Pith. sign in

REVIEW 3 major objections 6 minor 57 references

Pedestrians treat other people's conversations as invisible walls — reading gaze, talk, distance, and body orientation to decide whether to pass — and the walls fall once a single pedestrian walks through.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 09:05 UTC pith:ST77DY5A

load-bearing objection Core avoidance findings hold up well; the collective-breach result is real but the contagion reading is not yet supported because group following was never modeled. the 3 major comments →

arxiv 2607.20876 v1 pith:ST77DY5A submitted 2026-07-23 physics.soc-ph

Invisible walls: how pedestrians navigate around social interactions

Jack Terwilliger , Julia Di Silvestri , Seika Murase , Anne Elizabeth Clark White , Federico Rossano This is my paper
classification physics.soc-ph PACS 89.65.-s
keywords pedestrian dynamicsproxemicssocial normsinteractional territorysocial contagioninvolvement signsfield experimentspedestrian navigation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Pedestrian traffic is usually modeled as a physical system of moving bodies avoiding collisions, but this paper argues walkers also navigate a social landscape made of invisible walls: the territories created by other people's conversations. Across four field experiments with roughly 4,900 observed pedestrians, the authors show that whether a stranger walks between two people depends on the signs those people display — mutual gaze, speech, close standing, and face-to-face orientation each lower the chance of being walked through. When two people simply stand close facing each other, passersby detour around them almost as reliably as around a physical obstacle, while the same space between two chairs gets crossed nearly every time. The paper's second claim is that the norm is collectively reversible: a pedestrian is nearly three times more likely to walk through an interaction if someone breached it within the previous five seconds, a pattern the authors read as social contagion of rule-breaking.

Core claim

On the paper's own terms, the discovery is that pedestrians continuously infer a type of spatial claim the field has mostly ignored: interactional territory, an invisible boundary around an ongoing social interaction that non-participants treat as off-limits. Unlike personal space, whose boundary is attached to a single body, this territory is latent — it exists only in what the interacting people make observable — and pedestrians compute its presence from involvement signs such as where others look, whether they talk, how their bodies face, and how far apart they stand. The authors show each sign independently shifts breaching probability, that mutual gaze saturates the contribution of talk

What carries the argument

The central object is interactional territory: an invisible spatial boundary around an interaction that non-participants normally avoid crossing. The mechanism that carries the argument is involvement-sign reading — pedestrians use gaze direction, speech, body orientation, and interpersonal distance as observable evidence of whether an interaction exists and how strong its claim on the space is. The experiments work by manipulating one sign at a time (gaze versus phone, talk versus silence, five versus ten feet, face-to-face versus offset versus back-to-back, two people versus two chairs versus a sign) while holding physical occupancy constant, so changes in breaching rate isolate the social

Load-bearing premise

The collective-breach result assumes the five-second prior-breach count is the cause of followers' behavior, not a proxy for something else — a crowd surge, a gap in the actors' performance, or flow direction — that independently makes later pedestrians more likely to walk through.

