How Studying Works | Levels of Processing — Why Meaning Usually Builds Stronger Memory Than Surface Repetition

HSW-0164 · How Studying Works

You read the same page three times.

The sentences become familiar. The words look easy. The page starts to feel learned.

Then someone closes the book and asks:

What does this idea mean, how does it connect to what you already know, and when would you use it?

The answer is much harder.

Levels of processing is a memory framework proposing that later retention depends strongly on the kind of processing performed during encoding. Meaning-based or semantic processing often produces stronger memory than processing limited to surface appearance or sound.

The practical lesson is powerful but easy to oversimplify. “Deep” does not mean “spend longer.” It does not mean “think philosophically about everything.” And it does not mean that surface form never matters.

It means that what the learner does with the information changes what becomes retrievable later.

The 50-Second Read

• Memory depends on processing, not exposure alone. Seeing something repeatedly is not the same as analysing its meaning.
• Semantic processing usually improves later retention. Asking what something means and how it relates to other knowledge often outperforms focusing only on appearance or sound.
• Deeper encoding can create a stronger starting memory without necessarily changing the rate of forgetting. Recent research found better immediate and delayed recognition after deep processing but similar forgetting slopes.
• Surface processing still matters when surface form is the target. Spelling, notation, pronunciation, exact wording and visual symbols sometimes require attention to form.
• Transfer-appropriate processing limits the slogan. The best encoding is partly the encoding that prepares the operations required at retrieval.
• Meaning needs structure. Elaboration, comparison, explanation, examples and prediction can make semantic processing richer and more distinctive.
• Study quality is not measured by minutes or repetitions. Audit the operations the learner actually performed.

1. The Same Word Can Be Processed in Different Ways

Take the word ecosystem.

You could ask:

• Is it printed in capital letters?
• Does it rhyme with another word?
• How many syllables does it have?
• What does it mean?
• How is it different from a habitat?
• What would happen to an ecosystem if one population disappeared?

All six questions make you process the same word.

They do not create the same memory representation.

The classic levels-of-processing framework associated shallower tasks with structural or phonological analysis and deeper tasks with semantic analysis. The enduring insight was that memory is strongly shaped by the operations carried out on information during learning.

2. “Deep” Is a Processing Description, Not a Moral Ranking

Students hear “deep learning” and often assume deeper must always mean better.

That is too crude.

If the future task asks you to spell accommodation, visual and orthographic detail matters.

If the future task asks you to pronounce a new vocabulary word, phonological encoding matters.

If the future task asks you to explain why inflation changes purchasing power, semantic and relational processing becomes central.

The useful question is not:

“Was this processing deep?”

It is:

What representation did this processing build, and is that representation useful for the future task?

3. Recent Evidence: Deep Processing Starts Stronger

A 2024 study published in Memory & Cognition and appearing in the 2025 volume tested whether levels of processing change not only initial memory but also the rate of later forgetting.

Across three experiments, semantic processing produced substantially better immediate and delayed recognition than shallower rhyme or visual-pattern processing. See Peng, Logie and Della Sala, Effect of levels-of-processing on rates of forgetting.

That supports the practical value of semantic encoding.

But the next result is equally important.

4. Deep Processing Did Not Necessarily Slow the Forgetting Rate

In those experiments, deeper processing created a stronger memory level, but the slopes of forgetting were not significantly different between deep and shallow conditions.

That means a common study story needs repair.

It is tempting to say:

“Deep learning lasts longer because it decays more slowly.”

The new evidence suggests a more careful description:

deep processing can build a stronger memory from the start, and that advantage can remain visible later even if the underlying rate of decline is similar.

For students, this matters. Deep processing is not a substitute for spacing, retrieval and maintenance. A stronger starting point can still weaken with time.

5. Individual Learners Do Not Gain Identically

The levels effect is reliable at group level, but people vary.

A 2024 study in the journal Memory examined individual differences in levels of processing using semantic and rhyme judgments. It found the expected overall advantage for semantic processing while also documenting meaningful variation across individuals. See Unsworth and Miller, An examination of individual differences in levels of processing.

The educational implication is familiar but important: a reliable average effect does not justify assuming identical benefit for every learner, task or subject.

6. Maintenance Repetition and Elaborative Processing Are Different

Repeating a phrase can keep it active.

“Mitochondria release energy. Mitochondria release energy. Mitochondria release energy.”

That may help temporary accessibility.

Elaborative processing asks more:

• What process is actually happening?
• What inputs and outputs are involved?
• How does this connect to respiration?
• What would change if oxygen were limited?
• What observation would support the model?

Now the learner is building relationships rather than only preserving a phrase.

