The Three Steps of Memory — Putting It In, Keeping It, Taking It Out

A Phone Number Heard Once, a Home Address from Long Ago

A person you meet for the first time at a Friday evening gathering reads out their phone number. "010-0482-7391." This is a number made up for this piece, and it appears to be one that is not actually assigned. Since it is a Korean mobile phone number, it starts with 010. Your phone is inside your bag. You mutter the number under your breath as you open your bag, when the person next to you speaks to you. When you answer and turn back around, the number is already blurry. You do not know if the last four digits were 7391 or 7319.

Yet the address of the house you lived in when you were young flows out smoothly even decades later. A number heard once scatters at a single word of conversation, while an address from long ago stays for a lifetime. Both are memories, but they are this different.

This piece follows the path where memory is made in three stages. They are the stages of holding it, leaving it, and pulling it out. For the explanation, we will use a desk and a library bookshelf. Only a few books you are about to open sit on the desk, while books to keep for a long time are shelved in the bookshelf. We will carry the phone number heard at the gathering all the way to the end of the piece. The person who read out the number is also a made-up example for the sake of the explanation.

The Desk and the Bookshelf — Working Memory and Long-Term Memory

While you repeat a number in your mouth, the number is on the desk. Working memory is the memory you hold onto in your head for a moment to use right now. Holding onto the number until you take your phone out of your bag is the job of working memory. The amount of information you can hold onto in your head like this is very small (PMID 20362427). The desk is small.

Your childhood home address is on the bookshelf. Long-term memory is the memory stored longer than the brief moment on the desk, lasting anywhere from a long time to a lifetime. The menu you ate for lunch today, known words, a friend's face, and the commute to work are also shelved here. The ability to store experiences from a few days to a lifetime belongs to this bookshelf (PMID 40258487).

There is a path going back and forth between the desk and the bookshelf. The number on the desk has to move to the bookshelf so it remains next week, and to use the address on the bookshelf, you have to pull it back out onto the desk. Memory research divides this path into three stages: encoding, storage, and retrieval (PMID 19300446).

The number you just heard is on the narrow desk, and the old address is on the bookshelf.

How Many Items Can Fit on the Desk

A phone number has 11 digits. Holding all 11 digits individually is a handful. Then, how many items can fit on the desk?

The old answer is "around 7." A 1956 paper dealing with the limit of memory that immediately repeats what was just heard was titled "The Magical Number Seven, Plus or Minus Two" (DOI: 10.1037/h0043158). However, even the author of that paper deferred judgment on whether the number 7 appearing everywhere—like the seven seas and the 7 days of the week—had a deep meaning, and wrote that it was likely a coincidence.

After that, the view setting the number smaller gained traction. There is an assessment that 7 was an overly large estimate and the estimate "4 plus or minus 1" took its place (DOI: 10.1515/cog-2015-0112).

Counting by the number of items itself is a point of contention. There is a view that for memory holding what is seen with the eyes, three or four is the upper limit. Conversely, there is a view that there is no upper limit counted by items, and a limited amount is shared across multiple items (PMID 20362427). There are experiments showing that the difference between a model viewing it as 4 and another model viewing it as 7 can be explained in large part by the difference in what is counted as a single item (DOI: 10.5334/joc.387).

How many items can fit on the desk is still debated, and the only thing we can say for sure is that the amount we can hold at once is very small. There is no need to memorize 7±2 as a fixed rule.

Why the desk is so narrow is also still unsolved. One review compared three explanations. The explanation that items on the desk naturally fade over time was deemed hard to sustain, at least for verbal materials. The explanation that the available power is limited is still possible, but it was seen as inconsistent with several results. It concluded that the explanation where the items being held interfere with one another explains the results best, but even here, unresolved problems remain (PMID 26950009).

Encoding — Putting a Label on It and Bringing It In

The moment you hear a phone number at a gathering, the digits enter your ear as a human voice. Encoding is the process of converting what you see and hear into a form that can be stored in memory and bringing it in. In terms of a library, this is the job of putting a label on a new book and shelving it. Depending on how you label it, the amount you can handle on your narrow desk changes.

There is a way to put 11 numbers onto a narrow desk. If you take them in chunks of three, like 010, 0482, and 7391, it becomes much easier to hold onto them than if you laid out all 11 in a single line. Chunking is the process of binding individual pieces of information into meaningful bundles and handling them as a single unit. Another claim of the paper that talked about the magic number 7 was that if you organize incoming items into bundles and recode them, you can clear the narrow bottleneck or at least widen it (DOI: 10.1037/h0043158).

