Xinyu Zhang¹ (), Jinglan Wu¹, Tengfei Wang¹, Mowei Shen¹, Zaifeng Gao¹; ¹Zhejiang University

The binding problem is a central issue in cognitive science, addressing how integrated information is processed and stored in mind to support cognition and behavior. Despite its significance, whether a dedicated storage buffer exists for bound representation in working memory (WM) remains an open question. This study aims to address this question through latent variable modelling (Study 1) and psychophysical experiments (Study 2). In study 1, we adopted 13 tasks to tap the WM storage of visual objects, verbal stimuli, and bound representations, the central executive, and the storage of bound representations in long-term memory. A total of 184 participants completed these tasks. Confirmatory factor analysis revealed no evidence for an independent buffer dedicated to bound representations storage in WM. The storage of bound representations in WM closely linked to the buffer that holds their constituent elements. Meanwhile, a clear distinction emerged between WM binding and long-term memory binding. Study 2 extended the findings of Study 1 by conducting three psychophysical experiments. We employed a dual-task paradigm in which participants were required to simultaneously memorize two types of stimuli: single features and bound representations. If these two types of representations share the same buffer in WM, increasing the load on one would reduce memory performance for the other (i.e., dual competition effect). We observed dual competition between orientation and letter-location binding (Experiment 1) and between orientation and color-letter binding (Experiment 2). Meanwhile, no competition was observed between location and color-letter binding (Experiment 3), suggesting that location may have a unique role in WM. These findings align with the prediction of Study 1 and provide complementary evidence. Taken together, these studies suggest that there is no dedicated buffer for bound representations in WM, offering new insights into the structure of current WM models.