Abstract
Massive studies have explored biological motion (BM) crowds processing for their remarkable social significance, primarily focused on uniformly distributed ones. However, real-world BM crowds often exhibit hierarchical structures rather than uniform arrangements. How such structured BM crowds are processed remains a subject of inquiry. This study investigates the representation of structured BM crowds in working memory (WM), recognizing the pivotal role WM plays in our social interactions involving BM. We propose the group-based ensemble hypothesis and test it through a member identification task. Participants were required to discern whether a presented BM belonged to a prior memory display of eight BM, each with distinct walking directions. Drawing on prominent Gestalt principles as organizational cues, we constructed structured groups within BM crowds by applying proximity and similarity cues in Experiments 1 and 2, respectively. In Experiment 3, we deliberately weakened the visibility of stimuli structures by increasing the similarity between subsets, probing the robustness of results. Consistently, our findings indicate that BM aligned with the mean direction of the subsets was more likely to be recognized as part of the memory stimuli. This suggests that WM inherently organizes structured BM crowds into separate ensembles based on organizational cues. In essence, our results illuminate the simultaneous operation of grouping and ensemble encoding mechanisms for BM crowds within WM.
Introduction
Human working memory (WM) is notoriously limited, typically holding only a few items at a time. However, we navigate a world filled with "biological motion" (BM)—the complex movements of other living things—often in crowded settings like busy streets or sports games. Previous research showed that for uniform crowds, the brain uses "ensemble coding" (averaging features like walking direction) to save space. But real-world crowds aren't uniform; they have structure (e.g., two opposing teams on a field). This study addresses whether our memory respects this structure: does the brain create one giant, blurry average of everyone on screen, or does it maintain distinct "summary reports" for different sub-groups?
Methods
The researchers conducted three experiments using Point Light Displays (PLDs)—animations where 13 moving dots represent the joints of a walking human. The stimuli were generated using walking sequences from the Carnegie Mellon University (CMU) Motion Capture Database, a standard benchmark in biological motion research.
Participants performed a "member identification task." They viewed a memory display of eight walkers with different directions, followed by a 1000ms delay. They were then shown a single "probe" walker and asked: "Was this specific walker in the set you just saw?"
The researchers used Gestalt principles to create structure:
- Experiment 1 (Proximity): Walkers were physically separated into two groups (left vs. right side of the screen).
- Experiment 2 (Similarity): Walkers were distributed uniformly in a circle but were grouped by their walking directions (e.g., four walking generally leftward, four generally rightward).
- Experiment 3 (Robustness): Similarity cues were weakened to see if the grouping effect would persist when the structure was less obvious.
Key Results and Insights
The core finding of the study is the "Mean Bias." Across all experiments, participants showed a "sharp tuning" of responses around the subset means. Endorsement rates for the subset mean were significantly higher than for the actual individual directions present in the display ($p < 0.001$ in most conditions).
This suggests a process of Hierarchical Compression: the brain doesn't choose between grouping and averaging; it does both. It uses organizational cues (like where people are standing) to "segment" the crowd and then calculates an average for each segment. Notably, participants were much more likely to endorse the local mean (the average of a sub-group) than the global mean (the average of all eight walkers). This indicates that our mental representation of a crowd is a collection of structured summaries rather than one single "gist."
Furthermore, this process appears to be automatic. Participants were not instructed to average directions; the task encouraged them to remember individuals. The persistent gravitation toward the "mean" suggests that ensemble coding is an involuntary, "built-in" feature of working memory.
Discussion and Future Work
While this study demonstrates that we store multiple ensembles, it does not define the upper limit of how many distinct groups can be "summary-coded" before memory fails. Additionally, the study utilized basic physical cues (proximity and similarity). Future work should incorporate "social Gestalt" principles, such as joint gaze, to see how social context influences grouping.
These findings could eventually inform crowd management strategies or the design of visual signage in high-traffic environments like airports, where understanding how people "chunk" moving crowds is vital for safety and navigation.
Summary updated