한국어

Does the Memory Palace Work? What the Studies Show

At a glance
  • The memory palace (method of loci) works. A meta-analysis of 13 randomized trials found a medium effect (g = 0.65).
  • After 30 minutes a day for six weeks, people went from recalling 25 to 62 of 72 words.
  • Four months later they still recalled 50 after studying a list again. The skill stayed.
  • The evidence is mostly for word lists. For understanding concepts, it is thin.
  • Keep lists in the palace, and walk it again with gaps in between, or it fades too.

The memory palace works. A meta-analysis of 13 randomized trials found a medium-sized benefit, and people who practiced it 30 minutes a day for six weeks went from recalling 25 to 62 words out of 72. But what was tested was mostly lists of words. For learning that depends on understanding, the evidence is thin, and what you put in a palace still has to be recalled again to last.

What is the memory palace technique?

You place the things you want to remember along a route through a place you know well, then walk that route in your mind to get them back. Researchers call it the method of loci.

Say your shopping list is milk, eggs, apples. You picture milk spilled across your front door, eggs cracking on the shoe rack, apples rolling over the sofa. To recall, you walk in from the front door. The places hold the order, and each scene is a cue. The method has been used for thousands of years, and many successful memory athletes rely on it (Twomey & Kroneisen 2021).

Does the memory palace really work?

Yes. Pooled across randomized trials, the effect is medium-sized.

Twomey and Kroneisen (2021) analyzed 13 randomized controlled trials that compared the method of loci with other conditions, most of them at universities. The effect size was g = 0.65 (95% CI 0.45 to 0.85), and it held at similar levels after adjusting for publication bias, dropping studies one at a time, and splitting by setting or control condition.

They also flagged a weakness. Most studies did not report how participants were randomized or how allocation was concealed, so the risk of bias was high. They called it good initial support, not a settled answer.

How much can six weeks of practice change?

A lot. People who practiced 30 minutes a day for six weeks more than doubled the words they recalled out of 72, and the skill was still there four months later.

Wagner and colleagues (2021) split 50 men with no mnemonic experience into three groups. One trained the method of loci, one did working memory training for the same time (40 sessions of 30 minutes), and one did nothing. Before and after the six weeks, everyone studied 72 concrete nouns and wrote down all they could about 20 minutes after leaving the scanner where they had studied them. Four months on, they came back once more.

Only the group that practiced the memory palace recalled far more words

  • Memory palace
  • Working memory
  • No training
018365472Memory palace · Before · 25.225.2Working memory · Before · 30.730.7No training · Before · 28.928.9BeforeMemory palace · After six weeks · 62.262.2Working memory · After six weeks · 41.741.7No training · After six weeks · 36.436.4Aftersix weeksMemory palace · Four months later · 50.350.3Working memory · Four months later · 30.430.4No training · Four months later · 27.427.4Four monthslaterSimilar at the start
Mean words recalled out of 72, in any order, written down about 20 minutes after the scan session in which they were studied. At four months, people studied the first-day list again and recalled it 15 minutes later. Training was 40 sessions of 30 minutes. Memory palace 17, working memory 16, no training 17, all men (45 of the 50 at four months).Source: Wagner et al. (2021), Table 2
See the numbers
Memory palaceWorking memoryNo training
Before25.230.728.9
After six weeks62.241.736.4
Four months later50.330.427.4

The memory palace group went from 25.2 to 62.2 words. The working memory group went from 30.7 to 41.7, and the group that did nothing from 28.9 to 36.4. The memory palace group's gain was significantly larger than both (Dresler et al. 2017). At four months, the memory palace group recalled 50.3 words, the other two 30.4 and 27.4.

How you read "50 words at four months" matters. It is not memory for words studied four months earlier. It is the score when people studied a list again, four months later. What this shows lasting is the skill. Memory for the lists learned during the study was not tested that late. The authors noted as a limitation that the retest reused the first-day list, but that was true for all three groups.

The skill has a cost. Forty sessions of 30 minutes is 20 hours (our calculation). And the practice itself was mostly memorizing lists of random words (Dresler et al. 2017).

Do memories from a memory palace last?

More of their words were still there the next day. Some were still lost.

The same study asked for the words twice after training, 20 minutes later and 24 hours later by phone. The rise in words still recalled at 24 hours was significantly larger in the memory palace group than in either other group. The change in words lost overnight did not differ significantly between groups. The authors concluded that the method boosted durable memories rather than ones that fade quickly.

