How Studying Works | Schema Congruency — Why Expected and Surprising Information Can Be Remembered Differently Depending on the Test

HSW-0188 · How Studying Works

You walk into a kitchen and see a kettle.

You barely notice it.

You walk into the same kitchen and see a violin floating above the sink.

You notice immediately.

Which one will you remember better tomorrow?

The intuitive answer is “the surprising one.”

Memory research gives a more interesting answer:

it depends on what kind of memory you test.

Schema congruency describes how well new information fits an existing structured expectation or knowledge framework, and memory for congruent versus incongruent information can differ depending on whether the test asks for item recognition, perceptual detail, source, context or another feature of the episode.

Two 2026 Memory & Cognition studies sharpen this point. Suarez, Lara and Ciria found that contextually congruent object–scene combinations supported better item recognition and processing efficiency, but did not produce the same advantage for fine-grained perceptual detail. A later 2026 paper by Yacoby and colleagues found that item recognition tended to favour expected information while source memory showed a different pattern, including stronger memory at some ends of the expectedness–surprise continuum. See Suarez, Lara and Ciria, 2026 and Yacoby and colleagues, 2026.

The lesson is not “make everything surprising.” It is that prior knowledge changes encoding and retrieval, and different tests reveal different benefits.

This article owns the narrow study problem of how expectedness and schema congruency change different forms of memory depending on the retrieval demand. It does not replace HSW-0176, The Prior Knowledge Paradox, which owns the broader effects of existing knowledge on new learning, or HSW-0140, The Distinctiveness Effect, which owns memory advantages created by standing out relative to context.

The 50-Second Read

• Expected information is not automatically forgettable. Schema-consistent material can be processed efficiently and remembered well at the item level.
• Surprise is not automatically superior. Incongruent information can capture attention or preserve contextual detail, but effects depend on the test and degree of surprise.
• Memory is multidimensional. Remembering the item, its exact appearance and where it came from are different outcomes.
• 2026 studies show dissociation. Congruency can help recognition without equally helping detail memory, while source-memory patterns can differ from item-recognition patterns.
• Prior knowledge is both compressor and filter. It helps expected information fit quickly but can cause learners to miss what is distinctive about a new exception.
• Good studying alternates fit and mismatch. First connect new knowledge to a stable schema; then deliberately contrast exceptions and boundary cases.
• The practical rule: connect → predict → expose mismatch → retrieve item → retrieve source/context → explain why the mismatch matters.

1. A Schema Is an Expectation Structure

A schema is not simply a list of facts.

It is an organised representation of what usually belongs together and how parts relate.

Your “restaurant” schema may include:

• tables;
• menus;
• ordering;
• food;
• payment;
• staff and customers.

When new information fits the structure, comprehension can be fast because many relations are already available.

When information violates the structure, the mismatch can attract processing because the current model predicts poorly.

2. Congruent Information Can Be Easy to Encode

Expected information arrives with somewhere to go.

A kettle in a kitchen is supported by many existing associations.

The learner does not need to build every relation from scratch.

That efficiency can support item recognition because the new item is encoded inside a familiar relational structure.

This connects to—but is narrower than—The Prior Knowledge Paradox: prior knowledge can lower the cost of understanding when the new information fits the existing model.

3. Incongruent Information Can Demand More Processing

A violin above the sink does not fit the kitchen schema.

The system has more work to do.

• Was the object correctly perceived?
• Why is it there?
• Does the context need updating?
• Is the event exceptional?

That extra processing can strengthen some aspects of memory.

But it can also make integration harder.

Incongruency therefore does not have one inevitable effect.

4. What Suarez, Lara and Ciria Found in 2026

The 2026 Memory & Cognition study by Suarez, Lara and Ciria examined memory for objects presented in congruent or incongruent scene contexts.

Congruent objects showed an advantage in recognition accuracy and processing efficiency.

However, that congruency advantage did not extend in the same way to retrieval of fine-grained perceptual details.

