Train Your Brain Like a Memory Champion | Hacker News Reader

Train Your Brain Like a Memory Champion

nytimes.com

175 pointsprostoalex49 comments

There's a bigger story hidden behind the achievements of memory champions. The groundbreaking result from the past few decades is not that there is a way for a human to memorize a thousand numbers, or that a particular method of memorization (say, memory palace) is the way to do it. It's nice, but not that relevant for everyday life and not applicable to most life problems.

The true achievement is recognition and verification, across many disciplines, that: 1) there are better and worse ways to improve skill 2) anyone can improve any skill [if they do it the right way] 3) mental representations are key to high performance. Some are better than others. 4) one should study how to learn a given skill, or get someone who knows that to teach, e.g. to study people who are best at something - to find out their training regimen, their way of structuring the information/skill/work/memory/... - this is likely to work everywhere 5) If that doesn't help, i.e. little progress is made, one can still figure out, discover, create their own ways of practice to advance. 6) the upper limit of skill is way way higher thank we think 7) Many fields don't have clear criteria for success, so little feedback is available on low-level performance details, which limits the progress of training methods.

If you're interested in all that, I recommend reading the book of a renowned scientist who actually discovered a lot of this stuff and who worked with early memory champions, provoking them to push the boundaries of what was thought possible -- Anders Ericsson, "Peak: Secrets from New Science of Expertise".

(3) is fascinating and begs the question of how we discover these representations. I imagine - like science - it's a bit of theory, a bit of empiricism. Constantly tinkering with our theories and our data to evaluate if new models improve and reliably predict performance.

(7) is also interesting. Ambiguous success criteria is one problem. I'd also include delayed feedback and multivariate feedback loops. The longer the feedback loop, the more likely confounds can enter into the training regimen, rendering it harder to discern whether the new training ("treatment") actually has a significant effect. This is a problem in all fields: when we are trying to measure long-term results, it gets very difficult to decide whether any proposed intervention (new diets, new exercise regimens, new marketing campaigns) have any effect.

3 sounds to me like mental models. I'm fascinated by the idea of how we can learn/teach using mental models directly, rather than expecting the learner to build the mental model over time by working through rote memorisation and experimentation until it "clicks". I strongly believe that opening up mental models would make learning new skills vastly quicker and more accurate.

What a coincidence!

https://alexpetralia.github.io/newsletters

> At your left, there’s a map of Minnesota, dangling precariously from the wall. You’re certain it wasn’t there this morning. Below it, you find a plush M&M candy.

> If none of this makes sense, stick with us; by the end of this piece you’ll be using the same techniques to memorize just about anything you’ve ever wanted to remember.

No, it won't ever make sense and it may continue to come as a surprise to many that not all of us have the ability to form visual based imagery in our minds. A common term for this is Aphantasia.

Not everyone thinks the same way. Any attempt at mass producing some means to know something better/faster is probably not going to work on a subset of the population.

I was able to use the Memory Book's number to letter technique a few years ago to memorize short lists of objects, but my recall for that is pretty good anyway so it's not worth the time taken to memorize stuff using the process (which itself requires memorizing certain objects for the numbers). For those who visualize, I would imagine such techniques could be quite useful.

I've had stronger and weaker Aphantasia at different points in my development and in different learning/work situations.

Given that styles of learning has largely been debunked, I would suspect that the vast majority of us actually have roughly the same capabilities and they are just unexercised, exercized to fitness or from too much stimulus to exhaustion by our specific environments, diets, motivation levels, etc.

After I read “Moonwalking With Einstein”, I pretty much think it’s a gimmick and still a lot of work. Spaced repetition is probably a better technique.

The main idea that seemed to work is the visual association. The visual of Moonwalking with Einstein, for example, was one of his mneumonics.

It seems hard to apply spaced repetition to numeric information or names of people you've just met at a party.

> It seems hard to apply spaced repetition to numeric information or names of people you've just met at a party.

Spaced repetition is one of the only ways to learn names in my experience. If I learn someone’s name and then don’t see them again for 2 weeks, chances are I will have forgotten. Even if I see them once a month for a year, there’s a good chance I will forget each time.

If I see them the next day after I learned their name I will remember. If I see them a week after that I will still remember. If I see them a month after that chances are I will still remember.

If I had first e.g. read the person’s academic paper or a few of their blog posts or the like, there’s a good chance putting a face to the name once will be sufficient.

* * *

If you can get a photo of the person, then you can work learning their name into some kind of deliberate flash card routine, and that will be more effective overall than relying on chance encounters and less embarrassing than repeatedly asking.

I would recommend that teachers of lecture courses put a few minutes a day into doing spaced repetition of students’ faces/names, starting before the start of the term: teachers who can remember students’ names make a big impression.