When creating bundles, things already on the bookshelf do the work. Since 010 is the prefix attached to every Korean mobile number, it comes in all at once as a single chunk already on the shelf. The new things you need to hold onto are the remaining eight digits, and even these are easier to handle when broken into groups of four than as individuals. However, meaningless four-digit numbers do not immediately become a single unit. For a bundle to be handled like a single chunk, it must be meaningful or familiar.

There is a theory that people who are skilled in a field borrow what they already know on a much larger scale. According to this theory, in skilled tasks like chess, mental arithmetic, and medical diagnosis, the results produced during work are placed into long-term memory, and only the cues to retrieve them are left on the desk (DOI: 10.1037/0033-295x.102.2.211). This theory explains why skilled people can handle so many things with a narrow desk when reading or working.

When you chunk things as you store them and tie them to what you already know, you can handle more with the same desk.

Consolidation — Until a New Book Finds Its Place on the Shelf

For a number you just took in to stay until next week, one more step is needed. Consolidation is the process where newly taken-in memories become firm over time and turn into long-lasting memories. The idea that new memories do not set all at once but slowly harden over time has guided memory research for over 100 years (DOI: 10.1126/science.287.5451.248).

To put it in terms of a library, new books are first placed on a temporary shelf near the entrance, and over time they are moved to their permanent shelves. The hippocampus is a small area deep inside the brain that serves as the initial resting place for new memories, like things you experienced or facts you learned. This corresponds to the temporary shelf in a library. When the hippocampus and its surrounding areas are damaged, recent memories are broken while very old memories remain intact. This observation became the basis for the explanation that certain types of memories initially rely on the hippocampus and gradually find a place elsewhere in the brain (DOI: 10.1073/pnas.91.15.7041). However, since there are findings that the hippocampus is used even when recalling old events, this explanation is still being refined today (PMID 30528612). This story will come up again later.

Sleep helps with this moving and re-shelving. The fact that sleep is beneficial for retaining memories is a well-established truth from over a century of research, and the currently leading explanation is that during sleep, the memories taken in that day are replayed and moved to long-lasting locations (DOI: 10.1152/physrev.00032.2012). However, how memories change as they are moved is still a matter of debate (PMID 40875205). Even while awake, studies show that taking a short break without doing anything right after learning helps solidify new memories (PMID 30683602).

Some books find their places exceptionally firmly. The fact that experiences that shake the heart greatly—whether joyful or distressing—stay for a long time is a finding pointed to by both animal and human studies (DOI: 10.1146/annurev.neuro.27.070203.144157). However, emotions do not always help memory, and there are also findings that they sometimes hindered it depending on the conditions (DOI: 10.3389/fpsyg.2017.01454).

The memories you take in find their places on the shelf through time and sleep.

Retrieval — Finding It by Its Label and Bringing It to the Desk

A month later, as you walk past the place where you met, the conversation you shared that day suddenly pops into your mind.

Retrieval is the act of finding a memory on the bookshelf and bringing it back out onto your desk. Just as you find a book in a library by its label, you find memories using cues. The label attached when storing something mixes the contents of the book with what was happening around you at the time.

The context reinstatement effect is a phenomenon where remembering works better if the surroundings from when the information was stored are recreated when you try to retrieve it (PMID 41525397). When pooling the results of 22 papers and 2,177 people, this benefit appeared consistently in both young and older adults, its size was relatively small, and it did not differ by age (PMID 41525397). Compared to when items to be memorized were stored separately one by one, this benefit was smaller when two items were paired and memorized together. Even if you were in a different place during retrieval than during storage, if that place was familiar, the benefit gained from recreating the original place tended to decrease (PMID 41525397).

This benefit does not always appear. In experiments, the benefit mostly appeared under conditions like these: when the surroundings were rich in meaning, when the items to be memorized were unfamiliar, and when there were not many cues given all at once. Also, it appeared when the items to be memorized were interwoven and stored together with the surrounding appearance, and when quite a bit of time had passed between storage and retrieval. Because methods vary from study to study, the results sometimes conflict (PMID 42550362).

Whether a memory that does not come to mind has vanished from the bookshelf or cannot be found because the label does not match is a story covered in "Forgetting — What Fades and What We Just Can't Get Out".

Memories are retrieved using cues, and if the surroundings from when they were stored are present again when retrieving them, they generally come to mind a little better.

The Scene You Recall May Have Been Rebuilt

You recall that Friday evening a month later. The scene is quite vivid. You even remember who sat where and what they talked about.

