Visual Working Memory: Large-Scale Experiment, Comprehensive Exploration, and Integrative Framework

Two decades of research on visual working memory have yielded substantial yet fragmented knowledge, akin to many pieces of a puzzle. Here, I attempt to assemble these pieces into an overall picture, creating an integrative framework. Based on a large-scale behavioral experiment that consists of 40 million responses to 10,000 color patterns, a quasi-comprehensive exploration model of visual working memory, termed QCE-VWM, has been constructed. Despite its markedly reduced complexity (57 parameters as opposed to 30,796), the QCE-VWM outperforms neural networks in data fitting. This model provides an integrative framework for human visual working memory, reconciling previously conflicting concepts such as “slot” and “resource.” It modifies traditional methods, including Bayesian integration and the von Mises distribution, and identifies new mechanisms like “red advantage,” “red disadvantage,” “Matthew effect,” and “crosstalk.” This study underscores the value of large-scale behavioral experiments in crafting comprehensive models of cognitive mechanisms.