What would settle it

A randomized field test would settle the contagion claim: have trained confederates breach the two actors' interaction at randomly chosen moments on matched days, while never breaching on control days, and compare the breach rate of naive pedestrians in the following ten seconds; if the rate does not rise after confederate breaches, the social-contagion claim collapses. The chair-versus-actor contrast already tests the avoidance claim, so the sharpest unsettled test is the breach-window one.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If walkers treat interactional territory as a navigation constraint, then robots, delivery drones, and autonomous vehicles sharing walkways must infer the same signs — gaze, talk, orientation — to avoid socially disruptive paths.
  • Pedestrian-flow models built purely from collision avoidance will systematically mispredict where people walk, since breaching probability changes by up to a factor of 17 on social signs alone, independent of physical occupancy.
  • The collective-breach result implies conversational groups create only provisional barriers; crowd managers cannot rely on a conversation group to keep a lane clear once any pedestrian has crossed it.
  • The sign-versus-mural contrast shows the inference is content-sensitive: pedestrians do not avoid everyone who looks at something, only people engaged in interactions whose space others can disrupt.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the contagion effect is causal, the same release mechanism should appear in other proxemic norms such as personal space or queue order; a direct test would compare follower breach rates after an unambiguous confederate breach versus after a near-miss that leaves the territory intact.
  • The saturation of talk by gaze suggests a threshold model of interaction detection rather than a linear integrator; a testable prediction is that adding further involvement signs, such as touch or sustained joint attention, yields no additional protection once mutual gaze is present.
  • The mural result implies pedestrians model the affordances of what a person attends to; an unexplored extension is whether pedestrians also distinguish pairs of friends from strangers or task-oriented talk from social chat.
  • Fitting breach probability as a function of continuous variables — gaze angle, distance, talk duration — in the same field setup would turn the framework into a predictive spatial map of social cost usable for robot navigation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper reports four field experiments (total N=4,911) in which pairs of actors or actor-object displays were placed on busy pathways, and the outcome was whether pedestrians walked between the display ('breach') or around it. Experiment 1 manipulated mutual gaze, talk, interpersonal distance, and actor gender; Experiment 2 manipulated mutual body orientation and measured continuous trajectories; Experiment 3 contrasted a person attending to an informational sign versus an art mural; Experiment 4 contrasted two actors with two chairs and an empty baseline. The main avoidance effects are strong: pedestrians are much less likely to walk between actors displaying mutual gaze, talk, close spacing, and face-to-face orientation, and the chair/sign controls argue against a purely physical-obstacle explanation. The paper also reports a collective-norm-breaching effect in Experiment 2: a pedestrian was about 2.77 times more likely to breach if another pedestrian had breached in the preceding 5 s. The authors interpret these results as evidence that pedestrians infer a latent 'interactional territory' and that this territory can be collectively overridden.

Significance. If the conclusions hold, this is a useful contribution to pedestrian dynamics and social cognition: it provides naturalistic, experimental evidence that route choice is modulated by inferred social-interactional states, and it offers quantitative effect sizes (odds ratios) that could inform socially aware navigation models. The manuscript has notable strengths: four experiments, counterbalanced schedules, inter-coder reliability checks (κ≈0.84–0.92 in Exp. 2), physical-control conditions (chairs; sign vs. mural), continuous trajectory measurement with reported reprojection error, and public data/code. The main risk is that the headline collective-breach result may reflect within-group coordination rather than contagion among unrelated pedestrians, because group membership was coded but omitted from the contagion model. This is testable and fixable with the deposited data, so I do not see it as fatal to the paper, but it is load-bearing for the abstract's second claim.

