From separate items to an integrated unit in visual working memory: Similarity chunking vs. configural grouping

It is unclear whether sequentially presented items can be grouped into an ensemble by similarity chunking without the aid of the configural cue (e.g., Kanizsa figure). The current study systematically explored this issue by gradually weakening the configural cue until it was completely removed. The results showed that sequentially presented Pac-Men could be integrated into a Kanizsa figure for better memory only when they were successively presented in adjacent locations. When Pac-Men were presented at non-adjacent or unpredictable locations, they could be chunked by their orientation similarity (they all faced toward the fixation cross) to improve memory performance. Importantly, such orientation similarity alone was sufficient to evoke memory benefit, even when successively presented Pac-Men could not form a Kanizsa object structure. This memory benefit cannot simply be attributed to the statistic regularity or mental transformation based on the egocentric reference frame. However, the configural cue alone could not produce a memory benefit if these Pac-Men did not face toward the fixation cross. All of these results indicate that the memory benefit of sequentially presented items could occur by similarity chunking without the aid of the configural cue. The memory benefit based on the configural grouping was fragile when items were sequentially presented.