Right, spaced repetition isn't good at memorising things quickly. Where it shines is that it lets you decide the things you want to keep in your long term memory.

I see method of loci more as "ram", whereas spaced repetition is moving things into a hard drive. For example, when reading, I'll often use a mental room with 10 spots x 3 key points per spot to keep an outline.

Periodically, I'll go back and walk through the outline, looking up details where I have gaps (ie spaced testing). It's helpful that a spatial path lets you easily keep the temporal structure of the narrative.

One of the things I learned about memory when I was reading tons of papers on this stuff is that usually people need to have comprehension of something before they will remember it. As an example, when I was teaching English to high school students in Japan one time I told the students to bring their textbook the next class (we usually didn't use it). I spoke English. The next day, not one of the 30 or students brought their text book. I asked the students, "Does anyone remember me saying something last class?" Not one student remembered. I had a teacher teaching with me ("team teaching" -- very good idea) and he remembered me saying it. It was then that I really grasped the concept: if you don't understand, you won't remember. In fact, you might not even remember that something you didn't understand had occurred.

I think what visualisation (especially with story telling) does is create a framework for comprehension. It gives logic to the thing you are memorising. This is especially good when you are memorising a list of arbitrary facts that have no inner logic themselves. People who are studying Chinese characters often create stories (ala Hesig) and I can verify that it makes them easier to remember. Without an inner logic, it's really hard to hang on to.

My feeling (which corresponds to my experience) is that when memorising things that already have an inner logic, there is no need to visualise or create stories. For example, if you memorise example sentences rather than vocabulary or grammar, visualisation doesn't help. The logic already exists within the sentence.

> is that when memorising things that already have an inner logic, there is no need to visualize or create stories

That matches closely with what I observed of how it's difficult to memorize prose in a completely foreign language, yet it seems one can recall song lyrics in that language with just one or two listens. It seems that any sort of internal structure can serve as scaffolding and reduce the task from brute-force memorization to some sort of "interpolation."

I don't necessarily agree with this. As a french native who has grown up to english music, I realised years later as my english improve that I 'remembered' the english songs lyrics from previous years by basically 'mapping' the english to french equivalent phonemes, NOT by words. In fact the stuff I 'remembered' made no sense at all, or were quite funny, a posteriori.

I have to admit as a Canadian I only just recently learned what "Terror de nose eye you" meant in the French lyrics of the national anthem ;-) But I think even that creates a kind of weird internal logic that you can remember. That memory must be at least 40 years old for me.

Aside: 13 years of French lessons and I'm hopeless. A big pile of Japanese books and no lessons and I'm fluent. I always found that interesting.

Yeah, I read that book as well and felt the same way.

I think in many ways working memory is the most important thing - sure I can learn these methods to memorize 1000 random facts - but what 1000 random facts will help me in my daily life?

Maybe it's more useful for people doing medicine or something similar.

This. I wanted to make this comment. I was quite interested in improving my both long and short term memory and I came across "Moonwalking with Einstein" book. While the methods such as "menomics" and "memory palace" presented in this book were quite interesting, I learned over a period of time that it doesn't help much.

There is quite a lot of information out there and we need to compress it in someway, intelligently so that we could retrieve it in when the need arises. The important part of the retrieval process, I believe, is getting all the relevant information along with the required ones and I honestly don't see how menomics and memory palace could help.

*mnemonics

there is no peer reviewed study that corroborates the claims made in that book. it's mostly a gimmick. it's like trying to learn how to become taller, which s absurd but not much different from learning how to suddenly become good at memorizing stuff.

So the main tool that memory champions rely on is essentially 'visualizing/picturing' fake situations. Now, I have a solid memory outside of this. But when I picture things, it's pretty hard for me even for most familiar places like my home. Any memory experts with actual advice to see more vibrantly? I feel like I'm always fighting against the natural tendency to see black (as is natural with eyes closed) versus trying to focus on what I'm picturing.

I'm exactly the same! I've recently started to think that I don't actually have a minds eye. I can't hold an image in my head for more than a fraction of second - and even then, it's just parts of the image at a time (not a whole scene, and not in realistic color). The memory-castle type techniques seem to be visually "begging the question" - to remember things, first... remember things!

If you can see it for even a fraction of a second, you do have a minds eye; those with aphantasia can't see anything at all for any length of time. If you can see anything at all, you can improve that with practice.

Have you tried a strategy like method of loci? If not, I would try it first, to see how it goes.

Visualizing is one part of the strategy (depth of processing), but a huge other piece is that you are enlisting spatial cues to help prompt recall (cued recall).