major comments (3)
  1. [Results, 'Pedestrians collectively breach interactional territory'; Supplementary, 'Calculating predictors and controls] The collective-breach GLMM uses the count of all pedestrians who breached in the previous 5 s as the key predictor, but it does not include group membership, group size, or group-lead status, although these were coded for every pedestrian (Table S2). If the prior 'breacher' is in the same group as the target pedestrian, the predictor captures within-group coordination rather than exposure to an independent norm violation. The qualitative example in Fig. 5 begins with a group of three pedestrians breaching, so subsequent breachers may include group members. The reported OR ≈ 2.77 could therefore overstate contagion to unrelated pedestrians. Please re-fit the model with a same-group prior-breach indicator (or cluster by group ID) and report the effect restricted to pedestrians with no groupmate among prior breachers. If the effect vanishes, the collective-override claim in the abstract mus
  2. [Experiment 1 (Results, Fig. 2C) and Experiment 4 (Results, Fig. 4)] The interpersonal-distance effect in Experiment 1 — OR = 0.06 for 5 ft vs. 10 ft — is the largest reported effect, but it is potentially confounded with physical passage width: at 5 ft the gap between the actors is half as wide as at 10 ft. The chair control in Experiment 4 is informative, but it was run only at 5 ft; it does not establish whether a 5 ft vs. 10 ft gap between inert obstacles would also reduce breaching. As written, the abstract's claim that pedestrians 'integrate ... proximity' as a social cue is stronger than the evidence provided. Please add a chair-distance condition or an analysis (e.g., using the trajectory data to model path clearance and energy cost) that separates physical affordance from social inference.
  3. [Supplementary, 'Calculating predictors and controls for collective proxemic norm breaching'] The 5-second prior-breach window is a free parameter; robustness to window length is reported only in the Supplementary. More importantly, the collective-breach model omits pedestrian walking direction, even though direction interacts with body orientation in the same experiment (Fig. 2D), and crowd-flow direction could be a time-varying confound correlated with both prior breaches and the current breach. Please include direction (and group membership, as above) as fixed effects, or show in a sensitivity analysis that the odds ratio is unchanged.
minor comments (6)
  1. [Throughout] Typos and formatting: title has 'Invisible w alls'; abstract has 'Keywordspedestrians'; Introduction has 'challenging problem to for machines'; reference to 'Physics Review E' should be 'Physical Review E'.
  2. [Results, Fig. 5 vs. Supplementary Fig. S4] The anecdote is inconsistent: the main text says 12 of the 19 remaining pedestrians breached after the group of three, while Figure S4 says 15 of 19. Please reconcile the counts.
  3. [Results and Supplementary, Experiment 3] The sign vs. two-actors comparison is reported as OR = 3.57 in the Results and as '3.83 times more likely' in one sentence of the Supplementary. Please check which value is correct.
  4. [Results and Supplementary, Experiment 3] The phrase 'additive interaction' is confusing: an interaction on the log-odds scale is not additive in the ordinary sense. Please rephrase or explain the model scale.
  5. [Methods, Experiments 1 and 3] Exclusion proportions are large (458/2,319 ≈ 19.8% in Exp. 1; 641/1,515 ≈ 42.3% in Exp. 3). A CONSORT-style flow diagram and a robustness analysis retaining all coded pedestrians would help, since 'never traversed near the actors' may not be condition-independent.
  6. [Methods / Statistical analysis] No pre-registration is mentioned. For an observational field study with multiple models, it would be useful to state which analyses were confirmatory and which were exploratory.

Circularity Check

0 steps flagged

No circularity: all reported effects are fitted contrasts on experimentally manipulated stimuli; the collective-breach claim is a fitted temporal association, not a constructed prediction.

full rationale

No circularity found. The paper makes no formal derivation; every headline claim is a fitted GLMM contrast against experimentally manipulated conditions (gaze vs phone, talk vs silence, 5 vs 10 ft, body orientation, and actor vs chair/baseline). The outcome (breach vs no breach) is measured independently of the predictors, so the avoidance effects are not definitional. The collective-breach result in Experiment 2 is also not circular: the predictor (count of breaches in the prior 5 s) is temporally prior to and empirically distinct from the outcome (the target pedestrian's breach), and the OR = 2.77 is a fitted association, not a quantity constructed from the same outcome. The latent term 'interactional territory' is interpretive framing drawn from prior literature (Lyman & Scott, 1967; Goffman, 1963), not an equation defined from breach data and then re-predicted. There is no load-bearing self-citation or imported uniqueness theorem; the Kaufhold et al. 2025 self-citation is contextual and not used to justify the main results. The same-group-following confound in the contagion analysis is a potential external-validity threat, but it is not a circular reduction by construction, so under the hard rules it does not raise the circularity score.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 1 invented entities

No theoretical free parameters are introduced; all quantities of interest are statistical estimates from fitted GLMMs. The central latent construct, interactional territory, is not independently measured. The paper's support rests on the validity of manual coding, the absence of unmeasured confounds, and standard mixed-model assumptions.

