Working memory is waaaay more critical than you might think to all levels of functionality. There are many basic tasks, like walking to another room to get something and noticing something minor, like a pen on a table that should be put away, and doing both tasks, that depend on working memory. The same with mentaly reasoning through a complex system. The reason abstractions are so valuable is that they allow for compression of something into working memory.
For me, personally, this is why I often approach things by scaffolding them into relationships with existing structures (mentally) - by integrating with an existing structure, I avoid a sort of fragmentation overload in my working memory.
Anyway, I think it's one of those things you don't really notice until it goes bad somehow.
Multiple small kids are incredibly disruptive to this. Just, a continuous "happening", constant out-of-context asks and "situations". 5 different things can happen between noticing or thinking of a task and being able to do anything about it. God help you if you have to go from one room to another because that in itself requires explanations (the best case is they quietly follow you to find out whats going on).
There are long stretches of my day where functionally, I have no free working memory at all. The main way I stay barely functional is by keeping memory "in the state of the world". The way I remember I promised to fix the tap today is by placing the tool kit prominently next to the tap, etc. As a last resort I try to write things down.
> The reason abstractions are so valuable is that they allow for compression of something into working memory.
I wonder if people with smaller working memory necessarily have more organized mental models, to facilitate the compression. People with autism [1] and ADHD [2] tend to have decreased working memory. Are their mental models more "optimized" for compression?
ADHD makes a mockery of working memory. The number of times I'll have to go back to see what the the fourth, fifth and sixth digit of a six digit sequence were is truly frustrating.
The article indicates that working memory can be improved though, going to have to give that a try.
Honestly, I don't find I have much of a problem with working memory. Sure, my ability to recall a meaningless number several minutes later is absolutely terrible, but a handling an analysis or conversation about a complicated subject (complex is a different matter that relies on raw intelligence more than working memory) with lots of important detail seems to be easier for me, to the point where other people tap out with "information overload" when I'm moving along just fine.
Of course, executing a complicated process is a whole different matter, because the ADHD brain quickly loses interest and focuses on something else.
I took a properly administered IQ test as part of my ADHD diagnosis. It was eye opening.
All through the test I felt like I was crushing it. Spacial reasoning, pattern recognition, memory tasks. When the results came back I got 135 on spatial reasoning but 89 on processing speed and working memory.
Looking back on my life I realize I had always made up for limited working memory with systems, mnemonics and other techniques. When you've lived your whole life with a limitation you can have a huge blind spot. You've never known what it's like to have "normal" working memory.
Well you're not just noticing the problem or you don't have ADHD because the working memory impairment is well documented and it's essentially THE symptom of ADHD.
What you say about being more easily distracted is a side effect of impaired working memory for example.
Regarding unconscious thinking: I've known for many years that if I'm trying to implement something and what I'm doing just feels wrong and I'm not sure why, it's time to stop and come back first thing in the morning. Sleeping on it engages my back-brain; and invariably the next morning everything makes sense, sometimes immediately, and sometimes with just a small amount of work.
Mind you, the solution I have in mind when I wake isn't necessarily the right one; but I get to the right one pretty quickly.
I get this even on a shorter timescale with compiler errors. So many times I have a near-subconscious feeling that something's wrong with the code, and as soon as I run or compile it, there it is, a syntax error. I've been trying to train myself to pay more attention to these gut feelings, so I can act on them before the compiler gets to it.
Time, doorways, and ritualized actions like sweeping, brushing your teeth, washing your hair, or walking tend to trigger those connections. Even going for a walk might set you right.
That’s usually when I would go for coffee when I worked in an office.
> "The growing trend, especially among young people, to multi-task may seem wonderful. But actually, multi-tasking is most likely to interfere with focused attention and, in turn, degrade memory formation, recall, and thinking quality."
Eventually I realized that parallelization is not really possible, you end up making a mess of everything, and trying to be a rapid context-switcher - similar to the illusion of simultaneous multitasking on a single CPU core - just takes too much energy and time - 15-30 min to unload, clear the slate and reload with something else seems common.
