Characteristics of information decay in short-term visual memory and a diffusion model

Abstract

The effect of exposure duration of test stimulus on the decay rate in short-term visual memory (STVM) was investigated in a same-different task. In the model, it is assumed that memory noise, the variance of each convex value represented in STVM, increases with time, which causes the decay in STVM. In recognition experiments, the results showed that the decay rate is lower for longer exposure duration regardless of pattern complexity. Furthermore, the decay lasted gradually for 16 s with exposure duration of 1,200 ms, which suggests that prolonged exposure largely acts to reduce the decay rate but does not prevent the decay itself. These experimental data were well predicted by the model, which clearly indicates that memory noise increases as a linear function of retention interval and that the rate of increase of memory noise is inversely proportional to exposure duration. These results are interpreted by an extension of the integration model of Signal Detection Theory: The number of mental scanning, N, monotonically increases with an increase in exposure duration, s, and increase of memory noise is inversely proportional to N, so memory sensitivity d' is predicted to be multiplied by square route of s.