free parameters (1)
  • Prior-breach exposure window = 5 s (robust across 0.9–10 s)
    Hand-set temporal window for the contagion predictor in the GLMM; the paper reports the result is robust across this range, so it is not load-bearing, but it is a modeling choice.
axioms (4)
  • domain assumption Manual video coding of pedestrian attributes and breach outcomes is accurate enough for unbiased inference (gold-standard accuracy 89.8%; Cohen's kappa 0.84–0.92).
    Invoked wherever breach outcomes and exclusions are used; if coding errors correlate with condition, the reported ORs are biased. See Supplementary Methods and Tables S1–S3.
  • domain assumption Actor manipulations (mutual gaze, talk, distance, body orientation, sign/mural) are perceived as social-involvement signals and are not confounded by physical obstruction or pedestrian gaze-avoidance.
    The Exp4 chairs baseline and Exp3 sign/mural contrasts are the paper's evidence for this; if the displays also change the physical geometry or salience of the actors, the attribution to social inference is weakened. See Results, Experiments 3–4.
  • standard math GLMM/brms inference correctly models the hierarchical sampling design (random intercepts for location, date, minute-within-hour) and that observations are conditionally independent.
    All statistical conclusions depend on lme4/brms estimates; violations of distributional or independence assumptions would change the reported ORs. See Methods: Statistical analysis.
  • domain assumption The 5-second prior-breach count and the square-root crowding covariate adequately control for alternative explanations of the contagion effect; no other time-varying confound (e.g., actor fatigue, group membership, flow direction) drives the association.
    The collective-breach result uses these covariates in an observational design; robustness of the 5-s window is shown, but confound control is not experimental. See Results: 'Pedestrians collectively breach interactional territory' and Supplementary.
invented entities (1)
  • Interactional territory ('invisible wall') no independent evidence
    purpose: Latent spatial boundary around a social interaction that pedestrians are hypothesized to detect and avoid breaching.
    The construct is inferred from the same avoidance behavior it is invoked to explain; the paper provides no independent measurement of the territory itself (e.g., physiological, neural, or self-report handle).

pith-pipeline@v1.3.0-alltime-deepseek · 16051 in / 12883 out tokens · 132806 ms · 2026-08-01T09:05:55.591657+00:00 · methodology

0 comments
read the original abstract

Pedestrian dynamics are characterized as complex physical systems constrained by social norms, like personal space. Yet, prior research has ignored the spatial norms imposed by others' social interactions, such as conversations. Unlike physical obstacles, the boundaries of social interactions are latent --- even more so than those of personal space --- as if they are invisible walls, and must be inferred from signs of interactional involvement. In four field experiments with 4,911 participants, we show that pedestrians integrate others' gaze, proximity, body orientation, and talk to avoid interrupting possible interactions. However, pedestrians also collectively violate these spatial norms by walking through an interaction if other pedestrians had already done so. These results demonstrate how physical mobility depends on pedestrians' social computations.

Figures

Figures reproduced from arXiv: 2607.20876 by Anne Elizabeth Clark White, Federico Rossano, Jack Terwilliger, Julia Di Silvestri, Seika Murase.

Figure 1
Figure 1. Figure 1: A diagram of our field experiments. Actors (shown in orange) stood in busy pathways while [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (A) Video frames from experiment 1. (B) Video frames from experiment 2. (C) Breaching probabilities from experiment 1. Dots represent the estimated marginal mean and lines represent 95% confidence intervals. Empirical breaching rates are represented with white diamonds. (D) Breaching rates from experiment 2 plotted by condition and pedestrian walking direction. Conditions are depicted on the right side of … view at source ↗
Figure 3
Figure 3. Figure 3: (A) Video frames from experiment 3 of the 5 ft + mutual gaze + sign (left) and the 5 ft + no gaze + sign condition (right). (B) Breaching probabilities pooled from experiment 1 & 3 at 5 feet. Dots represent the estimated marginal mean and lines represent 95% confidence intervals. Empirical breaching rates are represented with white diamonds. also integrated into navigation decisions. Together, these result… view at source ↗
Figure 4
Figure 4. Figure 4: (A) Video frames from experiment 4. (B) Pedestrian trajectories from experiment 4 colored by pedestrian gender (green for man, purple for woman). Colored circles mark where each participant changed course around the actors. We tested the effects of gender on proxemic breaching. In experiment 1, pedestrians were twice as likely to breach the interactional territory of women than men (OR = 2.12 ± 0.61, z = 2… view at source ↗
Figure 5
Figure 5. Figure 5: On December 7th, 2023 at 12:31:35, over the span of 35.5 seconds, a crowd of 46 pedestrians [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

57 extracted references · 26 canonical work pages

  1. [1]

    Y. Tong, N. W. Bode, The principles of pedestrian route choice. Journal of the Royal Society Interface 19, (2022). doi:10.1098/rsif.2022.0061

    arXiv 2022

  2. [2]