Practically, this is why people working on difficult problems that require their full attention get really irritated by interruptions, and often prefer to work in isolation or only with like-minded individuals.+
From what I've read, the training isn't necessarily transferable. You just get better at these sorts of brain games, which doesn't necessarily mean your working memory is increasing.
Even while reading gwern's blog that seemed pretty positive of this kind of training, there was limited evidence that you shouldn't learn just a new instrument or language or new sort of math discipline.
Why do people keep thinking that "training" can improve working memory?
I agree that if you want to get better at specific things you should just do those things.
Usually when you "learn" something you improve your understanding of the domain, you start chunking things up into patterns and structures. This reduces your mental load and lets you use your "working memory" more effectively.
I think the intuition with say, "n-back" is that there's supposed to be no structure beyond the memory task, so any increase in performance _must_ be an improvement in some sort of generalised "working memory".
As I understand it people have shown that there is "transferance" between these various types of working-memory based brain games (i.e, getting good at one can improve your performance on others that you haven't done before). But no one has shown that getting good at (say) dual n-back produces a strong improvement in "real tasks" that aren't just memory games.
I can attest to the benefits of n-back. I've been doing it for a couple of years now, five days a week for 20-25 minutes. I've noticed a tangible positive difference in both my verbal fluency and my processing speed on days where I engage this protocol.
The evidence is mixed, but some studies (e.g. Jaeggi) did find transfer effects from n back training to fluid intelligence.
It only takes 40 mins a day for 8 weeks to test it out. Much less time than the commitment to learn a new language.
Having tried it, I wouldn't be surprised if the mixed results were due to improper adherence and misunderstanding of how n back works by some study participants. In other words, I think it's possible that results would be less mixed for someone who is already starting from a point of solid intelligence and who is driven enough to put in the hard,focused work to get to higher n back levels.
Gwern has a lot of research and meta-research about this (if memory serves, hah), and in general I seem to recall that doing brain games like dual-n-back makes you better at brain games like dual-n-back.
Concur seeing same research, a thought in the back of my head - does learning one of these games makes you faster at learning other games? Does that translate at learning faster in areas that would use similar recall patterns. There is some research that when you learn N>5 languages (https://news.mit.edu/2024/mit-study-polyglots-brain-processi...) the brain layout changes. What happens when you practice / master >X memory recall games.
I have found that setting to myself a limited, 30 or 40min window, where i am not allowing myself to do anything else but the task I have to do, is actually a sensible way to trick the brain to work on things. Then i swith context. Like coding for 30min,then filling a visa application
I didn't know any better. I just looked things up. Including variables all the time. intellisense was a god send.
Then I sorted my sleep apnea and Celiac and my memory started working. It's like life on easy mode. I'm not smarter but having that mental scratch pad of memory is awesome. Even little things like getting an auth code and not having to read out each number one by one.
Being a caretaker for someone with Alzheimer’s has been a real eye opener to how critical working memory is. I think a large portion of our personal success in life is contingent on our working memory or the ability to manage it effectively.
"The training task to expand working memory capacity consisted of presenting a digit or a word item for a second, with one-second intervals between items. For example, a sequence might be 5, 8, 4, 7, with one-second intervals between each digit. Test for recall could take the form of “Where in the sequence was the 4?” or “What was the third item?” Thus students had to practice holding the item sequence in working memory. With practice, the trainers increased the number of items from 3 to 8.
After training, researchers tested the children on another working memory task. Scores on this test indicated that working memory correlated with IQ test scores. That is, children with better working memory ability also had higher IQs."
You open up other spaces. Society mostly distracts you with busy work. It’s really psychotic to have someone do anything for 8 hours a day.
Think of it like this. You have other rooms in your house that you don’t know what to do with. You might know how to decorate your first room, maybe even your second, but by the time you are lucky enough to have the fifth room, you won’t really know what to fill it with.
I mean yeah, it gets brought up every time this topic is discussed. I’m not going to try to argue Plato was totally right, but I do think there is some merit to the argument.
At least in my experience, the more stuff I can keep my working memory, the better I can solve problems. Contrast to popular opinion, I actually think memorization is a great learning strategy.
