Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55–81.
In {de Groot} ({1965})’s experiments, chess masters could {reliably reconstruct chess boards} after only 5 seconds of viewing (novices couldn’t at all)—but critically, they couldn’t {reconstruct random chess boards}.
The authors suspect that de Groot’s results mean chess masters have an {ordinary limited} Span of working memory, but that they perceive chess boards with {larger “chunk” sizes (“Chunks” in human cognition, Human channel capacity increases with bits-per-chunk)}. The implication is that chess mastery comes from {internalizing increasingly sophisticated representations of the board (Recoding can increase chunk size)}.
They ran two experiments to test this hypothesis, using a master chess player, an advanced chess player, and a beginner.
The subject was presented with two chess boards: one with a pre-set board position, and one empty. Their task was to duplicate the pre-set board as quickly as possible. The experimenters recorded the subjects with a camera to see how many pieces they perceived with each glance at the pre-set board (as a proxy for chunk boundaries).
The time spent looking at the pre-set board in each glance varied inversely with chess experience, but the number of pieces set per glance didn’t vary appreciably with player skill.
The perception task was repeated, except with a partition separating the pre-set board and the blank board. It was lifted for 5 seconds at a time to allow the subject to view the pre-set board. Once the subject had placed as many pieces as they could, the could request that the partition be lifted again.
The better chess players placed more pieces per trial (and hence took fewer trials to complete the board) when the boards represented an authentic game, but not when they were random, largely in accordance with de Groot’s data.
The experimenters noticed that when players were in the recall phase, they’d place several pieces rapidly, then pause around two or more seconds, then repeat. They hypothesize that these pauses represent the time needed to recall each successive chunk. These patterns suggested that the better chess players were recalling more pieces in each chunk.
Using this technique, the data suggest that better players were able to recall both more chunks and larger chunks in each trial. The authors hypothesize that better players don’t really have larger Span of working memory, but they have a catalog of common configurations stored in long-term memory, and that some of the pauses represent some kind of hierarchical processing which relates the chunks to these memorized configurations.