The Forgetting Curve Explained: What It Shows and How to Remember More

The forgetting curve describes a simple but important pattern: after learning something, our ability to retrieve it often declines with time, and much of that decline can happen relatively early.

But the familiar diagram of a smooth downward curve is easy to misunderstand.

There is no universal rule saying that everyone forgets a fixed percentage after one hour, one day, or one week. The original forgetting curve came from a very specific experiment by Hermann Ebbinghaus, using artificial material and a particular way of measuring memory.

The useful lesson is broader: learning once is usually not enough for knowledge you want to retain for a long time. Returning to material after a delay—and trying to retrieve it rather than only rereading it—can substantially change what happens next.

What is the forgetting curve?

The forgetting curve is a model of how memory performance changes as time passes after learning.

In its familiar form, the curve falls quickly at first and then more gradually.

The broad pattern matters more than the exact percentages.

It suggests that forgetting is often nonlinear. The difference between remembering something immediately after learning and remembering it later can be large, while additional time does not necessarily produce the same amount of further forgetting.

That pattern has appeared in many memory experiments, although the exact shape depends on what was learned, how well it was learned, how memory is tested, and what happens between learning and testing.

So the forgetting curve is better understood as a general pattern of retention over time, not a personal countdown clock for your memory.

What Hermann Ebbinghaus actually measured

The forgetting curve is associated with German psychologist Hermann Ebbinghaus, whose experiments on memory were conducted in the late nineteenth century and published in Über das Gedächtnis in 1885.

His work was unusual for its time because he tried to study memory quantitatively.

He used nonsense syllables

Ebbinghaus wanted material that carried as little existing meaning as possible.

Normal words are difficult to control because some are familiar, emotional, easy to visualize, or already connected to things you know. He therefore created large sets of artificial consonant-vowel-consonant syllables.

He then memorized lists of them.

There is an important limitation here: Ebbinghaus was also his own participant.

His experiments were rigorous and historically important, but they were not large modern studies involving hundreds of independent learners studying meaningful university material.

That matters when interpreting the exact values associated with his curve.

He measured relearning, not simply “percent remembered”

One of the most misunderstood parts of the forgetting curve is its vertical axis.

Ebbinghaus used a measure called savings.

Imagine that learning a list for the first time takes 20 minutes. A week later, you cannot reproduce it perfectly, but relearning the same list takes only 12 minutes.

You have saved 8 minutes compared with the original learning:

Initial learning: 20 minutes
Relearning:       12 minutes
Saved:             8 minutes

Savings = 8 / 20 = 40%

That 40% savings indicates that the original learning still had an effect.

It does not necessarily mean that you could freely recall 40% of the material before relearning it.

This distinction explains why many modern versions of the forgetting curve are misleading. They relabel Ebbinghaus's savings values as if they directly measured the percentage of information remaining in memory.

They did not.

What the forgetting curve actually shows

Ebbinghaus tested memory after delays ranging from less than an hour to 31 days.

His results showed a recognizable pattern: savings declined considerably during the earlier intervals and then changed more slowly over longer delays.

A close replication published in 2015 produced a broadly similar curve.

More generally, research conducted long after Ebbinghaus has repeatedly found that retention often decreases as the delay between learning and testing increases.

But there is not one exact forgetting curve that describes every kind of memory.

A useful interpretation is:

Forgetting can be substantial relatively soon after learning, but the rate of change is not constant.

That is different from saying that every person forgets the same percentage of everything at the same times.

Why the “70% forgotten after one day” rule is misleading

You may have seen claims such as:

• “We forget 50% of new information within an hour.”
• “We forget 70% within 24 hours.”
• “Only 10% remains after a week.”

These numbers are often presented as if they were a universal law discovered by Ebbinghaus.

They should not be treated that way.

First, Ebbinghaus measured savings during relearning, not a direct percentage of information still available for conscious recall.

Second, his experiments used nonsense syllables and one participant: Ebbinghaus himself.

Third, meaningful learning is affected by variables that his controlled material deliberately minimized.

