Abstract
<p>Visual working memory (VWM) is central to human cognition, providing an interface between visual processing and higher-level cognition. While several cognitive models have been proposed to explain VWM, typically assuming that recall error distributions in continuous reproduction tasks reflect a mixture of memory and guessing processes, they suffer from two key limitations. Specifically, these models (i) focus on response precision, while ignoring response times or failing to capture them accurately, and (ii) suggest guessing as a theoretically motivated process while leaving its underlying process unspecified or overconstrained. Here, we propose the Noisy Guessing Mixture Circular Diffusion Model (NGM-CDM), which accounts for both the response error and response time distributions, and assumes that the guessing process is noisier than the memory process, reflected by a larger diffusion coefficient in the guessing process. Our simulations show that the NGM-CDM reliably estimates the relative noisiness of guessing and memory processes, demonstrating its potential as a measurement tool. Moreover, by reanalyzing seven datasets (total N = 443), we show that the NGM-CDM captures several key qualitative patterns, such as minimal increases in mean response time with increasing memory load and changes in the shape of the recall error distribution across memory load. Quantitative and qualitative model comparisons indicate that the NGM-CDM provides a superior account of empirical data relative to other previously proposed models. Therefore, we believe that the NGM-CDM provides an important theoretical and measurement advancement for VWM research, and serves as a key building block for future, increasingly comprehensive models of VWM.</p>