before calculation:
“In this question, x is ________, and I need to use it to ________.”
That small act makes the new task set explicit before the old one drives the next step.
English: The Same Sentence Can Be Read for Different Features
An English learner may first analyse a sentence for grammar, then for tone, then for evidence of character.
Each question changes what matters.
If the learner stays configured for grammar, they may notice verb agreement perfectly while missing the evaluative adjective that carries tone. If they stay configured for tone, they may quote a vivid phrase that does not answer a reference question.
The skill is not generic “attention.” It is controlled reprioritisation.
Science: Correct Feature, Wrong Object
Binding errors are especially easy to see in diagrams with repeated structures.
A learner knows that oxygenated blood has a particular property and knows the correct blood vessel names, yet attaches the property to the wrong vessel. Or the learner remembers which organ performs a function but places the function on the wrong labelled structure.
More exposure to the individual facts may not repair the relation. The learner needs feature-to-object discrimination.
Why “Pay More Attention” Is Too Vague
A learner can pay a lot of attention to the wrong dimension.
Attention must be configured by the task:
- Which feature matters now?
- Which feature mattered before but is now irrelevant?
- What object or location must the feature remain bound to?
- What competing feature could intrude?
This turns “focus” into an operational instruction.
The Relevance Reset
Before a task switch, use a three-part reset:
- Old rule: What was I tracking?
- New rule: What matters now?
- Binding target: What must this feature stay attached to?
This can take five seconds. Its purpose is not ritual. It is to make reconfiguration explicit when the task structure makes carryover costly.
The Wrong-Object Diagnostic
When an answer is wrong, classify it.
- Missing feature: learner did not remember the value.
- Low precision: remembered value is approximate or distorted.
- Binding error: remembered a real value but assigned it to the wrong object, place or role.
- Rule-selection error: attended to a dimension that was not currently relevant.
These errors need different repairs.
If the learner knows all the individual values but swaps them, more fact drilling may be low-value. Practise relation and location.
The Contrast-Pair Drill
Build two neighbouring tasks that use the same material but change the relevant feature.
For example:
- same graph: identify magnitude vs identify rate;
- same paragraph: locate evidence vs infer motive;
- same diagram: identify structure vs explain function;
- same equation: solve for a variable vs interpret a parameter.
Before each item, ask the learner to name the currently relevant feature. After the answer, ask what would have been tempting if the previous rule had remained active.
The aim is controlled switching, not avoidance of switching.
Feature-Switch Binding Cost vs Auditory Feature Asymmetry
Auditory Memory Feature Asymmetry asks what happens when attention returns to an already remembered sound and some features benefit more than others.
Feature-Switch Binding Cost concerns a different moment: the relevant feature dimension changes across trials, and the switch can increase confusion about which feature belongs to which location.
One is about selective enhancement within a remembered object. The other is about reconfiguration and binding across changing task demands.
Feature-Switch Binding Cost vs Cognitive Flexibility
Cognitive Flexibility is a broad competence: change strategy, representation or rule while preserving the goal.
The present article does not claim to measure that broad competence. It isolates one cost in a specific laboratory setting: memory can become more vulnerable to non-target binding when the relevant visual dimension switches.
Feature-Switch Binding Cost vs Spatiotemporal Addressing
Spatiotemporal Working-Memory Addressing explains how temporary memories can be organised and retrieved by spatial and temporal coordinates.
The switch-cost evidence adds a complementary warning: a location cue is only useful if the correct feature remained bound to that location during encoding.
A Better Way to Design Mixed Practice
Mixed practice is valuable when learners must choose among methods or question types. But mixing should train discrimination, not merely create turbulence.
A high-quality mixed set makes the switch criterion learnable:
- the learner names what changed;
- the learner identifies the cue that makes the new rule relevant;
- the learner explains why the old rule no longer applies;
- the learner keeps the new information attached to the correct object or role;
- later items remove the explicit prompt.
This converts switching from random alternation into training of controlled reconfiguration.
The Delayed Switch Test
Immediate success after a cue may reflect the cue itself.
After a delay, give a small mixed set with no “this is a rate question” or “now analyse tone” labels. Ask the learner to:
- identify the relevant feature;
- state the rule;
- answer;
- explain one tempting carryover from the previous item.
If performance collapses only when labels disappear, the learner may have learned to obey the switch cue without learning to detect the need for a switch.
Parent and Tutor Section: When the Child “Knows It” but Keeps Swapping Things
Repeated swaps can look careless.
Before giving that label, test whether the learner can:
- state each item correctly in isolation;
- state which relation or location belongs to each item;
- detect when the task changes what feature matters;
- clear the old rule before starting the next item;
- repeat the performance after a delay.
If the pieces are known but repeatedly misassigned, the repair is not “memorise harder.” It is to train the relation and the switch that selects it.
What the Evidence Does Not Justify
- It does not prove that switching school subjects causes the same binding mechanism.
- It does not prove multitasking is the only or even primary source of classroom binding errors.
- It does not show that every switch damages memory precision.
- It does not show that colour/orientation tasks are equivalent to mathematics, reading or science.
- It does not imply learners should practise only one feature at a time forever.
- It does not establish a diagnosis for an individual learner from ordinary swap errors.
Evidence Boundary
The Moore and Grange experiments directly support a switch cost in controlled visual short-term-memory tasks. In the delayed-estimation experiment, modelling pointed to a selective increase in non-target responses rather than a general loss of precision or greater guessing. The authors propose impaired feature–location binding as a mechanism and note that some aspects of the broader account remain to be tested.
The classroom protocols above are educational applications, not direct effects demonstrated by the experiment. Their purpose is to make relevance shifts explicit, diagnose relationship errors and verify independent control after prompts are removed.
A Six-Minute Repair Protocol
- Choose two tasks using the same material but different relevant features.
- Have the learner state the rule before Task A.
- Switch to Task B and require an explicit “old rule / new rule” statement.
- Ask which object or location the new feature belongs to.
- Repeat with a new pair.
- Finish with an unlabelled mixed item and check whether the learner detects the switch alone.
If this improves performance, the evidence supports a task-control problem worth training. It still does not prove the laboratory mechanism caused the original error.
Return: Sometimes the Memory Is Sharp and the Relationship Is Wrong
That distinction changes teaching.
If memory were simply fuzzy, the answer might be more exposure or stronger encoding.
If the feature is remembered but attached to the wrong place, the learner needs binding and control.
Before the next task begins, name what matters now. Clear what mattered before. Keep the feature attached to the correct object. Then practise until the learner can reconfigure without being told.
Continue through Auditory Memory Feature Asymmetry, Spatiotemporal Working-Memory Addressing, How Cognitive Flexibility Works, the How Studying Works Numbered Series Reading Index, and the How X Works Hub.