    K. A. Bacik, B. S. Bacik, T. Rogers, Lane nucleation in complex active flows. Science 379, 923--928 (2023). doi:10.1126/science.add8091

    work page doi:10.1126/science.add8091 2023

  3. [3]

    F. Gu, B. Guiselin, N. Bain, I. Zuriguel, D. Bartolo, Emergence of collective oscillations in massive human crowds. Nature 638, 112--119 (2025). doi:10.1038/s41586-024-08514-6

    doi:10.1038/s41586-024-08514-6 2025

  4. [4]

    Helbing, P

    D. Helbing, P. Moln\' a r, Social force model for pedestrian dynamics. Physics Review E 51, 4282--4286 (1995). doi:10.1103/PhysRevE.51.4282

    doi:10.1103/physreve.51.4282 1995

  5. [5]

    G. L. Brown, N. Seethapathi, M. Srinivasan, A unified energy-optimality criterion predicts human navigation paths and speeds. Proceedings of the National Academy of Sciences 118, e2020327118 (2021). doi:10.1073/pnas.2020327118

    work page doi:10.1073/pnas.2020327118 2021

  6. [6]

    Y. Ma, E. W. M. Lee, M. Shi, R. K. K. Yuen, Spontaneous synchronization of motion in pedestrian crowds of different densities. Nature Human Behaviour 5, 447--457 (2021). doi:10.1038/s41562-020-00997-3

    doi:10.1038/s41562-020-00997-3 2021

  7. [7]

    Gilbert, Walking together: A paradigmatic social phenomenon

    M. Gilbert, Walking together: A paradigmatic social phenomenon. MidWest Studies in Philosophy 15, 1--14 (1990). doi:10.1111/j.1475-4975.1990.tb00202.x

    arXiv 1990

  8. [8]

    A. L. Ryave, J. N. Schenkein, ``Notes on the art of walking'' in Ethnomethodology: Selected Readings (Penguin, 1974), pp.\ 265--274

    1974

  9. [9]

    R. M. Alexander, Energetics and optimization of human walking and running: the 2000 Raymond Pearl memorial lecture. American Journal of Human Biology 14, 641--648 (2002). doi:10.1002/ajhb.10067

    work page doi:10.1002/ajhb.10067 2000

  10. [10]

    Murakami, C

    H. Murakami, C. Feliciani, Y. Nishiyama, K. Nishinari, Mutual anticipation can contribute to self-organization in human crowds. Science Advances 7, eabe7758 (2021). doi:10.1126/sciadv.abe7758

    work page doi:10.1126/sciadv.abe7758 2021

  11. [11]

    Karamouzas, B

    I. Karamouzas, B. Skinner, S. J. Guy, Universal Power Law Governing Pedestrian Interactions. Physics Review Letters 113, 238701 (2014). doi:10.1103/PhysRevLett.113.238701

    work page doi:10.1103/physrevlett.113.238701 2014

  12. [12]

    Y. Ma, M. Shi, W. Xie, Z. Hu, Y. Wei, T. Zeng, E. W. M. Lee, Unraveling human crowd dynamics through the foot tracking of pedestrians. Science Advances 11, eadw2688 (2025). doi:10.1126/sciadv.adw2688

    work page doi:10.1126/sciadv.adw2688 2025

  13. [13]

    D. F. Sherry, L. F. Jacobs, S. J. Gaulin, Spatial memory and adaptive specialization of the hippocampus. Trends in Neuroscience 15(8), 298--303 (1992). doi:10.1016/0166-2236(92)90080-R

    work page doi:10.1016/0166-2236(92)90080-r 1992

  14. [14]

    Nathan, W

    R. Nathan, W. M. Getz, E. Revilla, M. Holyoak, R. Kadmon, D. Saltz, P. E. Smouse, A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences 105(49), 19052--19059 (2008). doi:10.1073/pnas.0800375105

    doi:10.1073/pnas.0800375105 2008

  15. [15]