The bit about training working memory to boost IQ is cool, but I wonder how long those effects actually last. Like, do you keep the gains, or does it fade if you stop training?
I must admit I was a little surprised when I first realised this to be true. It's rather unnerving to discover that the most intelligent part of your mind is operating without you knowing about it.
> We know that the subconscious mind is processing information (i.e. “thinking”) all the time, even while we sleep. The evidence for this kind of “sleep learning” is incontrovertible and summarized in my memory improvement book...
No. Just the opposite. A quick search shows majority of the scholarly papers question "sleep learning". Maybe 16 years of new data?
For others: it is supposed to be a website promoting the article author's book, but since it is from (2008), the link is now to an Indonesian slots website.
The referencesd book is “Thank You Brain, for All You Remember: What You Forgot Was My Fault” by William Robert Klemm, Benecton Press, 2004, ISBN 0975522507.
Absolutely. I find having a notebook open when writing/debugging code is a godsend. Just keeping a scratchpad of design notes/ideas/tests while I work let's me keep a lot more in my head. Think of it as increasing your level 1 cache. :)
A notebook is how you compensate for hitting the limit of working memory, and everyone can easily hit their limit. This is especially the case in software, where it is trivial to write code thats inter-leavings cause working memory overflow. That’s why good abstractions are so crucial; the human brain has a teeny tiny stack.
That’s not what this is about though, it’d still be nice if possible, to increase the stack size. The following is a link to “Simple Made Easy”, by Rich Hickey; which is the best talk I’ve ever seen on complexity in software and how to manage it. https://youtu.be/SxdOUGdseq4?feature=shared
Working memory is waaaay more critical than you might think to all levels of functionality. There are many basic tasks, like walking to another room to get something and noticing something minor, like a pen on a table that should be put away, and doing both tasks, that depend on working memory. The same with mentaly reasoning through a complex system. The reason abstractions are so valuable is that they allow for compression of something into working memory.
For me, personally, this is why I often approach things by scaffolding them into relationships with existing structures (mentally) - by integrating with an existing structure, I avoid a sort of fragmentation overload in my working memory.
Anyway, I think it's one of those things you don't really notice until it goes bad somehow.
Multiple small kids are incredibly disruptive to this. Just, a continuous "happening", constant out-of-context asks and "situations". 5 different things can happen between noticing or thinking of a task and being able to do anything about it. God help you if you have to go from one room to another because that in itself requires explanations (the best case is they quietly follow you to find out whats going on).
There are long stretches of my day where functionally, I have no free working memory at all. The main way I stay barely functional is by keeping memory "in the state of the world". The way I remember I promised to fix the tap today is by placing the tool kit prominently next to the tap, etc. As a last resort I try to write things down.
> keeping memory "in the state of the world"
This is what I do as well. If things aren't exactly where I left them, I'll not only never find them, but I'm likely to forget they ever existed.
2 replies →
Checklists are the only way I survive. Todos yes, but also basic checklists of "don't forgets"
“Don’t fuck with my mise” (Anthony Bourdain’s paraphrase of a line chef co-worker)
mise is kitchen French for like the IDE of a cook’s station
1 reply →
Working memory being terrible is one of the biggest issues I experience with my ADHD.
I forgot why I went somewhere, or worse I do something different upon arrival.
> The reason abstractions are so valuable is that they allow for compression of something into working memory.
I wonder if people with smaller working memory necessarily have more organized mental models, to facilitate the compression. People with autism [1] and ADHD [2] tend to have decreased working memory. Are their mental models more "optimized" for compression?
[1] https://pmc.ncbi.nlm.nih.gov/articles/PMC7071553/
[2] https://pmc.ncbi.nlm.nih.gov/articles/PMC7483636/
...or do they just perform worse on these memory-intensive tasks? The latter seems more likely to be common.
1 reply →
> Anyway, I think it's one of those things you don't really notice until it goes bad somehow.
So true!
Yeah, totally agree, working memory is basically the brain's RAM, and when it's running low, everything slows down.