This is the key dissociation:

A schema can help you remember that the item was there without equally strengthening every detail of how it appeared.

5. What Yacoby and Colleagues Added in 2026

Yacoby and colleagues published a later 2026 Memory & Cognition paper examining expectedness and surprise across four experiments involving 138 university students.

The pattern depended on the memory test.

Item recognition generally favoured expected information.

Source memory behaved differently, with evidence for a non-linear pattern across expectedness and surprise rather than a simple “more surprising = better memory” rule.

The authors discuss retrieval strategy as an important part of interpreting these effects.

The educational lesson is strong precisely because it is conditional: what looks memorable depends partly on what you later ask memory to recover.

6. Item Memory and Source Memory Are Different Jobs

Suppose a student remembers the statement:

“Increasing pressure can affect equilibrium position in gaseous systems.”

That is item memory.

Now ask:

“Was that principle stated in the teacher explanation, the textbook example, your own note, or an AI-generated summary?”

That is source memory.

A learning experience can strengthen one more than the other.

7. Detail Memory Is Another Job Again

You may remember seeing a graph without remembering whether the y-axis started at zero.

You may remember a diagram without remembering which arrow was dashed.

You may remember the example without remembering the boundary condition that made it valid.

Recognition of the object or concept can survive while fine-grained detail weakens.

Study design must therefore test the level of detail that future performance requires.

8. Expected Information Can Become Compressed

When something fits a schema, the mind may not need to encode every feature independently.

The schema supplies structure.

This compression is efficient.

But compression has a price: unusual detail can be normalised toward expectation.

In studying, this means a familiar topic can feel easy while the exact exception is missed.

9. Surprise Can Protect the Mismatch

When an event violates expectation, the discrepancy itself can become part of the memory.

That can help preserve:

• where the information appeared;
• what made it unusual;
• which prior expectation failed;
• the context in which the exception occurred.

But surprise can also become decoration.

If the learner remembers only the dramatic image and not the principle it was meant to illustrate, attention has been captured without the target being secured.

10. Schema Congruency vs Distinctiveness

HSW-0140, The Distinctiveness Effect, owns the general problem of an item standing out relative to its neighbours.

Schema incongruency is more specific.

The mismatch is defined relative to an existing expectation structure, not merely perceptual oddness.

A bright red word among black words is distinctive without necessarily violating a knowledge schema.

A result that contradicts a well-understood scientific expectation is schema-incongruent even if it looks visually ordinary.

11. Schema Congruency vs Context Reinstatement

HSW-0147, Context Reinstatement, owns the retrieval benefit that can come from restoring relevant context.

Schema congruency concerns what happened during encoding and how the item fitted the context and expectations then.

The two interact, but they are not the same question.

12. Schema Congruency vs Prior Knowledge

The Prior Knowledge Paradox explains why existing knowledge can help or hinder new learning through many mechanisms.

Schema congruency zooms in on one of those mechanisms:

does the new information fit what the learner already expects, and how does that relationship affect different memory outcomes?

13. Schema Congruency vs Self-Derived Knowledge

HSW-0173, Self-Derived Knowledge, owns how separate memories can be integrated to generate a new inference.

Schema congruency changes the conditions under which new information is encoded relative to what is already known.

A surprising fact can force schema revision; an expected fact can strengthen an existing model.

Both can later participate in self-derivation.

14. Mathematics: The Dangerous Familiar Problem

A student sees a problem that looks like a standard quadratic.

The familiar schema activates quickly:

expand → rearrange → solve.

But one condition changes the problem: perhaps the domain is restricted, a parameter alters the number of roots, or the target is an inequality rather than an equation.

The expected structure speeds entry and creates risk.

Train the student to ask:

What is familiar here, and what is the one feature that could invalidate the familiar method?

15. English: Genre Schemas Help Until the Writer Breaks Them

Readers use schemas for genre.

A news report, argument, story and advertisement create different expectations.

Those expectations improve comprehension because they help predict structure.