    Strandburg-Peshkin, D

    A. Strandburg-Peshkin, D. R. Farine, I. D. Couzin, M. C. Crofoot, Shared decision-making drives collective movement in wild baboons. Science 348, 1358--1361 (2015). doi:10.1126/science.aaa5099

    doi:10.1126/science.aaa5099 2015

  16. [16]

    Sayin, E

    S. Sayin, E. Cousin-Fuchs, I. Petelski, Y. G\" u nzel, M. Salahshour, C. Y. Lee, L. Li, O. Deussen, G. A. Sword, I. D. Couzin, The behavioral mechanisms governing collective motion in swarming locusts. Science 387, 995--1000 (2025)

    2025

  17. [17]

    S. P. Kaufhold, J. Terwilliger, F. Rossano, Socially situated navigation: social rank and sex influence spatial navigation strategies in Japanese macaques. Proceedings of the Annual Meeting of the Cognitive Science Society vol.\ 47 (2025)

    2025

  18. [18]

    E. T. Hall, The Hidden Dimension (Doubleday & Co, 1966)

    1966

  19. [19]

    Sommer, Studies in personal space

    R. Sommer, Studies in personal space. Sociometry 22, 247--260 (1959). doi:10.2307/2785668

    work page doi:10.2307/2785668 1959

  20. [20]

    C. Zhou, M. C. Miao, X. R. Chen, Y. F. Hu, Q. Chang, M. Y. Yan, S. G. Kuai, Human-behaviour-based social locomotion model improves the humanization of social robots. Nature Machine Intelligence 4, 1040--1052 (2022). doi:10.1038/s42256-022-00542-z

    work page doi:10.1038/s42256-022-00542-z 2022

  21. [21]

    Sommer, Personal Space

    R. Sommer, Personal Space. The Behavioral Basis of Design (Prentice-Hall, 1969)

    1969

  22. [22]

    B. J. Dietrich, M. L. Sands, Seeing racial avoidance on New York City streets. Nature Human Behaviour 7, 1275--1281 (2023). doi:10.1038/s41562-023-01589-7

    work page doi:10.1038/s41562-023-01589-7 2023

  23. [23]

    F. N. Willis Jr., J. A. Gier, D. E. Smith, Stepping Aside: Correlates of Displacement in Pedestrians. Journal of Communication 29, 34--39 (1979). doi:10.1111/j.1460-2466.1979.tb01739.x

    arXiv 1979

  24. [24]

    J. M. Dabbs, N. A. Stokes, Beauty is Power: The Use of Space on the Sidewalk. Sociometry 38, 551--557 (1975). doi:10.2307/2786367

    work page doi:10.2307/2786367 1975

  25. [25]

    R. S. Sobel, N. Lillith, Determinants of Nonstationary Personal Space Invasion. Journal of Social Psychology 97, 39--45 (1975). doi:10.1080/00224545.1975.9923310

    arXiv 1975

  26. [26]

    Kendon, ``The F-formation system'' in Conducting Interaction: Patterns of Behavior in Focused Encounters (Cambridge University Press, 1990), pp.\ 209--237

    A. Kendon, ``The F-formation system'' in Conducting Interaction: Patterns of Behavior in Focused Encounters (Cambridge University Press, 1990), pp.\ 209--237

    1990

  27. [27]

    S. M. Lyman, M. B. Scott, Territoriality: A Neglected Sociological Dimension. Social Problems 15, 236--249 (1967). doi:10.1525/sp.1967.15.2.03a00090

    work page doi:10.1525/sp.1967.15.2.03a00090 1967

  28. [28]

    Goffman, Behavior in Public Spaces (The Free Press, 1963)

    E. Goffman, Behavior in Public Spaces (The Free Press, 1963)

    1963

  29. [29]

    S. C. Levinson, ``The Interaction Engine and Social Life'' in The Interaction Engine: Language in Social Life and Human Evolution (Cambridge University Press, 2025), pp.\ 107--148. doi:10.1017/9781009570343.006

    work page doi:10.1017/9781009570343.006 2025

  30. [30]

    Dingemanse et al., Beyond Single-Mindedness: A Figure-Ground Reversal for the Cognitive Sciences

