Saturday, October 11, 2008

Memory & Chess

"When the dawn comes, tonight will be a memory too..."

Scientists know what role the parts of our brain participate in, via brain scans and studies of people and animals with partially-malfunctioning brains.
  • Amygdala - Stores emotionally-charged memories and reacts when you re-enter a dangerous situation for the second or third time.
  • Basal Ganglia - Stores skills and habits that we've repeated many, many times such as driving our car, using a spoon, or exercising in the morning. Pros: Anything in procedural memory can be done on auto-pilot and it's hard to forget. Cons: It's hard to forget!
  • Hippocampus - Forms and stores long-term semantic memory, like principles, techniques, and tactical patterns. The longer we know these, the less likely the hippo's involved. It also stores visual-spatial memories, no matter how long we know them.
  • Frontal Cortex - Stores long-term semantic memory. If we know something well, and we know it for months or years, the knowledge will transfer completely from the hippocampus and into the frontal lobes.
A brain scan of players 1700-2600 shows that the stronger the player, the more they use their frontal lobes and the less they use their hippocampus. They remember while we calculate.

What does this mean to me?
  1. The secret to mastery is not in learning to play chess on auto-pilot.
  2. The secret to mastery is not studying drawing of 3d boards.
  3. The secret to mastery is not how many positions we can calculate.
Since most of our chess knowledge is in forgettable, semantic memory we should be selective about what we learn and remember to review it (think: spaced repetition).


likesforests said...

Caveat: I'm not a neuroscientist, but I read a few articles written by neuroscientists then while working late at work Thursday night. ;)

Polly said...

In the book Deep Survival that I discussed a few weeks ago there are some excellent descriptions of the function of those various parts of the brain, and how they impact the ways we learn things, process situations and respond to stress.

There are times I feel like there is a huge hippotumus sitting on top of my hippocampus. Now that I've actually finished the book I want o go back to look closer at some of the brain functions as they relate to stress, and see how that relates to chess induced stressed.

Temposchlucker said...

Does this imply that Magnus Carlson at age 13 has stored more in his semantic memory than an amateur with 40 years of experience in both playing and studying? I simply cannot believe that.

likesforests said...

I really enjoyed "Deep Survival". Compelling stories. And yeah, it definitely piqued my interest about how the brain operates. Although some of it went over my head since I had never studied what these various brain structures do before.

"There are times I feel like there is a huge hippotumus sitting on top of my hippocampus."

I should know what you mean, but only bits and pieces of the parts dealing with the brain sunk in. ;)

Regarding stress in chess, I've only been stressed in two OTB games so far and in both games it led to me making irrational choices. Those games are also my most vivid memories--of course!

Is it generally a bad thing to be stressed during a game? Should we be seeking to remain calm--or does the adrenaline serve a purpose?

If you write such an article, I will be very curious to read it!

likesforests said...

tempo, it's incredible. Fantastic! Studies tend to focus on "average" GMs and masters rather than super-exceptions. It makes more sense to me that his chess skills would be in semantic memory as that takes less time and effort to form than procedural memory, and that's where most GMs/masters have much more going on in the brain scans. But if you're suggesting science doesn't even yet grasp how Carlsen and other super-GMs rise so quickly, then I might agree. :)

* Specific Positions *

I use endgame examples, because it's the one phase of my game that I feel is somewhat mature.

White: e2, Ke1. Black: Ke8. White to move. What is the result?

When I see the picture, my brain marks the key squares d4,e4,f4. Then it counts 1-2-3 steps for the White king to reach there, and in 1-2 steps the Black king cannot stop him in time. I'm certain it's won in less than 1.5s. After this long think, I could play the remaining moves to the key square instantly. This has involved semantic memory, calculation, and short-term memory.

Converting once my K is on a key square is automatic (procedural). But suppose I did not have this advantage and had to think for 1s every few moves. Would it matter OTB, or in the 25+5 online games I play? I don't think so. Having the semantic knowledge is the key. The procedural memory is a bonus.