Here we return to the story of the hippocampus. New memories initially lean on a temporary shelf called the hippocampus and gradually settle elsewhere in the brain. Given that, you might think the hippocampus could rest when recalling old things, but in reality, the hippocampus is at work even then. This is the case despite evidence that the trace of a memory—namely, the mark left in the hippocampus when a memory is first stored—fades over time. Addressing this mismatch, it has been proposed that the hippocampus gathers pieces solidified across the brain without the original trace and rebuilds them into a single scene (PMID 30528612). In library terms, since the original book on the temporary shelf is already gone, it is closer to gathering pages scattered across various shelves and binding them into one volume on the spot.

People sometimes get the plot right while getting the details wrong, and pre-existing knowledge can twist memories in a certain direction. A computational model viewing memory as such a rebuilding process successfully reproduced these phenomena (PMID 41794201). There is also the hypothesis that a retrieved memory can go through a reconsolidation process. Although evidence emerged challenging the idea that once a memory solidifies it is final (PMID 16675390), two studies published around the same time reached different conclusions on just how much old memories can change at this time, and the debate continues (PMID 12408838).

Because a recalled memory may have been rebuilt, even a vividly recalled scene can contain incorrect pieces.

When You Try to Recall a Name at a Reunion

Two months later, you cross paths again at the gathering where you first met the person. You recognize the face, but the name does not come right away. Let's apply the four things we've looked at to this moment.

First is the desk. If you heard the name once that day and immediately talked to the person next to you, the name may have stayed briefly on the narrow desk and then scattered. What didn't make it to the bookshelf cannot be retrieved either.

Next are packing and leaving. If you split the phone number into three parts like 010, 0482, and 7391 that day, holding on to it would have been much easier than holding all 11 digits in a single line. If you packed the name that day and slept a few nights after that, the name had time to settle into the bookshelf.

And that is retrieval. The very place where you first met serves as a cue. This help, mostly confirmed by experiments, tends to be modest and does not always work, but it can slightly raise the likelihood that the name comes to mind.

Finally, it is rebuilding. "Was it Min-ji, or was it Min-jung?" Even if the recalled name is clear, wrong pieces might have mixed in. Checking a business card or contact info once is safer.

If You Hold It Up to the Week Before the Exam

We can hold the same picture up to studying, too. Imagine an exam is one week away, and there are about 20 terms to memorize.

The desk being narrow means you cannot put all 20 of them out at once and hold onto them. The more you group things that make sense together into chunks and tie them to what you already know, the more the amount you can handle at a time may increase.

The leading explanation is that sleep is the time that helps what was taken in that day settle into the bookshelf. If you pull an all-nighter the night before the exam, you miss out on that help sleep gives. Because consolidation does not happen only during sleep, taking a short break right after finishing studying without crowding in other tasks can also help consolidate new memories.

The exam room is different from the study room. Viewed through the context-dependent memory effect, taking the exam in the study room might be slightly more advantageous, but that help tends to be small and does not always appear.

Where to Look Next

It is not yet answered why desks are so narrow, or how memories transferred during sleep change.

"Forgetting — What Fades and What We Just Can't Get Out" distinguishes between cases where a book disappears from the shelf and cases where its label cannot be found. "Retrieval Practice — The More You Pull It Out, the More It Stays" covers the study method of intentionally pulling things out of memory for the purpose of learning, and "The Spacing Effect — Spread It Out and It Lasts" covers how to study by spreading it out and mixing it up. "The Science of Studying: What Actually Works" gathers these learning methods in one place along with the strength of their evidence.