    M. Dingemanse et al., Beyond Single-Mindedness: A Figure-Ground Reversal for the Cognitive Sciences. Cogn.\ Sci. 47, e13230 (2023). doi:10.1111/cogs.13230

    work page doi:10.1111/cogs.13230 2023

  31. [31]

    C. Zhou, M. Han, Q. Liang, Y. F. Hu, S. G. Kuai, A social interaction field model accurately identifies static and dynamic social groupings. Nature Human Behaviour 3, 847--855 (2019). doi:10.1038/s41562-019-0618-2

    work page doi:10.1038/s41562-019-0618-2 2019

  32. [32]

    Setti, C

    F. Setti, C. Russell, C. Bassetti, M. Cristani, F-Formation Detection: Individuating Free-Standing Conversational Groups in Images. PLOS One 10, e0123783 (2015). doi:10.1371/journal.pone.0123783

    work page doi:10.1371/journal.pone.0123783 2015

  33. [33]

    McMahon, L

    E. McMahon, L. Isik, Seeing social interactions. Trends in Cognitive Science 27, 1165--1179 (2023). doi:10.1016/j.tics.2023.09.001

    work page doi:10.1016/j.tics.2023.09.001 2023

  34. [34]

    Skripkauskaite, I

    S. Skripkauskaite, I. Mihai, K. Koldewyn, Attentional bias towards social interactions during viewing of naturalistic scenes. Quarterly Journal of Experimental Psychology 76, 2303--2311 (2023). doi:10.1177/17470218221140879

    work page doi:10.1177/17470218221140879 2023

  35. [35]

    Vestner, S

    T. Vestner, S. P. Tipper, T. Hartley, H. Over, S. A. Rueschemeyer, Bound together: Social binding leads to faster processing, spatial distortion, and enhanced memory of interacting partners. Journal of Experimental Psychology: General 148, 1251--1268 (2019). doi:10.1037/xge0000545

    work page doi:10.1037/xge0000545 2019

  36. [36]

    X. Ding, Z. Gao, M. Shen, Two Equals One: Two Human Actions During Social Interaction Are Grouped as One Unit in Working Memory. Psychological Science 28, 1311--1320 (2017). doi:10.1177/0956797617707318

    work page doi:10.1177/0956797617707318 2017

  37. [37]

    Brown, S

    P. Brown, S. C. Levinson, Politeness: Some Universals in Language Usage (Cambridge University Press, 1987)

    1987

  38. [38]

    Goffman, On face-work: an analysis of ritual elements in social interaction

    E. Goffman, On face-work: an analysis of ritual elements in social interaction. Psychiatry 18, 213--231 (1955)

    1955

  39. [39]

    Krause, P

    J. Krause, P. Romanczuk, E. Cracco, W. Arlidge, A. Nassauer, M. Brass, Collective rule-breaking. Trends in Cognitive Sciences 25, 1082--1095 (2021). doi:10.1016/j.tics.2021.08.003

    work page doi:10.1016/j.tics.2021.08.003 2021

  40. [40]

    J. J. Faria, S. Krause, J. Krause, Collective behavior in road crossing pedestrians: the role of social information. Behavioral Ecology 21, 1236--1242 (2010). doi:10.1093/beheco/arq141

    work page doi:10.1093/beheco/arq141 2010

  41. [41]

    Keizer, S

    K. Keizer, S. Lindenberg, L. Steg, The Spreading of Disorder. Science 322, 1681--1685 (2008). doi:10.1126/science.1161405

    work page doi:10.1126/science.1161405 2008

  42. [42]

    Garfinkel, Studies in Ethnomethodology (Prentice-Hall, 1967)

    H. Garfinkel, Studies in Ethnomethodology (Prentice-Hall, 1967)

    1967

  43. [43]

    Rossano, ``Gaze in Conversation'' in The Handbook of Conversation Analysis J

    F. Rossano, ``Gaze in Conversation'' in The Handbook of Conversation Analysis J. Sidnell, T. Stivers, Eds. (Wiley, 2012) pp.\ 308--329. doi:10.1002/9781118325001.ch15

    work page doi:10.1002/9781118325001.ch15 2012

  44. [44]