Consider the difference between learning:

• an arbitrary syllable,
• the capital of a country,
• a programming concept connected to years of prior knowledge,
• the logic of a mathematical proof,
• a story that has personal meaning.

It would be surprising if all of them followed exactly the same retention curve.

The exact percentage is therefore less useful than the underlying principle: without later retrieval or additional learning, access to newly learned information often becomes more difficult with time.

The curve describes forgetting, not a single cause of forgetting

Another common mistake is to interpret the forgetting curve as proof that memories simply decay because time passes.

The curve itself cannot establish that.

It describes the relationship between a memory measure and the time since learning. It does not identify one universal biological mechanism responsible for every failure to remember.

Memory performance can also be affected by factors such as:

• how strongly the material was learned initially,
• prior knowledge,
• similarity to other information,
• interference from later experiences,
• available retrieval cues,
• sleep and consolidation,
• the way memory is tested.

This matters practically.

Time is important because longer delays create more opportunities for retrieval to become difficult, but time alone does not determine what you will remember.

Reviewing does not simply “reset the curve”

Many forgetting-curve diagrams show a sequence like this:

Learn
↓
Forget
↓
Review
↓
Curve resets
↓
Forget more slowly
↓
Review again

This is a useful illustration, but it should not be interpreted literally.

A review is another learning event. What happens afterward depends on what you actually do during that review.

Looking at the answer for five seconds is not necessarily equivalent to trying to reconstruct an idea from memory.

Successfully retrieving information after a delay can strengthen later retention. Repeated learning episodes separated in time can also outperform the same amount of practice compressed into a single session.

These ideas lead to two related study methods:

• spaced practice — distributing learning across time;
• retrieval practice — actively trying to bring information back from memory.

How spaced repetition relates to the forgetting curve

Spaced repetition takes the basic problem illustrated by the forgetting curve and turns it into a scheduling problem:

If important knowledge becomes harder to retrieve over time, when should you return to it?

The answer is not a universal sequence of fixed intervals.

Research on distributed practice shows that the useful spacing between learning events depends partly on how long you eventually want the material to remain available.

If you need information next week, an appropriate study schedule will not necessarily be the same as one designed to retain it for a year.

This evidence supports spacing as a learning principle; it does not imply one universally optimal scheduling algorithm.

This is why rigid rules such as:

Review after 1 day
Review after 3 days
Review after 7 days
Review after 30 days

should be treated as heuristics, not laws of memory.

In practice, spaced-repetition systems can use recall performance or self-assessment to adjust intervals instead of following one universal sequence.

If retrieval remains easy, the next review can move farther away. If you can barely retrieve the material—or cannot retrieve it at all—it may make sense to return sooner.

That is the practical idea behind spaced repetition for notes: spend fewer reviews on material that remains stable and more attention on knowledge that is becoming difficult to retrieve.

Active recall: reviewing is not the same as rereading

Spacing answers when to return.

Retrieval practice changes what you do when you return.

Suppose you are trying to remember a topic from a biology course.

You could open your notes and immediately reread them.

Or you could first ask yourself:

• What are the main stages of this process?
• Why does each stage happen?
• Which concepts can I explain without looking?
• Where does my explanation become vague?

Only after trying to retrieve the material do you open the note and compare.

The second approach gives you information that rereading cannot: whether you can actually reconstruct the knowledge without seeing it.

Experimental research on the testing effect and retrieval practice has repeatedly found benefits for later retention compared with additional study under many conditions.

This does not mean rereading is useless.

You still need the source material to learn, understand, correct errors, and fill gaps.

The distinction is that exposure and retrieval are different learning activities.

A practical review can use both:

Try to retrieve
↓
Notice gaps
↓
Check the material
↓
Correct or deepen understanding

A practical way to use the forgetting curve

You do not need to estimate your personal forgetting curve mathematically to benefit from the idea.

A useful workflow is simpler.

1. Learn for understanding first

Spaced repetition is not a substitute for understanding material in the first place.

Build the mental model first.

Connect new information to what you already know, work through examples, and resolve the parts that remain unclear.

2. Try to retrieve before looking

After some delay, do not begin by immediately rereading the source.