'visualizing' isn't just 'seeing'. Try to think back in your mind to the house in you grew up in. Do you remember where the kitchen, bedrooms, and/or bathrooms were? Do you remember which way the beds/furniture were facing in most of the rooms, which side the sinks and counters were in the bathrooms and kitchens? Do you remember where in the house any other furniture, such as a desk, couch, television, or coat rack was? Most people can remember these things, even if they can't conjure forth before their mind's eye a vivid mental picture of their house.

Try answering all of those questions about, say, any of your neighbors houses growing up which you may have been in once or twice. About a recent home, building or room in which you may have only been in once, sometime within the past 6 months to a year.

While it's not perfect, and some particular facts might elude you, most people will find it surprisingly easy to answer most of these questions, even about buildings they may have only been in once or twice a decade or more prior. Yet, if they were to try to, say, answer detailed questions about a painting they may have seen around the same time, most will struggle.

We seem to have a certain kind architectural/location memory which is used for remembering the relative layouts of places we've been, and this sort of memory seems to have some different properties compared to just visual imagery. It seems to be retained long-term fairly effortlessly, with very little time actually spent 'memorizing' it.

This is the basis of a lot of the tricks used by memory champions: https://en.wikipedia.org/wiki/Method_of_loci

> even if they can't conjure forth before their mind's eye a vivid mental picture of their house

It should be noted that there's also a small percentage of people who physically don't have a "mind's eye":

https://en.wikipedia.org/wiki/Aphantasia

Most people can practice and develop this skill.

You might enjoy reading http://www.amblesideonline.org/CM/vol1complete.html#048

speaking from personal experience I relied heavily on mnemonics memorizing skills during my years in academia.

At the beginning I had time visualizing all those "fake situations" and images. That's OK. You're OK. It gets much easier as you practice.

What helped me in this process is avoiding movies(especially modern action movies) and reading books.

People are different. Each mind is different. Only over 30% of the population is highly visual.

When people talk about "images" in the memorization world, they are not just talking about visual memories. But also touch(we have millions of touch sensors distributed along the skin) "images". Speech images. Spacial 3d "images", "smell images","kinestesic or muscle memory and so on.

A persona that is born blind can't make visual images, a baby with malfunctioning ears will not develop sound images. In fact, you could restore vision or cochlea "sensors" later in life and they will not be able to hear or see because the brain as a baby did not have it and neural connections were not made. This is the reason they test for hearing on babies and operate them as soon as possible.

But that a blind or deaf person could not see or listen, it does not mean that he could not make an "image" of the world, in fact some areas of their brains will be actually stronger than on people that can see or hear.

You have to know thyself, what are your strong areas, and favor those sensory images. I am deeply kinesthetic for example, so I always favor space relationships and movement, actions and so on.

It is not just about visualizing but also the choice of visuals. One of the things that Harry Lorayne stresses on, is to visualize something out of the ordinary, so crazy that it sticks out in the memory (for e.g, to remember a sequence of man and pencil, imagine a man with pencil sticking out his head or a pencil walking like a man on the street). The second best option is to imagine something sensory, tactile or olfactory or anything visceral. Mundane items or events are the ones most likely to escape our complete attention and consequently slip through our memory. Truth is that the attention is the most important bit of memorizing. More often than not the act of processing the information registers it much more effectively than any of these details. Try it for yourself, you will be surprised.

Look up "aphantasia". It's a relatively new (and trendy) term referring to an inability to "see" things in your mind. Apparently, most people can engage the visual cortex and actually form mental images, though I haven't really experienced this myself. I just get brief flashes, but mostly remember things factually.

I've personally found memory palaces and whatnot to be largely useless, probably because I can't visualize anything. Soundalike-based mnemonics work better for me (cockney rhyming slang type stuff, etc.)

If you want to run through some pre-built memory palaces to get a sense of how they work, I recommend you check out https://www.memorypalace.com. They have a platform where you can build them and post them for public viewing. I use memory palaces all the time and it really does take some practice, but they help me store things into long term memory. Part of the trick I feel is that you can review frequently in your head without accessing a book or digital, and so you can repeat it often to yourself so you can make that jump from short term to long term memory.

Thanks that seems interesting. What kind of information you are learning with this method and how exactly you go about doing it? I feel brute force learning is very important before diving deep.

I am trying to evolve my personal framework and I feel us starting with "Why" type questions is counterproductive and frustruating. Trying to ask why is something before even asking what is something is poor use of time.

To understand the complex system, I feel going from what to how to why makes much more sense than asking why questions and just getting overwhelmed.

- What Phase Once we know how things work on mechanical level. Asking what each component is build of and what are its parts.

- How Phase This phase is like taking a leap of faith and comitting the steps to memory. At the end of this phase a big picture of how various components work and hang together.

- Why Phase This phase should be reserved for things we want to deeply understand.