They also make deliberate violations meaningful.

An author who withholds the expected resolution or changes viewpoint can become memorable because the text violates the active model.

Strong reading asks both:

• What does the genre make me expect?
• Where does the text refuse that expectation?

16. Science: Anomalies Matter Only Against a Model

A surprising result is surprising because something else was expected.

That means scientific anomalies require a prior model.

Teach students to state the expected result before revealing the anomaly.

Then ask:

• Was the prediction wrong?
• Was the measurement unreliable?
• Was a hidden variable present?
• Does the model need a boundary condition?
• Is there an alternative mechanism?

The mismatch becomes a reasoning event, not merely a memorable trick.

17. The Prediction-First Method

If you want a mismatch to teach, expose the expectation first.

1. Present enough information to form a prediction.
2. Ask the learner to commit to the prediction.
3. Reveal the actual case.
4. Compare expectation with outcome.
5. Explain the mechanism behind the match or mismatch.
6. Retrieve the explanation later.

Without the prediction, “surprise” may be merely novelty.

18. The Exception-Tagging Method

When a fact violates a well-established rule, do not store it as an isolated curiosity.

Tag it to the rule.

Use the structure:

default → exception → condition that creates exception → evidence.

This preserves both schema efficiency and boundary knowledge.

19. The Source-Memory Audit

After learning several claims, test more than content.

• What was the claim?
• Where did it come from?
• Was it evidence, example, speculation or explanation?
• Which part was surprising?
• Why was it surprising relative to the prior model?

This connects to HSW-0134, Source Monitoring, without replacing it. Here source is tested because schema congruency can produce different outcomes for item and contextual memory.

20. The “Make Everything Weird” Trap

If every fact is made bizarre, nothing is meaningfully surprising relative to the learning structure.

Worse, the learner may allocate attention to decorative novelty instead of the causal distinction.

Use surprise where there is a real expectation worth correcting.

Do not manufacture theatrical mismatch merely to create arousal.

21. The “Everything Fits” Trap

The opposite problem is teaching only examples that fit the rule perfectly.

Learners then build a clean schema with no knowledge of its edges.

Add deliberate boundary cases:

• one standard example;
• one near neighbour;
• one exception;
• one case that looks like the rule but fails a condition.

The schema becomes useful because it carries both center and edge.

22. The Center-to-Edge Route

1. Center: build the normal schema with clear representative examples.
2. Predict: ask what the schema expects in a new case.
3. Contrast: introduce a meaningful mismatch.
4. Explain: identify which assumption or condition changes.
5. Source-check: remember where the exception came from and what evidence supports it.
6. Mix: interleave congruent, near-congruent and incongruent cases.
7. Edge: classify an unfamiliar case and justify whether the schema applies.

23. The School Route: Sequence Typicality Before Exception

A curriculum that starts with exceptions can leave novices without the schema needed to understand why the exception matters.

Often the stronger sequence is:

representative case → repeated structure → prediction → exception → boundary condition.

This is not a universal instructional law. Some misconceptions need early confrontation. But an exception earns its explanatory power from a model stable enough to be violated.

24. The Systems Route: Schemas Compress Normal Operations; Alerts Preserve Exceptions

Large systems often treat routine events differently from exceptions.

Routine events can be compressed because they fit the expected model. Exceptions receive more explicit logging because they may signal a change in state.

Human memory is not an enterprise monitoring system. The analogy simply highlights a useful design principle:

let the schema carry what is normal, and explicitly tag what changes the rule.

25. The Financial Route: Base Case and Tail Risk

Financial analysis distinguishes a base case from scenarios that violate ordinary expectations.

Studying can use the same conceptual split.

Know the base rule well enough that exceptions are visible.

Then invest extra attention where a low-frequency exception has high consequence.

The analogy is about prioritisation, not evidence for memory mechanisms.

26. The Learning Route: Use Two Retrieval Questions

After studying a concept, retrieve two kinds of answer.