ADHD makes a mockery of working memory. The number of times I'll have to go back to see what the the fourth, fifth and sixth digit of a six digit sequence were is truly frustrating.
The article indicates that working memory can be improved though, going to have to give that a try.
> ADHD makes a mockery of working memory.
Honestly, I don't find I have much of a problem with working memory. Sure, my ability to recall a meaningless number several minutes later is absolutely terrible, but a handling an analysis or conversation about a complicated subject (complex is a different matter that relies on raw intelligence more than working memory) with lots of important detail seems to be easier for me, to the point where other people tap out with "information overload" when I'm moving along just fine.
Of course, executing a complicated process is a whole different matter, because the ADHD brain quickly loses interest and focuses on something else.
I took a properly administered IQ test as part of my ADHD diagnosis. It was eye opening.
All through the test I felt like I was crushing it. Spacial reasoning, pattern recognition, memory tasks. When the results came back I got 135 on spatial reasoning but 89 on processing speed and working memory.
Looking back on my life I realize I had always made up for limited working memory with systems, mnemonics and other techniques. When you've lived your whole life with a limitation you can have a huge blind spot. You've never known what it's like to have "normal" working memory.
1 reply →
Well you're not just noticing the problem or you don't have ADHD because the working memory impairment is well documented and it's essentially THE symptom of ADHD.
What you say about being more easily distracted is a side effect of impaired working memory for example.
10 replies →
Might still be worth trying, though!
Regarding unconscious thinking: I've known for many years that if I'm trying to implement something and what I'm doing just feels wrong and I'm not sure why, it's time to stop and come back first thing in the morning. Sleeping on it engages my back-brain; and invariably the next morning everything makes sense, sometimes immediately, and sometimes with just a small amount of work.
Mind you, the solution I have in mind when I wake isn't necessarily the right one; but I get to the right one pretty quickly.
I get this even on a shorter timescale with compiler errors. So many times I have a near-subconscious feeling that something's wrong with the code, and as soon as I run or compile it, there it is, a syntax error. I've been trying to train myself to pay more attention to these gut feelings, so I can act on them before the compiler gets to it.
If you use a modern IDE it will just tell you immediately
1 reply →
Time, doorways, and ritualized actions like sweeping, brushing your teeth, washing your hair, or walking tend to trigger those connections. Even going for a walk might set you right.
That’s usually when I would go for coffee when I worked in an office.
Yup. Also going for a drive, taking a shower, etc.
Ever tried deliberately priming your brain before bed? Like reviewing the problem briefly before sleeping?
Almost daily. For years.
I fell into this trap for a while:
> "The growing trend, especially among young people, to multi-task may seem wonderful. But actually, multi-tasking is most likely to interfere with focused attention and, in turn, degrade memory formation, recall, and thinking quality."
Eventually I realized that parallelization is not really possible, you end up making a mess of everything, and trying to be a rapid context-switcher - similar to the illusion of simultaneous multitasking on a single CPU core - just takes too much energy and time - 15-30 min to unload, clear the slate and reload with something else seems common.
Practically, this is why people working on difficult problems that require their full attention get really irritated by interruptions, and often prefer to work in isolation or only with like-minded individuals.+
I love multitasking - I feel like I'm doing more than 100% and I like being productive.
The trick is to pick a combinations that works:
* listening to a language lesson when cycling (learning+sport)
* repeating flash cards in a bus, instead of doomscrolling (commute+learning)
* listening to a language lesson when cycling to work (learning+sport+commute - whoa!)
* thinking about my programming project when cleaning my home (work+brainless menial work)
In most cases this involves something that doesn't require to much too much conscious attention and something that does.
I'm what you described 100%. I wonder if there's a different type of multitasking term that describes this because I swear it's a thing
Multitasking two coding problems at once completely doesn't work for me but what you described works and I do that all the time.
There seems to be a language disconnect for the type of multitasking that works vs. doesn't work.
1 reply →
Constantly ctrl-z just to see what I was doing a second ago.
ADHD is a blessing and a curse. I can hold every line of the codebase in my head but I can't remember what I was just doing...
You figure out how to work without working memory. Just offload it all immediately.