Despite all the media attention about memory training, there's actually little evidence this stuff works.

https://greyenlightenment.com/bullshitting-with-einstein/ Either memory is a direct function or IQ or due to 'savant abilities' and not something that an be replicated. There are no reputable studies controlling for IQ that replicate this.

The author started losing me pretty early in the post, starting from where he implies that IQ is an "innate, biological trait." Notably, he (intentionally?) completely misses the point of a memory palace:

"Using a mnemonic device (such as a ‘memory palace’) still requires one memorize the mnemonic. If I ask you to memorize ten historical dates, a trick may be to associate these dates with a mental visualization, but you must still remember ten associations, which is still not easy."

It sounds like the author has not even once run a self-experiment to test the efficacy of mnemonic devices. It's patently easier to memorize these "associations" that he criticizes than it is to directly memorize an equal number of bland, meaningless digits, for example. Memorizing the mnemonic is a one-time, up-front investment of effort. Once you've written the mnemonic (encoder-decoder machinery) into long-term memory and become fluent in its use, you can fire it up at any time and encode meaningless sequences into highly meaningful and memorable equivalents.

Check out Alex Mullen's site where he explains this in more detail - https://mullenmemory.com/

There's a great app to train your memory: https://play.google.com/store/apps/details?id=com.doggoapps....

Having read Moonwalking with Einstein (which is fascinating and worth the read), I think of these skills as similar to studying chess - effective for the discipline but probably not generally helpful.

Memory experts sold their discipline as if it would radically transform an average person’s daily life, which has not been my experience.

I want to know if these champs can do like 20-back or higher(the n-back trainer) with three or more different types of inputs. If they can achieve 80% success rate, that will be super impressive.

Do such techniques work for mathematical stuff, like equations or formulae?

Sure; you just need a mapping of math symbols to memorable objects/colors/verbs/whatever. Mnemonics are just symbol mapping.

So say I want to remember what Tallagrand's Inequality was https://en.wikipedia.org/wiki/Talagrand%27s_concentration_in... . How would one go about it?

When I was ten years old I was proud to have memorized lots of odd things like the name of the highest peak in Sri Lanka or the capital of Honduras . Now 50+ years on I still remember them because I wanted to remember, but I honestly can say nobody ever has asked me for it . I also learnt 5 languages , because I wanted to be able to communicate or just understand what was written there. I never used mnemonics , I always forget them.

While not exactly an answer to your question, you could go through the process described here: http://cognitivemedium.com/srs-mathematics

Which would probably burn that into your brain pretty well.

Unfortunately, no one can give you a way that would work for you, because that's kind of the point: you adjust your internal representation to your strengths. But here's an example of how I'd map it to a memory palace, based on what I did for my driver's license test.

In my case, I'd walk into the living room of my old house and I "see" that someone wrote "ct24" on the floor. That's all I need to remember that I'm trying to reconstruct that pr[A].pr[A^c_t]<=e-(t^2)/4, because I know I'm reconstructing an inequality, exponents of e are often negative, and I know that I need to parse that text as [][ct][t24] (which are the exponents and sub-indices I need).

I then picture that a guy comes to me and says "Hi, I'm Paxton", and he's wearing a t-shirt with an Omega symbol. With that, I can reconstruct A_t={x in Omega|p(A,x)<=t} (in case you didn't catch it, I map p(A,x)<=t to "Paxton").

And so on. Note that I can take some shortcuts here because I'm playing to my strengths of being used to equations, and therefore I don't need to memorize that the second step defines A_t because it comes naturally.

For a completely different approach, you can picture a kid trying to say "Practice", but he mumbles instead "Pra... Pract...et... 24!", which you can map back to the equation ("Pract" = Pr[a^c_t]). The fact that the "24!" at the end comes out of nowhere only makes the scene more memorable, and therefore easier to remember.

Thank you! I just wanted an example so I could conceptualize what this system was talking about.

They should! This could even be a cool way to apply something like Gödel numbering. [1]

[1]: https://en.wikipedia.org/wiki/G%C3%B6del_numbering

"Others may contain misspellings and factual errors. It doesn’t matter. This system is designed to create rich imagery, not accurate representations." Wait, what?!

That's right. The system isn't a lossless compression algorithm. It's designed to give your brain a bunch of hooks to hang interesting snippets onto, which you can use to reconstruct the short story that unravels into a phone number or whatever. For this reason a misspelling or impossibility is just as a good as any other notable attribute, because it makes the story element more memorable.

Taking the article's examples, you might construct a story like "toilet paper is used as socks in Minnesota." It doesn't matter whether that's true. It only matters that it makes it incredibly easy to reconstruct the number 190732.

Does it work for less well structured Data?

Digits of pi you can translate to characters and twine a story, but loosely associated facts are more tricky.

Reply on news.ycombinator.com