Cognitive mechanisms underlying the formation of offline representations in visual working memory

Abstract: Visual working memory (VWM) plays a foundational role in advanced cognitive functions. The state-based models propose a hierarchical organization of functional states, where memory representations with high attentional priority are retained in an online state (i.e., active state), while those with lower priority are kept in an offline state (i.e., passive state) for later use. The memory representations can be dynamically transferred between the two states according to the task demands. However, there was rare work to explore how the memory representations transitioned into the offline state from the online, generating the offline representations. Here, we put forward two hypothesis, the consolidation hypothesis and the fade-away hypothesis.
To explore this question, participants were instructed to remember two sequential memory arrays, with Memory array 2 being detected before Memory array 1. In this memory task, Memory array 1 was held in the offline state during the active maintenance of Memory array 2. Colored squares served as memory stimuli. 30 healthy college students participated in each experiment. We primarily modulated the temporal context related to the state transformation of memory representations: the interval delay between the two memory arrays in Experiment 1 and the presentation time of Memory array 2 in Experiment 2. The load of online memory varied between two and four in each trial. These variables were within-subject factors. Experiment 1 aimed to verify that the shortage of interval delay between memory arrays led to the failure of state transformation in the condition of 0.8s-interval. Experiment 2 attempted to determine which hypothesis, consolidation or fade-away, aligned better with the state transformation process.
The exploration of representational state transformation was built on the resources-dissociation account, which proposed that the offline representations are independent of the active processing of online representations. Memory arrays 1 and 2 were used to test the offline and online memory, respectively. The results of Experiment 1 showed that variations in online load did not affect offline memory when extending the interval delay from 0.8s to 1s. This indicated that the state transformation of Memory array 1 continued beyond 0.8s after its disappearance and could complete within a 1s-interval. In Experiment 2, the interval was designed at 0.8s. We observed that the online load variation had no impact on offline memory when extending the presentation time of Memory array 2 from 0.2s to 0.5s. This supported the consolidation hypothesis, indicating that the sufficient presentation time of Memory array 2 allowed for the state transformation of Memory array 1 to complete before the subsequent processing of Memory array 2. Thus, we concluded that the state transformation involved a consolidation processing to transfer the online representations to the offline state, rather than natural fade-away of persistent neural activity.
In summary, the state transformation acts as a process of consolidating online memory representations into the offline state, thereby forming offline representations. This process can be completed within a sufficiently long retention interval, or continue during the presentation of subsequent stimuli when providing a deficient interval. The current findings provide fresh insights into the mechanisms of representational maintenance in the two distinct states.