Fig. 3: Structure of QCE-VWM model.

This model aims to simulate observers’ responses by blending knowledge-based responses with a proportion of random guesses. The former comprises three types of normally distributed components: eight color-category-biased components, one unbiased component, and three swap-based components. Throughout the operation of the model, the weights of these components, along with the means and SDs of their distributions, as well as the proportion of random guesses, are calculated and used collectively to make predictions about the distribution of responses. Blue, green, red, and yellow arrows respectively represent the effects of interaction on retention, interaction on bias, chunking, and spatial attention.