Abstract. With the help of the neural network model of heteroassociative memory and the buffer model, which stores information about all stored objects separately from each other, the analysis was performed of the distribution of errors in psychophysical experiments on memorization and immediate reconstruction of the sequences of movements of pieces on a chessboard. It is shown that the distribution of placement errors in the test does not depend on the task (the recognition or the reconstruction of the whole play). The results of model experiments suggest that when memorizing sequences in solving the both problems the positional coding is used. To describe in a single structure both processes  recognition or  reproducing  the neural network model of the heteroassociative memory type is more successful than the buffer model.

 

Key words: heteroassociative neural network, spatial memory, positional coding.