A neural network model of short-term sequence memory

As a principle of temporal pattern processing in brain to encode arbitrary sequences, we indicate the effectiveness of "a pointer whose speed is adaptively changed by transitions in sequences", and the fact that this pointer can be realized by "propagation of local firing on one-dimensional nerve field whose symmetry is externally controlled". By computer simulation, we show that this principle enables the sequence encoding with nonlinear time length compression, retaining the memory trace as static firing patterns, and recalling the sequence with its original speed.