1. Schema question: What usually happens and why?
2. Boundary question: What case violates that expectation, and what condition makes the difference?

This protects both efficient general knowledge and exception memory.

27. The Education Route: Ask What Kind of Memory the Assessment Requires

A multiple-choice recognition test and a source-evaluation task do not measure the same thing.

If instruction uses surprise to strengthen contextual memory but the exam only tests item recognition, the educational payoff may differ from what was expected.

Conversely, if learners must evaluate evidence provenance, source memory deserves explicit practice.

Match study to the future retrieval job.

28. The Training Route: Congruent–Incongruent Contrast Sets

1. Choose one stable rule or schema.
2. Give two standard examples.
3. Ask the learner to predict a third case.
4. Introduce one carefully chosen mismatch.
5. Ask what feature caused the mismatch.
6. Mix standard and exception cases without labels.
7. Test item recognition and source/context separately after a delay.

The goal is not to maximise surprise. It is to train accurate expectation plus accurate revision of expectation.

29. The Improvement Route: Track Which Memory Dimension Failed

Failure	What it may mean
Item not recognised	Weak item representation or access
Item recognised, source lost	Content survived but provenance/context weakened
Rule remembered, exception lost	Schema center stronger than boundary
Exception remembered, rule lost	Novelty dominated structure
Both remembered, condition confused	Relation between schema and exception not integrated

A single “correct/incorrect” score can hide these different learning states.

30. The World Route: Expertise Needs Expectations—and the Ability to Notice Their Failure

Experts rely on schemas constantly.

A clinician recognises a familiar presentation. An engineer recognises a standard failure pattern. A lawyer recognises a familiar doctrine. A programmer recognises a common bug family.

Expertise would be painfully slow without those expectations.

But expert judgement also depends on noticing when the case does not fit.

The mature learning target is therefore not “expect nothing.” It is:

build strong expectations, then preserve sensitivity to meaningful violations.

31. Parent and Tutor Guide: Ask for the Normal Case and the Exception

When a child says they understand a topic, ask two questions:

• What normally happens?
• What is one case where that expectation fails or needs qualification?

If the learner knows only exceptions, the schema may be weak.

If the learner knows only the standard rule, boundary knowledge may be weak.

Then ask for the condition that separates the two.

32. What Not to Do

• Do not claim surprising information is always remembered better.
• Do not claim expected information is boring and therefore weakly encoded.
• Do not collapse item recognition, source memory and perceptual detail into one memory score.
• Do not make every lesson bizarre in an attempt to manufacture memorability.
• Do not teach exceptions before learners have enough structure to understand why they are exceptions.
• Do not treat schema fit as proof that a claim is true; familiar expectations can be wrong.
• Do not turn two 2026 laboratory-study lines into a universal classroom prescription.

33. Evidence Boundary

The 2026 studies discussed here use controlled memory tasks with object–scene relationships and university-adult samples. They offer valuable evidence that contextual congruency, expectedness and surprise can affect different memory outcomes in different ways.

They do not establish one universal schema-congruency effect for school learning, nor do they show that surprise should be maximised in teaching. Effects depend on materials, degree of congruency, retrieval task and strategy.

The educational applications in this article are therefore mechanism-informed proposals: use prior knowledge to create efficient structure, make meaningful mismatches visible, and test both central content and contextual/source detail rather than assuming one kind of memory stands for all the others.

34. Return: Memory Likes Both a Map and a Reason to Redraw It

A schema gives new information somewhere to land.

That makes learning efficient.

A real mismatch tells the learner that the map is incomplete.

That makes learning adaptive.

The two are not enemies.

Build the expected structure first. Predict with it. Notice the case that refuses to fit. Remember not only the exception but where it came from and why it matters. Then update the schema without throwing away the structure that still works.

Continue through The Prior Knowledge Paradox, The Distinctiveness Effect, Context Reinstatement, Self-Derived Knowledge, Source Monitoring, the How Studying Works Numbered Series Reading Index and the How X Works Hub.