Source Papers

• "Discrete capacity limits in visual working memory." (2010) — PMID 20362427 — The amount of information that can be actively held in the mind is very limited. Views contend whether the upper limit is roughly three or four items or if memory is divided continuously without an upper limit on the number of items (review).
• "The molecular and cellular basis of memory engrams: Mechanisms of synaptic and systems consolidation." (2025) — PMID 40258487 — The ability to record and store life experiences anywhere from days to a lifetime. Consolidation is the process of stabilizing and long-term storing memories, which occurs at two levels: synaptic and systems (review).
• "The anatomy of memory: an interactive overview of the parahippocampal-hippocampal network." (2009) — PMID 19300446 — The hippocampus and parahippocampal subregions each contribute to the encoding, consolidation, and retrieval of declarative memory (rat anatomical connection review).
• "The magical number seven, plus or minus two: Some limits on our capacity for processing information." (1956) — DOI: 10.1037/h0043158 — The span of immediate memory places severe limits on information processing. Organizing inputs into chunks and re-encoding them can bypass or expand the bottleneck. Judgment is deferred regarding the multiple instances of 7, suspecting coincidence.
• "Usage-based linguistics and the magic number four" (2017) — DOI: 10.1515/cog-2015-0112 — Miller's 7±2 is viewed as having been an overestimate, replaced in cognitive science by Cowan's "4±1", and argues that various structures of linguistic data align with 4.
• "Modelling Working Memory Capacity: Is the Magical Number Four, Seven, or Does it Depend on What You Are Counting?" (2024) — DOI: 10.5334/joc.387 — Compares the 4-unit model and 7-unit model across five experiments. Estimates vary depending on the task and discusses ways to reconcile the two views.
• "What limits working memory capacity?" (2016) — PMID 26950009 — Among three explanations for capacity limits, time decay is difficult to sustain at least for verbal materials, the resource explanation is hard to reconcile with various results, and the interference explanation fits best while having its own difficulties (review).
• "Long-term working memory." (1995) — DOI: 10.1037/0033-295x.102.2.211 — A theoretical framework in which text comprehension and skilled performance (mental arithmetic, medical diagnosis, chess) store processing results in long-term memory and provide direct access to them as retrieval cues in short-term memory.
• "Memory--a Century of Consolidation" (2000) — DOI: 10.1126/science.287.5451.248 — The consolidation hypothesis, proposing that new memories slowly harden over time, has guided memory research for 100 years (review).
• "Memory consolidation and the medial temporal lobe: a simple network model." (1994) — DOI: 10.1073/pnas.91.15.7041 — Following medial temporal lobe damage, recent memories are impaired while very old memories remain intact. Memories initially rely on this system before gradually taking root elsewhere.
• "About Sleep's Role in Memory" (2013) — DOI: 10.1152/physrev.00032.2012 — The benefit of sleep in maintaining memory has been established by a century of research. Current theories emphasize active consolidation, during which new memories are reactivated and transferred to long-term storage while sleeping (review).
• "Sleep's contribution to memory formation." (2026) — PMID 40875205 — Although most findings align with the concept of active consolidation during sleep, how memories are transformed and what is stored remain debated (review).
• "Memory Consolidation during Waking Rest." (2019) — PMID 30683602 — Recent studies showing that short breaks immediately after learning help consolidate new memories (review).
• "THE AMYGDALA MODULATES THE CONSOLIDATION OF MEMORIES OF EMOTIONALLY AROUSING EXPERIENCES" (2004) — DOI: 10.1146/annurev.neuro.27.070203.144157 — Animal and human studies together show that the amygdala is critically involved in making emotional experiences (both pleasant and unpleasant) memorable for a long time. Amygdala activation due to emotional arousal during encoding highly correlates with subsequent recall (review).
• "The Influences of Emotion on Learning and Memory" (2017) — DOI: 10.3389/fpsyg.2017.01454 — Emotion aids encoding and retrieval, but depending on various factors, it can either enhance or impair learning and long-term memory retention (review).
• "Context reinstatement effects in younger and older adults' memory: A meta-analysis." (2026) — PMID 41525397 — 22 papers (33 studies, 2,177 people). Restoring the encoding context at retrieval improved memory for both younger and older adults (g = 0.32) with no age difference. The effect was weaker when paired items were memorized with associative instructions.
• "Context brings back all the memories: a review of the physical context reinstatement effect on episodic memory." (2026) — PMID 42550362 — Context restoration aids memory when the context is semantically rich, stimuli are unfamiliar, cue burden is low, context and stimuli are bound together during encoding, and the retention interval is long. Methodological differences may be the cause of conflicting results (review).
• "Remote Memory and the Hippocampus: A Constructive Critique." (2019) — PMID 30528612 — The hippocampus is used even when recalling the distant past. It proposes that the hippocampus gathers consolidated neocortical pieces without their original traces to reconstruct them as a scene.
• "GENESIS: A Generative model of Episodic-Semantic Interaction." (2026) — PMID 41794201 — A computational model reproduces gist-based distortions and systematic distortions caused by semantic processing, explaining memory as an active and constructive process.
• "The secret life of memories." (2006) — PMID 16675390 — Challenging the view that memories become permanent through a single consolidation event, it discusses evidence that memories undergo "reconsolidation" after being recalled (commentary).
• "Systems-level reconsolidation: reengagement of the hippocampus with memory reactivation." (2002) — PMID 12408838 — Even memories that have completed consolidation can re-engage the hippocampus when recalled. Two studies in the same issue reach different conclusions regarding the modifiability of old memories (commentary).