Ask yourself what you can reconstruct.

For a small fact, that may be a direct question.

For a larger topic, it may mean explaining the main ideas, drawing the structure, recreating a process, or answering several questions.

3. Use the note to check yourself

Then return to the original material.

Look for:

• facts you missed,
• concepts you confused,
• details that no longer make sense,
• places where your explanation was incomplete.

The goal is not to prove that you remembered everything perfectly. It is to find the difference between what feels familiar and what you can actually retrieve.

4. Return later instead of repeatedly studying immediately

An immediate second reading may feel productive because the material is still highly accessible.

But if your goal is long-term retention, you eventually need to succeed after some forgetting has occurred.

Spacing creates that delay.

5. Let the interval grow when recall remains strong

If the material remains easy to retrieve, reviewing it again tomorrow may add little value.

Move the next review farther away.

When retrieval becomes difficult, shorten the interval.

This makes repeated learning more selective instead of treating every piece of knowledge as equally fragile.

6. Keep reviewing only what is worth remembering

Not everything you read needs to become permanent knowledge.

This is easy to miss when learning systems turn every highlight or sentence into another item to review.

A better question is:

Which knowledge will still be useful to me months or years from now?

Spaced repetition is most useful when applied intentionally.

How Yuhimo uses this idea with notes

Many spaced-repetition systems are built around individual flashcards.

That works well when the learning unit is an isolated fact, definition, translation, or short question-answer pair.

But knowledge is not always naturally atomic.

A topic may be better represented as:

• a short explanation,
• a structured set of notes,
• a technical concept,
• a long study page,
• a group of related ideas.

Yuhimo uses the page itself as the main learning unit.

You can keep ordinary notes in your knowledge base and decide which pages are important enough to revisit.

When reviewing a learning page, you can first try to reconstruct the material from memory. Questions can be attached to the page as retrieval prompts, but they are optional.

After the review, you assess your own recall:

• Forgot
• Barely
• Remembered
• Easily

That self-assessment is then used to decide when the page should return for another review.

The system does not attempt to grade your knowledge automatically. A large page may contain context that you want to keep but only several ideas that you actually need to remember. You remain responsible for deciding whether you knew what mattered.

This is one way to use spaced repetition without turning everything into flashcards, while keeping the material in the same notes-first workflow rather than maintaining a separate set of atomic cards.

The forgetting curve provides the motivation for returning to knowledge. The review system handles the scheduling. The learner still decides what is worth knowing.

What the forgetting curve cannot tell you

The forgetting curve is useful precisely when its limitations are kept clear.

It cannot tell you:

Exactly how much you will forget tomorrow. Retention depends on the learner, material, learning conditions, and how memory is measured.

The precise moment you should review something. Spacing research supports delayed, distributed practice, but there is no single optimal calendar for all learning goals.

Why every memory was forgotten. A retention curve describes performance over time; it does not establish one cause of forgetting.

Whether something is worth memorizing. Learning systems can schedule reviews. They cannot decide which knowledge deserves years of your attention.

Whether familiarity equals knowledge. Material can look obvious when it is in front of you and still be difficult to reconstruct without cues.

The most useful interpretation of the forgetting curve is therefore not “memory loses X percent every day.”

It is this:

Important knowledge becomes harder to retrieve when it is left unused, so long-term learning should include deliberate opportunities to retrieve it again over time.

Frequently asked questions

How quickly does the forgetting curve happen?

There is no single rate. Forgetting can be relatively rapid after initial learning and slower later, but the timing depends on the material, learner, and way memory is tested.

Is the forgetting curve supported by research?

Yes. Forgetting over time is well documented, and an Ebbinghaus-like pattern has been replicated. What research does not establish is one exact curve or set of percentages that applies to every learner and every type of material.

Does spaced repetition reset the forgetting curve?

“Reset” is a metaphor, not a precise scientific description. A spaced review creates another learning event, and retrieval during that review can improve later retention.

How often should you review something to avoid forgetting?

There is no universal schedule. Review after a delay, test what you can retrieve, and adjust later intervals according to how well the material is holding up and how long you need to retain it.