White: Nc1, Kh8. Black: c2, Kc3. White to move. Who wins? What move?

A 1600 online blitz player posted this position and a long sequence of moves that draw. To me, it's absurd because I recognize (semantic knowledge) on sight this as drawn (specifically, N in front of pawn, not a 7th-rank rook pawn).

In less than 1s (calculation) I considered the 'drawing replies'. Na2+ (semantic knowledge: laterally controlling the promotion square draws), Nc2+, Kg8, Kg7, Kh7. To me they are all equally good because they all preserve the simple draw.

likesforests said...

I should add, that my *habit* of looking at key squares and counting is due to procedural memory, though the actual recognition of key square and counting to them are not.

That is probably one of the best uses of procedural memory in chess--developing good habits that guide both our study and play.

Temposchlucker said...

I will have a better look at your comment here later. In the mean time have a look at an old post of mine.

Temposchlucker said...

It seems that the studies we are referring to aren't consistent with each other. Hence I tend to use the word procedural where you use the word semantic. We might mean the same though.

darkhorse said...

Interesting. As a suggestion, it would be great if you could cite your sources.

Also, I think that you meant to write "Hippocampus: Forms and stores short-term semantic memory" (instead of long-term).

likesforests said...

darkhorse, "Also, I think that you meant to write Hippocampus: Forms and stores short-term semantic memory (instead of long-term)."

I meant exactly what I said. Here are some insights based on observations of a man w/o a hippo:

(1) A person with a missing hippocampus has a functional short-term memory.

(2) A person with a missing hippocampus can recall many long-term semantic memories.

(3) A person with a missing hippocampus can't force new long-term semantic memories.

The hippocampus is definitely pivotal to transferring what is in your short-term memory into your long-term memory.

But its role goes beyond that. There's also strong evidence that while most long-term memory can be accessed without the hippo, it helps in some way, and some LTMs (especially recent ones) cannot be accessed without it. Its role in LTM retrieval is still somewhat mysterious compared to its role in forming new memories.

likesforests said...

tempo, here is a quote from the page where you interpret the study:

QUOTE> "In amateurs, focal gamma bursts were most prominently detected in the medial temporal lobe and in grandmasters researchers measured the bursts most often in their frontal and parietal cortices, parts of the brain linked with long-term memory and the ability to perform complex motor skills."

Here is the study:

The "medial temporal lobe" is where your hippo is. It's involved in converting STM to LTM; ie, the formation of new LTMs. Its also essential to retrieving recently-learned semantic LTMs (past hours or days, but occasionally longer).

"The frontal and parietal cortices" are the frontal lobes I refer to. It stores long-term semantic memories, especially ones that have been known for a long time.

What you said above characterizes the study very well, but then you add, "parts of the brain linked with... the ability to perform complex motorskills".

This is *not* a conclusion of the study. And that does not fit our current understanding of the brain! Procedural memory--habits and motorskills--are stored in a different area of the brain called the brain stem, cerebellum, or basal ganglia. That is the type of memory that you never forget. This study did not show any difference in that area of the brain between GMs and amateurs.

likesforests said...

The semantic memory is a form of "explicit, conscious" m emory.

darkhorse said...

I see. This is a subtle point, which definitely didn't come across in post. You say:

Hippocampus - Forms and stores long-term semantic memory,
Frontal Lobes - Stores long-term semantic memory.

...oh, so both are involved in long-term memory

The longer we know these, the less likely the hippo's involved.

...wait, wasn't the hippocampus for long-term memory?

If we know something well, and we know it for months or years, the knowledge will transfer completely from the hippocampus and into the frontal lobes.

... I get it, hippocampus is for short-term memory and the above definition is a typo.

See what I found confusing? Furthermore, we used to think hippocampus = short-term, which you are disagreeing with (based on recent research -- interesting!)

likesforests said...