7. Meaning Is Not Enough if It Is Vague

A learner can “think about meaning” very weakly.

“This is about energy.”

That is semantic, but it may still be too shallow for useful transfer.

Stronger semantic processing often includes:

• precise definition;
• cause and effect;
• comparison with neighbours;
• examples and counterexamples;
• links to prior knowledge;
• prediction;
• explanation of boundary conditions.

The quality of the semantic work matters.

8. Levels of Processing vs Elaboration

How Elaboration Works owns the learning mechanism of making new knowledge connect to prior knowledge through explanation, relationships and additional meaning.

Levels of processing is the broader encoding framework: it asks how the nature of processing—from surface form toward meaning—changes later memory.

Elaboration is one important way to make semantic processing richer.

9. Levels of Processing vs Self-Explanation

How Self-Explanation Works owns the act of generating explanations for steps, concepts and relationships.

Self-explanation often creates semantic and relational processing, which can help build stronger memory.

But not all deep processing requires self-explanation, and not every self-explanation is accurate or useful.

10. Levels of Processing vs Transfer-Appropriate Processing

This is the most important boundary.

Transfer-Appropriate Processing argues that memory performance depends partly on the match between operations at encoding and operations required at retrieval.

Robert Lockhart’s review of the relationship between levels of processing and transfer-appropriate processing concluded that no simple “deeper is universally more robust” account is sufficient. See Levels of processing, transfer-appropriate processing, and the concept of robust encoding.

The practical synthesis is:

• semantic processing is often an excellent general foundation;
• but study should also practise the exact operations future performance will demand.

11. Mathematics: Process the Relationship, Not Only the Formula Shape

A student can memorise the quadratic formula visually.

That surface memory may help reproduce the symbols.

Deeper mathematical processing asks:

• What equation type is this formula solving?
• Why does the discriminant change the number of real roots?
• How is the formula related to completing the square?
• What does each term represent?
• When is another method faster?

Now the formula is embedded inside a decision system.

But exact symbol form still matters. Mathematics therefore needs both semantic structure and precise surface control.

12. English: Meaning, Form and Register Need Different Processing

Vocabulary learning exposes the limit of “meaning only.”

To own a word, the learner may need:

• meaning;
• spelling;
• pronunciation;
• collocation;
• register;
• connotation;
• grammatical behaviour.

Semantic processing supports meaning. Orthographic and phonological processing support form. Usage tasks connect the word to contexts where it must later be produced.

A complete learning plan therefore processes the word at several useful levels rather than replacing all form work with semantic discussion.

13. Science: Turn Labels Into Mechanisms

Science revision often becomes a surface-label exercise:

• memorise the diagram;
• memorise the keyword;
• memorise the model answer.

Deeper processing asks what the model predicts.

For diffusion:

• What is moving?
• Why is net movement directional?
• What changes the rate?
• What would equilibrium mean?
• What observation would contradict the model?

The keyword becomes a compressed handle for a mechanism rather than a substitute for one.

14. The Processing Audit

After studying, ask what operations actually occurred.

Processing	Example	Likely value
Surface	Notice font, spelling, symbols	Useful when exact form matters
Phonological	Say, rhyme, pronounce	Useful for spoken form and sound discrimination
Semantic	Define meaning	Strong general memory foundation
Relational	Compare, connect, explain cause	Builds organised knowledge
Transfer	Use in a new task	Tests whether processing became usable capability

A study session dominated by surface and recognition tasks may feel long while producing a thin memory structure.

15. Why Rereading Can Mislead

Rereading increases familiarity with surface form.

That can be useful for orientation, but repeated exposure can create a feeling of progress without forcing semantic reconstruction.

A stronger rereading protocol changes the operation:

1. Read one short section.
2. Close the source.
3. State the meaning.
4. Connect it to one prior idea.
5. Generate one example.
6. Ask what future question could require this knowledge.

The page has not changed. The processing has.

16. The School Route: Ask Questions That Force Meaning

Question design changes processing.

Compare:

• “Underline the definition.”
• “Explain why this definition excludes the neighbouring concept.”

Or:

• “Copy the formula.”
• “Predict what happens if one variable doubles and explain why.”

The second prompt creates a different encoding event.

Teachers can therefore improve retention not only by adding more practice but by changing what learners must do cognitively during practice.

17. The Systems Route: Input Volume Is Not Transformation Depth

A system can process a huge volume of data through a shallow transformation.

Students can do the same.

Fifty pages scanned is not equivalent to ten pages analysed for relationships, mechanisms and transfer.

The systems question is:

What transformation happened between input and memory?

If the answer is only “the information passed through my eyes again,” throughput may be high while learning conversion remains low.