Treat reality as working memory substrate
Especially on home projects, I push commits frequently as a second layer of Ctrl-Z.
Every time I have googled for it, the only method of working memory training that comes up is N-Back.
So I'm happy that the article mentions another method. Apart from playing "Simon" (yes that circular game with lights), those are the only two I know.
Anyone know of any other methods?
I wanted to look at one of the links in that page about memory training in adults.
It's ironic that the page didn't exist and I had to go to a backup from 2014 in archive.org to find it.
https://sharpbrains.com/blog/2008/04/22/working-memory-train...
https://web.archive.org/web/20140713110043/http://sharpbrain...
I am a bit confused by this approach.
From what I've read, the training isn't necessarily transferable. You just get better at these sorts of brain games, which doesn't necessarily mean your working memory is increasing.
Even while reading gwern's blog that seemed pretty positive of this kind of training, there was limited evidence that you shouldn't learn just a new instrument or language or new sort of math discipline.
Why do people keep thinking that "training" can improve working memory?
I agree that if you want to get better at specific things you should just do those things.
Usually when you "learn" something you improve your understanding of the domain, you start chunking things up into patterns and structures. This reduces your mental load and lets you use your "working memory" more effectively.
I think the intuition with say, "n-back" is that there's supposed to be no structure beyond the memory task, so any increase in performance _must_ be an improvement in some sort of generalised "working memory".
As I understand it people have shown that there is "transferance" between these various types of working-memory based brain games (i.e, getting good at one can improve your performance on others that you haven't done before). But no one has shown that getting good at (say) dual n-back produces a strong improvement in "real tasks" that aren't just memory games.
I can attest to the benefits of n-back. I've been doing it for a couple of years now, five days a week for 20-25 minutes. I've noticed a tangible positive difference in both my verbal fluency and my processing speed on days where I engage this protocol.
I've benefited so much from this protocol that I created a mini app just for myself (https://mind-workout.pages.dev/)* as I was unable to find a suitable app for mac. [On windows there 's Brain Workshop (https://brainworkshop.sourceforge.net/)
*This is a variation of the n-back task called triple n-back, which is much harder than dual n-back.
The evidence is mixed, but some studies (e.g. Jaeggi) did find transfer effects from n back training to fluid intelligence.
It only takes 40 mins a day for 8 weeks to test it out. Much less time than the commitment to learn a new language.
Having tried it, I wouldn't be surprised if the mixed results were due to improper adherence and misunderstanding of how n back works by some study participants. In other words, I think it's possible that results would be less mixed for someone who is already starting from a point of solid intelligence and who is driven enough to put in the hard,focused work to get to higher n back levels.
Yeah, I've always heard about N-Back too, but I've never actually tried it. I had no idea "Simon" was considered memory training
Gwern has a lot of research and meta-research about this (if memory serves, hah), and in general I seem to recall that doing brain games like dual-n-back makes you better at brain games like dual-n-back.
Which is perhaps not without merit, but...
Concur seeing same research, a thought in the back of my head - does learning one of these games makes you faster at learning other games? Does that translate at learning faster in areas that would use similar recall patterns. There is some research that when you learn N>5 languages (https://news.mit.edu/2024/mit-study-polyglots-brain-processi...) the brain layout changes. What happens when you practice / master >X memory recall games.
Link to Gwern's article: Dual n-Back FAQ - https://gwern.net/dnb-faq
I have found that setting to myself a limited, 30 or 40min window, where i am not allowing myself to do anything else but the task I have to do, is actually a sensible way to trick the brain to work on things. Then i swith context. Like coding for 30min,then filling a visa application
Reminds me of the "Hammock Driven Development" talk from Rich Hickey:
https://www.youtube.com/watch?v=f84n5oFoZBc&t=1270s
>Imagine dialing a phone number by having to look up each digit one at a time in the phone book
This was me for the first 38 years of my life. I got a degree and worked as a software developer with memory that bad.
How did you manage it?
I didn't know any better. I just looked things up. Including variables all the time. intellisense was a god send.
