Abstract: PURPOSE: Whether visual working memory (VWM) is retained in the same brain areas involved in perception processing remains a subject of debate. However, studies examining this question have primarily concentrated on the representation of fundamental visual features, often neglecting the hierarchical and topological organization of the visual system. To address this gap in knowledge, we employed real-world objects as stimuli to systematically explore the representation of information in perception and working memory within the ventral visual hierarchy and association areas.
METHODS & RESULTS: We assessed the representation of 20 objects across three VWM and four perception tasks using fMRI recording. In experiment 1, twelve participants performed a retro-cue sequential VWM task in which they viewed two objects and were required to recall the cued object after a 10-second delay. We found that both the identity and categorical information of the cued objects could be decoded from the lateral occipital complex (LOC) and intraparietal sulcus (IPS), but not in early visual areas. Moreover, a common representational pattern between VWM and perceptual tasks was evident solely in the LOC through representational similarity analysis (RSA). These results suggest that task-relevant high-level sensory areas are specifically recruited for VWM maintenance. Experiment 2 further examined the extent to which representational properties were shared between VWM and perception. Six participants took part in a retro-cue spatial VWM task where they memorized a cued object from two objects presented separately in the left and right visual fields. This experiment reinforced the findings of Experiment 1 regarding the roles of LOC and IPS. Interestingly, while a strong contralateral bias was confirmed in the perception task, this bias was significantly reduced in the VWM task, as evidenced by enhanced ipsilateral representation of the cued object. Furthermore, we conducted an ablation analysis by excluding regions that exhibited a similar response profile between the two visual fields in a control perception task. The results revealed that more ipsilateral voxels were involved in representing memory item as compared to perception task, refuting the hypothesis that the ipsilateral representation in VWM was due to the large receptive field of the LOC. In Experiment 3, to force the participant to combine the identity of the memorized item and its corresponding location, we delayed the retro-cue at the end of the period. Persistent but significant reduced involvement of the ipsilateral LOC was observed in delay persisted, suggesting the representation of visual working memory can be flexibly modulated by task configuration.
CONCLUSIONS: Our study underscore the essential role of both the LOC and IPS in maintaining real-word object representations in VWM, while the sensory-based object representations in high-level visual areas may go beyond the feedforward visual information flow during VWM. The enhanced ipsilateral representation of objects during VWM expand the activation of memory-specific content and could aid in the stable retention of memory content.

Key words: visual working memory, object perception, functional MRI, lateral occipital complex