18. The Financial Route: Buy Representation Quality, Not Just Exposure Time

Study time is an investment.

Two hours can buy:

• twenty repetitions of a shallow task; or
• a smaller number of meaning-rich retrieval, comparison and transfer attempts.

The best allocation depends on the target. But when conceptual memory is the goal, quality of processing can dominate raw exposure count.

The learner should ask:

What kind of memory am I purchasing with this minute?

19. The Learning Route: Move From Meaning to Retrieval to Transfer

Semantic processing is a strong beginning, not the final test.

A useful progression is:

1. Meaning: explain the concept accurately.
2. Relation: connect it to neighbours and prerequisites.
3. Retrieval: reconstruct it without the source.
4. Application: use it in a familiar task.
5. Transfer: recognise and use it in a changed context.
6. Verification: check limits, exceptions and evidence.

This turns deep encoding into usable capability rather than stopping at an elegant explanation.

20. The Education Route: Teach Students to Name the Operation

Students often describe study methods by objects:

• notes;
• flashcards;
• videos;
• worksheets;
• AI;
• textbooks.

Levels of processing teaches a better question:

What operation did the tool make me perform?

• Did the flashcard require free recall or recognition?
• Did the video require prediction or passive watching?
• Did AI generate the explanation or did the learner construct and verify one?
• Did the worksheet require method choice or only repeated execution?

The tool does not determine depth. The learner operation does.

21. The Training Route: The Five-Layer Processing Drill

Choose one important concept and process it through five layers.

1. Form: state the exact term, notation or spelling.
2. Meaning: define it in your own accurate words.
3. Connection: link it to two prior ideas.
4. Discrimination: contrast it with the nearest confusable concept.
5. Transfer: solve or explain one unfamiliar application.

Then close all sources and retrieve the structure after a delay.

22. The Improvement Route: Audit Depth by What Survives

Do not infer processing depth from effort alone.

A difficult copying task can feel exhausting while remaining conceptually shallow.

Measure the output:

• Can the learner explain the concept without the source?
• Can they distinguish it from a neighbour?
• Can they generate an example?
• Can they retrieve it after delay?
• Can they recognise when it applies?
• Can they use it under changed surface features?

These receipts tell you more than how intense the study session felt.

23. The World Route: Experts Encode Meaningful Structure

Experts do not merely remember more facts. They often encode situations through meaningful structures that organise many details.

A doctor sees a pattern of symptoms. An engineer sees a load path. A lawyer sees a legal issue and controlling authority. A programmer sees a state transition. A teacher sees a misconception family.

Expertise changes what features are processed as meaningful.

That is why deep processing and prior knowledge reinforce each other: the more structure you possess, the more structure you can notice in new information.

24. When Surface Processing Is Exactly the Right Job

Do not turn levels of processing into contempt for detail.

Some tasks are surface-sensitive by design:

• spelling;
• musical notation;
• chemical symbols;
• mathematical signs;
• pronunciation;
• quotation accuracy;
• coding syntax;
• diagram labels.

A learner who understands meaning but ignores exact form can still fail.

The mature system processes both the meaning and the representation that must survive performance.

25. What Not to Do

• Do not equate deep processing with spending more time.
• Do not assume semantic encoding slows the rate of forgetting; recent evidence found stronger memory but similar forgetting slopes.
• Do not ignore spelling, notation or exact surface form when the future task requires them.
• Do not treat “depth” as a literal anatomical layer of memory.
• Do not use elaboration to generate inaccurate connections.
• Do not stop at meaning if future performance requires retrieval, method choice or transfer.
• Do not confuse levels of processing with transfer-appropriate processing.

26. Evidence Boundary

Levels of processing remains an influential framework rather than a complete theory of memory. The original idea of a simple ordered “depth” dimension has been refined by later work on elaboration, distinctiveness, retrieval conditions, prior knowledge and transfer-appropriate processing.

Current evidence still strongly supports the educational value of semantic encoding for many verbal and conceptual memory tasks. But “deep is always better” is too simple. The future test matters, the target representation matters, and stronger initial encoding does not automatically imply a slower forgetting rate.

27. Return: Ask What the Mind Had to Do

Students often count pages, minutes and repetitions.

Levels of processing asks for a different accounting system.

Did you notice the shape?

Did you hear the sound?

Did you understand the meaning?

Did you connect the idea?

Did you retrieve it?

Did you use it where the future task will demand it?

The quality of study is not only how much information passed through the learner. It is what the learner had to do with the information while it passed through.

Continue through Transfer-Appropriate Processing, Elaboration, Self-Explanation and the How Studying Works Numbered Series Reading Index.