Perhaps calling the hippocampus our "medium-term memory" is a bit of an oversimplification but makes getting its role much easier. :)

darkhorse said...

Wikipedia: The hippocampus is a part of the forebrain, located in the medial temporal lobe. It belongs to the limbic system and plays major roles in *short term* memory and spatial navigation.

(Just to show how confusing the hippocampus can be)

likesforests said...

It is confusing. We know some things, but there is much we still do not know and only guess at.

Wikipedia's article on STM says the role Hippocampus plays is to transfer info from STM to semantic and episodic LTM. As a link between the two systems, it could be labeled as a part of either, both, or neither subsystem.

The role in navigation.. aye! I had almost forgotten about that. Thanks for the discussion--it's helped me link some things.

Yes! It's a link between STM and LTM, but it holds all the keys in navigation (visual-spatial memory). The book Polly was talking about "Deep Survival" had some fascinating stuff on how one becomes lost and determines when we are lost, the hippo and its mapping ability being a key player in that scenario.

Glenn Wilson said...

Interesting post.

I was playing a student this past weekend. I was white and it was an Italian Game. We "play and talk"; make moves, try moves and experiment with the positions as we go.

At one point early in the game I was castled, she was not and she played Nf6xPe4. I pointed out that I had not even considered that move and that it "had to be bad."

So I showed her some winning lines for white from that point. But, the tactics were mostly over her head. She would likely not be able to win from that position as white in a real game. So, she "learned" some general stuff and she learned that she should have castled earlier. She can not learn what I know about that position (and similar positions) until her tactics are better.

So, I don't think I agree with your conclusion: The secret to mastery is not calculation. If you can't calculate the position then you can't really understand it to store it, to recall it later.

chesstiger said...

If i am not mistaken the right and left half of the brain has also something to do with all the processes of the brain. One side, i forgot which, controls the emotional side, the other more the technique side like math and such.

likesforests said...

chesstiger, absolutely, the side of the brain that we use is another interesting question! A small study of five chess players in 1995 indicated that if you're right-handed, you use the right hemisphere of your brain more often when playing chess... and if you're left-handed, you use the left hemi. But I can't find a useful training application of this knowledge.

likesforests said...

Glenn, fair enough. I think a more accurate way to state my meaning is, "The secret to mastery is not how many positions you can calculate."

Note, this study is focused on 1700 to 2600 players. For players below that threshold perhaps the conclusions are different.

Consider a position that arose OTB.

My opponent was not able to solve it--it was over their head. But due to my knowledge of patterns (forks and pins) and the principle of checking forcing moves first, I was able to solve it while only calculating a few positions.

A few months back, I gave you a position from a game that I could not solve in time... you actually solved it quickly... again by using patterns and principles to reduce your search radius.

(Another factor in missing tactics is having a good habit of looking for the seeds of tactical destruction, as Heisman likes to call them, on every move!)

When a pattern is over our head, it doesn't necessarily mean we can't calculate deeply enough--it may also mean we do not have the right basic patterns to simplify the solving of the position. As in Phaedrus's post last week! We both had a missing 'simple' pattern. :)

Kalan said...

Very interesting blog... :-)

chesstiger said...

Btw, did you read 'How to think in chess'? ( )

I dont know if it says something about how the brain is involved when playing chess since i didn't read it (although i own a copy that is collecting dust as we speak) but i heard it's not a mainstream book. With this i mean that it isn't a book full of analyzed games but more a study.

Blue Devil Knight said...

I haven't read the comments, but a couple on your taxonomy:

* Brain Stem - Stores skills and habits that we've repeated many, many times.

This is probably better to assign to the basal ganglia, which are the main engines of habit formation (I linked a few papers on this topic in the comments here).

The brain stem tends to do more mundane housekeeping (with some exceptions), such as breathing, waking/sleeping, acting as a relay from the brain to the body (and vice-versa).

* Frontal Lobes - Stores long-term semantic memory.