Then I sorted my sleep apnea and Celiac and my memory started working. It's like life on easy mode. I'm not smarter but having that mental scratch pad of memory is awesome. Even little things like getting an auth code and not having to read out each number one by one.
Being a caretaker for someone with Alzheimer’s has been a real eye opener to how critical working memory is. I think a large portion of our personal success in life is contingent on our working memory or the ability to manage it effectively.
"The training task to expand working memory capacity consisted of presenting a digit or a word item for a second, with one-second intervals between items. For example, a sequence might be 5, 8, 4, 7, with one-second intervals between each digit. Test for recall could take the form of “Where in the sequence was the 4?” or “What was the third item?” Thus students had to practice holding the item sequence in working memory. With practice, the trainers increased the number of items from 3 to 8.
After training, researchers tested the children on another working memory task. Scores on this test indicated that working memory correlated with IQ test scores. That is, children with better working memory ability also had higher IQs."
I actually think this is a danger with leaning more heavily on AI tools. What happens as you offload more of your thinking outside of your brain?
You open up other spaces. Society mostly distracts you with busy work. It’s really psychotic to have someone do anything for 8 hours a day.
Think of it like this. You have other rooms in your house that you don’t know what to do with. You might know how to decorate your first room, maybe even your second, but by the time you are lucky enough to have the fifth room, you won’t really know what to fill it with.
Luxury. It takes some imagination to fill it.
Ever hear of the curse of Thoth in relation to Plato on the merit of writing?
I mean yeah, it gets brought up every time this topic is discussed. I’m not going to try to argue Plato was totally right, but I do think there is some merit to the argument.
At least in my experience, the more stuff I can keep my working memory, the better I can solve problems. Contrast to popular opinion, I actually think memorization is a great learning strategy.
1 reply →
> when only four attributes were involved, subjects in the conscious-thought condition made the best choice of car.
> But when the complex condition of 12 attributes, results reversed. The best car was chosen most reliably in the unconscious-thought condition.
I failed to understand this.
The bit about training working memory to boost IQ is cool, but I wonder how long those effects actually last. Like, do you keep the gains, or does it fade if you stop training?
I must admit I was a little surprised when I first realised this to be true. It's rather unnerving to discover that the most intelligent part of your mind is operating without you knowing about it.
I wonder how these results would change if working memory was augmented, e.g. by giving participants a notepad.
I came here wondering if anybody was going to comment on the recent post that says that smoking marijuana reduces working memory
> We know that the subconscious mind is processing information (i.e. “thinking”) all the time, even while we sleep. The evidence for this kind of “sleep learning” is incontrovertible and summarized in my memory improvement book...
No. Just the opposite. A quick search shows majority of the scholarly papers question "sleep learning". Maybe 16 years of new data?
e.g., https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1365-2869....
and, as @joeyagreco noted the link is (might as well be) dead.
> see http://thankyoubrain.com
oh... oh god
For others: it is supposed to be a website promoting the article author's book, but since it is from (2008), the link is now to an Indonesian slots website.
The referencesd book is “Thank You Brain, for All You Remember: What You Forgot Was My Fault” by William Robert Klemm, Benecton Press, 2004, ISBN 0975522507.
Can working memory actually be improved as this article suggests? Links to apps to change my life please.
Absolutely. I find having a notebook open when writing/debugging code is a godsend. Just keeping a scratchpad of design notes/ideas/tests while I work let's me keep a lot more in my head. Think of it as increasing your level 1 cache. :)
A notebook is how you compensate for hitting the limit of working memory, and everyone can easily hit their limit. This is especially the case in software, where it is trivial to write code thats inter-leavings cause working memory overflow. That’s why good abstractions are so crucial; the human brain has a teeny tiny stack.
That’s not what this is about though, it’d still be nice if possible, to increase the stack size. The following is a link to “Simple Made Easy”, by Rich Hickey; which is the best talk I’ve ever seen on complexity in software and how to manage it. https://youtu.be/SxdOUGdseq4?feature=shared
See: https://gwern.net/dnb-faq#n-back-improves-working-memory and https://brainscale.net/
[flagged]