It is well known that while everyone has roughly the same capacity of visual working memory, expertise in an area drastically alters our efficiency in utilizing that memory. For example, chess masters are much better than novices at recalling chess positions when they look like typical middle-game positions, but they perform nearly the same as novices on random arrangements of pieces [1]. In the rubik's cube case, the expert is not memorizing 54 individual squares, but rather they are taking advantage of learned patterns to store fewer, larger chunks.

For another example, consider memorizing strings of letters: Which one do you find easier? GKAMIFTBW or IBMFBICIA

[1] Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive psychology, 4(1), 55-81.

Would Stephen Wiltshire's talent be attributed to the same explanations ?

http://www.stephenwiltshire.co.uk/

http://www.dailymail.co.uk/news/article-1223790/Autistic-art...

https://www.youtube.com/watch?v=bsJbApZ5GF0

Great question and I have no idea. My comment was geared toward those with normal function (should have specified). There is a large literature on memory and austism/savantism, but I'm familiar with none of it. However, my guess would be that the current best answer is as it is with nearly all of autism -- we don't know.

And likewise with the fact that a surprising number of people have 'aphantasia' and cannot visually imagine things at all. Humans obviously have a highly variable capacity for visualisation.

TFA also seems to conflate "imagining a scene containing a combination of objects" with "imagining an assortment of randomly oriented Gabor patches", which I don't think follows at all. You can't just disregard the semantic content of what's being imagined. It's like saying humans can't recall sentences containing 50 characters because they can't recite a string of 50 random letters.

It is well known that semantic groupings increase memory or working memory capacity. The author alluded to that. The Gabor patches are probably used in part because semantic strategies are more difficult to use and gain from using those stimuli.

Or perhaps just memorized the solution (that they figured out while still being able to see the cube).

I haven't ever tried doing it blindfolded, but I would expect there is not too much visualization behind it.

Solving blindfolded is just momorizing 2 sequences about phone number length. Not a whole lot of visualization. Might be different for the top tier bld solvers

Based on my own experience (not solving Rubik's cubes, but still), it's similar to how one can navigate with turn-by-turn directions despite not having a map. You just need to know how to "walk the graph" of the objects involved.

IIRC there are well established patterns to solve or nearly solve a cube from given starting conditions. Could that be it?

Edit: https://ruwix.com/the-rubiks-cube/how-to-solve-the-rubiks-cu...