Typically the frontal lobes are more associated with active working memory (remembering a telephone number someone just told you when you are fumbling with your phone).

As for long-term memory storage, that is still a subject of controversy. It seems quite distributed around the front of the cortex (which includes temporal lobes, frontal lobes). But it also may even be stored in the sensory cortices in the "back" of the brain (visual cortex, for instance).

Unfortunately I'm not sure how helpful it is to know this kind of stuff in terms of practical benefits for chess. More direct studies from psychologists tend to have more helpful information that can be applied in practice (e.g., the study that telling a story about a position helped subjects remember it--does it matter where the memories are stored, or the neuronal basis of such storytelling?).

Not to dis' neuroscience. It's the best!

Kalan said...


Thank you for your comment on my blog.

You can email me in ogn your 3 games, along with the student's notes/analysis if any. I will analyze the games and send them back.

The analysis for the first three games will be free of charge. If after that you feel that you are learning from my analysis and coaching techniques, we can discuss further coaching.

The analysis will be of the game as well as of the student's notes on the game (if any). I will provide analysis of variations as well as textual commentary.

My email:



Kalan said...

The rate for the analysis after the first three will be dependent on several factors:

1. Number of games
2. Your level of play (because I adjust my analysis accordingly to your level. No point teaching a beginner the nuances of playing a specific IQP position if he has just learnt making moves). I would prefer it if the student is above 1500 level.


Kalan said...

Re the comments on this post itself:

Well, as Glen Wilson points out, it is not enough to just "know"; because often, exact calculations are needed. What "knowledge" helps with is in narrowing down your field of calculation.

For instance, at a basic level, some readers will be aware of Lasker's famous double bishop sacrifice (Bxh7 and Bxg7). Knowing the idea behind that sacrifice will mean that on the board, the player will be OPEN TO LOOKING AT THE IDEA behind Bxh7. For an amateur, it is often easy to reject out of hand any move that is out of the ordinary - knowledge gets one over this problem.

Regarding the level of knowledge, it is estimated in psychological studies that grandmasters "know" 50000 patterns. One difference in the level of knowledge - many players know that an IQP is to be blockaded (ideally as Black with a Knight on d5). How many of the ordinary players know exactly what to do after Blockading? In which structure is it better to exchange a particular set of minor pieces as the attacking side? Are ALL iqp endings bad for the player with the IQP? With knowledge on these aspects, the player will KNOW what to calculate and what to immediately reject.

But at the same time, mere knowledge will be insufficient for anyone wishing to progress...

Rolling Pawns said...

Many years ago, back in Russia, I saw how the champion of the region where I lived, 20-year old guy, played a blitz game in the park.
He had a K+R vs. K, the thing is his flag was already hanging. So, his hand began to move like a rod in a fast running steam engine, "no thinking attached". When he finished, there was a checkmate on the board and the flag almost didn't move. What was it - procedural memory, semantic memory?

es_trick said...

I've heard Russian GM's use the expression "Learn it by hand"

that is learn to do certain things so well that your hand can seemingly move the pieces correctly without any thinking at all.

My understanding of "procedural knowledge" is that it's something you know how to do, without necessarily knowing how you know how to do it. It's contrasted wtih "declarative knowledge" which is something you can talk about with explanations.

For example, everyone knows how to speak at least one language (procedural knowledge). But relatively few people are able to go into much detail when it comes to citing grammar rules, or pronunciation rules, unless they are language teachers or linguists.

It's often the case that students of a foreign lanugage have a great deal of declarative knowledge concerning the language they study (more than native speakers of the same language) but relatively little ability to actually communicate in that language (procedural knowledge).

Perhaps it's similar for a lot of us chess players who study the game. We may have more declarative knowledge than procedural knowledge.

I recently read that Nakamura says he's hardly read any chess books.

A whiz like Magnus Carlson may be compared to a Mozart. How was Mozart able to "learn" so much music at such an early age? How did he "know" what he knew?