After training, the memory palace group still recalled more words the next day

  • After 20 minutes
  • After 24 hours
018365472After 20 minutes · Memory palace · 62.262.2After 24 hours · Memory palace · 56.256.2MemorypalaceAfter 20 minutes · Working memory · 41.741.7After 24 hours · Working memory · 30.530.5WorkingmemoryAfter 20 minutes · No training · 36.436.4After 24 hours · No training · 21.421.4Notraining56 vs 21 the next day
Mean words recalled out of 72 after training, at 20 minutes and at 24 hours (by phone). Words lost overnight were 6.0, 11.2 and 15.0 after training and 9.1, 11.3 and 10.4 before training (our calculation). The study found no significant group difference in how overnight loss changed.Source: Wagner et al. (2021), Table 2
See the numbers
After 20 minutesAfter 24 hours
Memory palace62.256.2
Working memory41.730.5
No training36.421.4

After training, the memory palace group lost 6.0 words overnight and the untrained group 15.0. Before training, the same groups lost 9.1 and 10.4, and the study found no significant group difference in how this loss changed. The memory palace group did not lose zero, and the study never tested the post-training list again after 24 hours. To carry a palace list to exam day, you recall it again with gaps in between, like any other memory. When to come back is laid out in the note on spaced repetition schedules.

Are memory champions born with better memories?

In one brain-imaging study, it was the method, not intellect or brain structure.

Maguire and colleagues (2003) studied people known for memory feats at events such as the World Memory Championships, using neuropsychological tests and brain imaging. Their superior memory did not come from exceptional intellect or differences in brain structure. They used a spatial learning strategy that engaged areas like the hippocampus, which matter for spatial memory.

In one line 17 of the world's top memory athletes recalled 70.8 of 72 words on average, and their matched controls 39.9 (Dresler et al. 2017). On the same task, people who had practiced for six weeks recalled 62.2 (Wagner et al. 2021).

Is the memory palace good for studying?

It is useful for lists. For learning that depends on understanding, the evidence is thin.

Most of what people memorized in these experiments were lists of concrete words. Wagner's team chose its tasks partly because they are common at memory championships and partly because they thought the method's strength lies in learning ordered sequences. Its main test, though, counted words recalled in any order. There is classroom research too. Qureshi and colleagues (2014) taught 78 second-year medical students at one college about insulin and diabetes. After two lectures, 28 randomly chosen students learned three memory palaces that two fellow students had built under a teacher's supervision, over an introduction and three sessions of 60 to 90 minutes. The rest of the class did a supervised worksheet session. On a 10-question multiple-choice quiz, the memory palace group averaged 9.31 and the others 8.10 (p = .003). But the memory palace group also got more class time, and it was one unit at one school.

A related imagery mnemonic, the keyword mnemonic (remembering a foreign word through a picture of a similar-sounding word), was rated low utility by Dunlosky and colleagues (2013). It is hard to use in some settings, and its benefits showed up for a limited range of materials and over short retention intervals.

What you need to learnFit for a memory palace
Lists of concrete wordsBest tested
Ordered steps, names, termsPlausible, little direct evidence
Principles, the whyThin evidence
How to solve problemsThin evidence

For material you need to understand, thinking about meaning and tying it to what you already know has a stronger record. That is the subject of the note on moving information to long-term memory.

A version you can try today

  1. Pick 10 spots in your home. The front door, the shoe rack, the sofa: places you always pass in the same order. Close your eyes and walk them once.
  2. Pick 10 items from a list you need. Ordered ones work best, such as the stages of cellular respiration or the order of a talk.
  3. Make one scene per spot. Big, moving, and strange works best. A mitochondrion cranking a generator on your shoe rack.
  4. Cover the list and walk the route. Look back only where you get stuck, and fix that scene.
  5. Walk it again tomorrow and a few days later. A palace you walk only once fades too.

In Aloom

A memory palace hangs new things on a structure you already know, then follows that structure to get them back. In Aloom, the structure is your own knowledge graph. You ask AI, keep only the answers worth keeping, and each one links to what you already know. Each day you recall 1, 3, or 5 of them, as a question first, before you see the answer. The more you recall a node, the darker its ink. Aloom has no places or images to memorize with, so the two can work side by side: lists in the palace, answers you want to understand in the graph. See how Aloom works

FAQ

How long does it take to learn the memory palace?

You can store a 10-item list in 10 spots today. The kind of gain in the study, more than doubling recall of 72 words, came after 30 minutes a day for six weeks, 20 hours in all.

Does the memory palace work for everyone?

The memory athletes studied did not have exceptional intellect or unusual brain structure, and people with no mnemonic experience improved a lot in six weeks. That training study, though, tested 50 male university students aged 18 to 30.

What are the downsides of the memory palace?

It takes practice, and the proven benefit is mostly for word lists. Evidence that it helps with understanding principles or solving problems is thin. And what you put in a palace still fades unless you recall it again.

References