When a Stranger’s Gaze Decides Which Gap Looks Wider

Every day, people squeeze through gaps that are not made of walls or furniture but of other human bodies: a crowd parting around a bus stop, a cluster of colleagues chatting in a corridor, strangers paused mid-conversation in a doorway. Deciding how, or whether, to slip past such human-made apertures involves far more than measuring physical distance. A new study published in Frontiers in Psychology suggests that the very same gap can be judged differently depending on what we intend to do with it, and that a single avatar’s facing direction can flip our choice from attraction to outright avoidance.

The research, conducted by Ryosaku Makino, Hiroshi Matsui, Kentaro Kodama, Takayuki Tomono and Katsumi Watanabe at institutions including Musashi University, Waseda University and Tokyo Metropolitan University, asked a deceptively simple question: when two gaps are physically identical, do the social postures of the people forming them change which one we pick? To find out, the team ran two online experiments with Japanese participants recruited through a crowdsourcing service, obtaining complete responses from 91 participants in the first experiment and 92 in the second. Crucially, no participant took part in both tasks, making this a between-subjects comparison of pure task context.

The stimuli were elegantly minimal. Each scene showed three human-like avatars arranged at equal intervals, creating two apertures of exactly the same width. The social configuration of each gap varied systematically: the two avatars flanking an aperture could stand face-to-face, suggesting an ongoing conversation; back-to-back, facing away from each other; or in the same direction, as if walking together. The researchers distilled these arrangements into a hypothesis-driven index they called the Degree of Preference, assigning back-to-back pairs a value of 3, same-direction pairs a value of 2, and face-to-face pairs a value of 1. The difference between the right and left apertures, termed ΔDoP, captured the relative social structure of the whole scene. A second cue, the center avatar direction or CAD, recorded which way the middle avatar, which bordered both gaps, was facing.

Participants saw each of eight stimulus conditions once, after a credit-card-based calibration procedure that approximately standardized apparent stimulus size across devices. In the first experiment, they simply chose which aperture appeared wider. In the second, they chose which aperture they would prefer to pass through. Everything else was identical, down to the five-second stimulus presentation, the randomized condition order, and the confidence rating collected after each choice on a five-point scale. The divergence in results between these two nearly identical tasks is where the study becomes genuinely striking.

Using generalized linear mixed models with a binomial link function, the team found that ΔDoP behaved consistently across both tasks. In the width-judgment experiment, each one-unit increase in ΔDoP nearly doubled the odds of choosing the right aperture as appearing wider, with an odds ratio of 1.94. In the passage-preference experiment, the effect was almost identical in magnitude, with an odds ratio of 1.82. In other words, gaps flanked by back-to-back avatars were not only preferred as routes but were also judged as spatially wider, even though every aperture in the study had exactly the same physical width. The relative social configuration of a scene, it seems, biases perception itself, not merely our behavioral preferences.

The center avatar told a very different story. In the width-judgment task, when the central avatar faced right, participants were about 2.18 times more likely to pick the right aperture, a modest attentional pull consistent with decades of research showing that gaze and head orientation automatically orient visual attention. But in the passage-preference task, the effect reversed dramatically: when the center avatar faced right, the odds of choosing the right aperture collapsed to just 14 percent of the odds when it faced left, an odds ratio of 0.14. Participants actively avoided the direction the avatar was looking. A statistical model combining both experiments confirmed this dissociation formally: the Experiment by CAD interaction was highly significant, while the Experiment by ΔDoP interaction was not.

The authors interpret this reversal through the lens of ecological psychology and social affordances. When judging abstract spatial properties, the avatar’s facing direction may act as little more than a spatial-attentional cue, drawing the eye toward whatever it looks at. But when participants imagine walking through the scene, the same orientation acquires action-relevant meaning: a person facing you may be about to move, may block your path, or may initiate an interaction. Avoiding the aperture aligned with the avatar’s gaze then becomes a strategy for dodging potential interference or collision, echoing real-world findings on pedestrian collision avoidance and the so-called hesitant avoidance phenomenon observed when two people approach each other head-on.

Exploratory analyses of response times and confidence ratings added a further layer. Participants responded faster and with higher confidence in the passage-preference task than in the width-judgment task, and the negative correlation between response time and confidence, the familiar pattern that fast answers tend to feel certain, was substantially stronger in the action-oriented task (a Spearman correlation of about −0.33 versus −0.12). This suggests that anticipating movement organizes perception more firmly around potential action, tightening the coupling between environmental structure and behavior. Both experiments also revealed a modest baseline tendency to choose the right aperture, consistent with previously documented rightward biases in spatial attention and navigation, though the authors treat this exploratory finding cautiously.

The study’s model fit tells its own story. In the width-judgment task, the best model explained only about 6 percent of variance through fixed effects, rising to roughly 29 percent with individual differences included. In the passage-preference task, those figures jumped to 40 and 50 percent respectively, implying that action-oriented judgments are far more systematically constrained by the social layout of a scene. When the body is hypothetically in motion, the social geometry of the environment appears to organize behavior much more tightly than when the body is a detached observer.

The authors are candid about the limitations. The experiments were conducted online, with button presses rather than actual locomotion, so the findings reflect anticipatory judgments rather than embodied behavior. Body dimensions were not measured, ΔDoP and CAD were strongly correlated in the stimulus set rather than orthogonally manipulated, and no non-social control condition was included to isolate social effects specifically. Demographic information was not collected, and the forced-choice design of the width task means the results demonstrate systematic bias rather than a true perceptual distortion of width. Future work using immersive virtual reality, real-world walking paradigms, and eye-tracking will be needed to test whether the dramatic CAD reversal survives when people actually move.

Even with those caveats, the implications are thought-provoking. Socially structured space is not perceived passively; its meaning is assembled on the fly from what we are about to do. A stranger’s turned head can make a gap look wider or make us swerve away, depending solely on whether we are judging space or planning to move through it. For architects, urban planners and anyone designing crowd flow through stations, stadiums or streets, the message is that human orientation is not decoration but infrastructure: the way people face within a space quietly reshapes the routes everyone else will take through it.

Subject of Research: Task-dependent effects of social cues on aperture width judgments and passage preference

Article Title: Left or right? How social cues and task context influence relative width judgments and aperture choice

Article References: Makino, R., Matsui, H., Kodama, K., Tomono, T., & Watanabe, K. (2026). Left or right? How social cues and task context influence relative width judgments and aperture choice. Frontiers in Psychology, 17, Article 1895232. https://doi.org/10.3389/fpsyg.2026.1895232

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1895232

Keywords: social cognition, aperture passing, affordances, spatial perception, gaze cueing, task context, passage preference, avatars, embodied cognition, pedestrian navigation, psychophysics, mixed-effects modeling