    N. H. Anderson, Foundations of Information Integration Theory (Academic Press, 1981)

    1981

  45. [45]

    Kendon, ``Description of Some Human Greetings'' in Conducting Interaction: Patterns of Behavior in Focused Encounters (Cambridge University Press, 1990), pp.\ 153--208

    A. Kendon, ``Description of Some Human Greetings'' in Conducting Interaction: Patterns of Behavior in Focused Encounters (Cambridge University Press, 1990), pp.\ 153--208

    1990

  46. [46]

    Papeo, N

    L. Papeo, N. Goupil, S. Soto-Faraco, Visual Search for People Among People. Psychological Science 30, 1483--1496 (2019). doi:10.1177/0956797619867295

    work page doi:10.1177/0956797619867295 2019

  47. [47]

    N. J. Emery, The eyes have it: the neuroethology, function and evolution of social gaze. Neuroscience & Biobehavioral Reviews 24, 581--604 (2000). doi:10.1016/S0149-7634(00)00025-7

    doi:10.1016/s0149-7634(00)00025-7 2000

  48. [48]

    von Uexk\" u ll, ``A stroll through the worlds of animals and men: A picture book of invisible worlds'' in Instinctive Behavior: The Development of a Modern Concept C

    J. von Uexk\" u ll, ``A stroll through the worlds of animals and men: A picture book of invisible worlds'' in Instinctive Behavior: The Development of a Modern Concept C. H. Schiller, Ed. and Transl. (International Universities Press, 1957), pp.\ 5--80

    1957

  49. [49]

    Rios-Martinez, A

    J. Rios-Martinez, A. Spalanzani, C. Laugier, ``Understanding human interaction for probabilistic autonomous navigation using Risk-RRT approach'' in 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (2011), pp.\ 2014--2019. doi:10.1109/IROS.2011.6094496

    arXiv 2011

  50. [50]

    Ettinger et al., Large Scale Interactive Motion Forecasting for Autonomous Driving: The Waymo Open Motion Dataset

    S. Ettinger et al., Large Scale Interactive Motion Forecasting for Autonomous Driving: The Waymo Open Motion Dataset. arXiv:2104.10133 [cs] (2021). doi:10.48550/arXiv.2104.10133

    Pith/arXiv arXiv doi:10.48550/arxiv.2104.10133 2021

  51. [51]

    Bates, M

    D. Bates, M. M\" a chler, B. Bolker, S. Walker, Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software 67, 1--48 (2015). doi:10.18637/jss.v067.i01

    doi:10.18637/jss.v067.i01 2015

  52. [52]

    Lenth, emmeans: Estimated marginal means, aka least-squares means

    R. Lenth, emmeans: Estimated marginal means, aka least-squares means. R package version 1.11.2-8 (2025)

    2025

  53. [53]

    R. W. Bohannon, A. Williams, Normal walking speed: a descriptive meta-analysis. Physiotherapy 97, 182--189 (2011). doi:10.1016/j.physio.2010.12.004

    work page doi:10.1016/j.physio.2010.12.004 2011

  54. [54]

    C. Y. Wang, A. Bochkovskiy, H. Y. M. Liao, ``YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors'' in IEEE/CVF Conference on Computer Vision and Pattern Recognition (2023), pp.\ 7464--7475. doi:10.1109/CVPR52729.2023.00721

    arXiv 2023

  55. [55]

    Bewley, Z

    A. Bewley, Z. Ge, L. Ott, F. Ramos, B. Upcroft, ``Simple online and realtime tracking'' in 2016 IEEE International Conference on Image Processing (ICIP) (2016), pp.\ 3464--3468. doi:10.1109/ICIP.2016.7533003

    arXiv 2016

  56. [56]

    P. C. B\" u rkner, brms: An R Package for Bayesian Multilevel Models Using Stan. Journal of Statistical Software 80, 1--28 (2017). doi:10.18637/jss.v080.i01

    doi:10.18637/jss.v080.i01 2017

  57. [57]

    Goffman, Relations in Public (Transaction Publishers, 2009)

    E. Goffman, Relations in Public (Transaction Publishers, 2009)

    2009