Abstract
The paper examines the merit of a small memory capacity on the detection and utilization of correlations within a multiple-variable environment that contained dummy variables. The participants were asked to predict a number (1 or 2) hidden within a figure with four features (color, shape, size, and number) on each trial. All features were binary variables, and only color was correlated with the hidden numbers. The results indicated that a small working-memory-capacity group performed better than a large working-memory-capacity group, which suggests that a smaller working memory capacity is advantageous for detecting and utilizing correlations. This implies an adaptive advantage of a smaller memory capacity.
