There are two books on this topic, that I have read long ago and that I can recommend:
Psychology in Chess by Nikolai Krogius
Schachpsychologie by Reinhard Munzert (in German)

As a training, you could try to play blindfold games against a friend. The free Tarrasch Chess GUI also offers this possibility. Make sure to use a weak engine.

>Does anyone know how to beat this? Has anything good been written on this topic? I notice that quite a few players are closing their eyes these days or turning away from the board - is that to visualise a 2D board in their mind which they can move the pieces around at will??

Chess for Zebras by Rowson talks about this a bit.

I believe what you're experiencing is the board not being abstract enough in your mind. It takes up too much mind space to visualize the board exactly, so you have to "compress" things so to speak.

I think initially everyone simplifies things a bit like not having colors to the squares, not having the board's geometry appear exactly as it would in real life, and some static/ignored aspects. A strong player though should be able to see much more complex patterns in abstract ways such as the relationship of important squares covered by pieces far away or dynamic factors such as a knight being a few moves away from a strong effect.

What struck me as odd was that Rowson asked players to describe the board in their head and found that the stronger players he asked had a very hard time with it. He himself said it was very abstract for him. I always thought I could describe it exactly but there's a large strength gap between Rowson and I. Now that I've studied a lot of tactics and chess in general since I read that book I think I'm beginning to see what he means.

In a lot of the tactics problems I did one side would naturally get a large advantage or winning position but in some the advantage wasn't so large so I would try to calculate playing the game through until I wasn't just sure of the win but I could see it. After this analysis I would put it into the computer and check to see how well I did. I did surprisingly well. At my best I felt that I was able to capture some of the more complex dynamic relationships and grab the essence or conclusion of their contribution in a given line and still continue analyzing without forgetting what I was working on. When this was the case a lot of times I'd have a very hard time describing the mental images involved. It's almost like they weren't there, the way information is stored in a computer system but remains invisible unless it has a medium through which it can be visualized.

Still there are a lot of challenges. I find that when I visualize something, anything really, that it's kind of cloudy. Either because 'clouds' are crossing the 'image', or the image itself is like a flash light being shone on different spots of a dark room. The hidden parts are just that, hidden. They have no color and offer no negative space, similar to the blind spot in the eye. There's no color to it and it doesn't define itself with clear borders.

The clouds, or dark areas not captured by the light, have a certain speed at which they move through the image. If I work with this speed I can remember certain bits and construct an operational picture. If I could always just take the parts that are important to the position even with my limited vision I think I would play very very strong but of course I cannot do this.

There's also a kind of size, and direction bias. Every time I imagine an image in my mind it tends to take on a certain size. If I'm asked to picture a compact disc for example, it is as if I am looking at from above at a height of 15 cm or so. If I picture it with something else it will be more like I'm looking at it from 15 meters. Of course I can call up images from memory of discs I've seen on a table or whatever, but this is something different. This is a problem because the 'clouds' have a different effect on each size. I have to struggle to see sizes in between to manipulate the effect the clouds have. Sometimes I've noticed that I can help this along by changing the size of the squares, the square borders if I choose to have them, or the pieces in a dramatic way. This is because it seems like I can imagine a tiny 1 cm person standing near a normal sized person, or my size relative to the city block, but I have trouble imagining size differences of less subtlety such as exactly how a 5'7" person would appear under a 10' ceiling.

The direction bias, or I guess you could call it a geometric bias is when I'm looking at a position but won't see certain movements because I'm automatically cutting them out. This is a real barrier towards clean and clear, high level calculation. Certain knight moves are often hard to consider, and I have trouble seeing a quick dynamic relationship when a piece is or will be between an active queen's "fingers". Like a queen on d4 is attacking the long diagonal, the d file and the 4th rank, but the squares it would attack if it were a knight ie. f5 are hard for me to tell how safe they are from the queen's next move especially on a busy board. It's like I see them as being "close" to the queen when dynamically they may be perfectly safe. Hard to explain.

Er I really started rambling there. Will just post now :-)

I would also recommend blindfold chess. This seems to help my son, although he is currently a long way from 2200 elo I am convinced it would help improve any level.