Novice acquisition of skilled recall of chess positions was studied in an e
xperiment in which two novices studied a series of five hundred chess posit
ions during a period of several months. They spent fifteen minutes to half
an hour a day leaching themselves these positions. As a result their skill
in recalling chess positions rose from sixteen percent to somewhere between
forty to fifty percent. The learning curve proved to have a shape which in
dicates that in the beginning learning is very fast but after some 100-150
studied positions the speed of learning decreases substantially.
A computer simulation was used to model the results and analyse alternative
explanations. Two alternative ways of thinking were tested. In the first,
chunk construction was assumed to be based on the neighbourhood of associat
ed pieces. The second model assumed a frequency-based correlative associati
on process. Although the learning curves of the two models are very similar
in shape to those of the subjects, the frequency-based associative model g
ave a better explanation for the data. This is why it is natural to suggest
that common co-occurrence in addition to easily recognizable chess-specifi
c characteristics, like colour and type of pieces, guide associative proces
ses during chess players' learning of chess-specific chunks.