Capablanca, Kasparov, and Josh Waitzkin are said to have learned how to play the game just by observering others.

Banatt said...

Didn't know that, to say the least. You just keep posting really interesting articles, don't you?

Keep it up!

likesforests said...

chesstiger, I haven't read How to Think In Chess. If you make a go at it and learn something interesting, please share. :)

BDK - Thanks for the corrections on the names of things. I was hoping you would chime in and now I will have a more accurate model. :)

"Unfortunately I'm not sure how helpful it is to know this kind of stuff in terms of practical benefits for chess..."

Here is the practical benefit. I was briefly believing that one of the keys to chess was to overtrain how to do certain maneuvers until I could do them instantly without thought--and at this point I would never forget the knowledge (from a neuroscience perspective, this is storing it in procedural memory).

These studies show that's not the primary way stronger players in the 1700 to 2600 range are different from each other. They are accessing long-term semantic memory much more often. Which means the chunking, #patterns AND principles known theory is still a better explanation for chess mastery (you, of course, are well-aware of these already) and that then impacts how I will train for chess.

Kalan - The double (and even single) bishop sacs are great examples. Being aware of these motifs (Art of Attack goes into the subtleties) greatly simplifies calculating similar positions (and of course, some base calculating ability is also necessary).

rolling pawns - Procedural memory, almost certainly. These results are focused on tournament chess. In blitz chess, the optimal way to train is probably different. For example, in a tournament chess you often have 5sec/move to finish off a K+Q vs K ending... in blitz you may have 3sec for all your moves. This changes fundamentally how well you need to know the ending.

es_trick - You have drawn a good contrast between procedural vs declarative memory, but these scientific studies don't support how you compare masters & amateurs.

Specifically, these studies did not reveal any differences between how often masters and amateurs use procedural memory. They did reveal a big different in that stronger players (in the 1700 to 2600 curve) use declarative memory during a chess game. Stronger players tended to spend most of their time retrieving declarative memories, while weaker players tended to spend most of their time forming new declarative memories (in other words, they see less that is familiar in the positions and have to find new ideas).

These Russians that are trying to "learn by hand"--are not learning in the most efficient way for tournament chess between 1700 and 2600 according to scientific studies. Of course, if they spend 8 hours/day studying in an inefficient way, they will likely best someone studying 2 hours/day in an efficient way. Mindset, time, and learning ability are also very important.

"I recently read that Nakamura says he's hardly read any chess books."

No wonder he can play at extremely rapid time controls! But he is outstanding at any time control. These studies are again, more relevant to players between 1700 and 2600 at tournament time controls.


We often attribute his results to innate ability. For certain, he had talent, but he also spent countless hours playing with his harpsichord and even before that listening to his parents play.

We sometimes do the same with Mohammad Ali and Michael Jordan. Before they were great, nobody would have said they had the ideal build for their respective sports. There is much more to their success than innate ability.

banatt - thanks! :)

Polly said...

Like: I think there is a fine line between too much adrenaline and not enough. When I won that tournament back in March I was amped enough to stay focused and not take anything for granted, but not overly so that I couldn't stay calm in the face of adversity. There were a number of games where it could have gone either way, but a nice mix of calm and adrenaline helped me through.

There are times where being too calm can leave one in a state of lethargy where if the position doesn't have much going on one can easily slip into a lost position without seeing it coming.

I'm interested in that study regarding left and right handed players. I'm dyslexic and the form I have has to do with non-dominance in the brain. I showed no preference for the left or right hand as a very young child. The therapists working with my mother and me said she could choose to make me either hand. My mother being a lefty and having gone through hell, and having put my oldest sister through hell as a lefty made me right handed. As I got older she feels I probably should have been left handed. I'm wondering how this has impacted my development and thought processes as a chess player.

Chessbumbus said...


You have a nice blog here! I like visiting it!
I offering you a link exchange - if you accepting it.

My blog:
Welcome to visit! - Please leave a message if you are OK with the link exchange!