Walking in Sync: Two is Company, Three’s a Crowd

Kleomenis Katevas
Jul 11, 2018 · 4 min read
Photo by Dan Newman on Unsplash

Crowds have several common behavioural patterns. For example, pedestrians inside a crowd tend to move together as a unit; with group members walking at the same speed, following the same trajectories and quickly reform after they become separated.

An interesting phenomenon in small groups is the synchronisation of stepping that often occurs in people walking side-by-side. This mutual engagement can either occur intentionally (e.g. a public procession that occurs by imitation) or unintentionally when that happens subconsciously, a behaviour known in psychology as mirroring. Previous studies suggest that the reason behind an unintentional sync in walking is that they share a common feeling of unity and close relationship.

In the next example, as recorded by the Royal Institution for the needs of last year’s Christmas Lectures “The Language of Life”, two children were asked to walk for a few minutes while discussing their favourite movie. At the same time, a smartphone placed in their left pocket was recording data using the device’s accelerometer sensor. After a few seconds, the children unconsciously synchronised their steps, reaching the moment that is shown in the next animation:

Tweet from the Royal Institution showing two children unconsciously synchronise their steps

The following visualisation shows the magnitude of their acceleration over time. You can see that the two signals are initially out of phase, but start overlapping after a few seconds showing that the children have the same walking rhythm.

Accelerometer signal from the two children collected using smartphones placed in their pocket

How does this interesting phenomenon happen and what triggers it? In most cases, a group walk also includes engagement in a conversation. People tend to look at each other while walking, interact with gestures and at the same time have a turn-taking conversation. We believe that these non-verbal social signals such as the gaze, head orientation and gestures influence the walking patterns and synchronisation of these walking groups. But does this phenomenon exist in all types of walking groups?

Group members of more than two people usually have a V-shaped walking formation that facilitates social interactions between members. According to Battersby and Healey (2010), gaze and head orientation become problematic in conversations between more than two participants. The reason is that the gaze can only focus on a single person at a time, especially when the participants are walking side-by-side.

To investigate this phenomenon further we conducted an experiment where two participants were asked to walk in a park for 12 minutes (Scenario 1) while discussing a topic of their choice. A smartphone device configured to collect accelerometer data in 100Hz sampling rate was placed on each participant’s left pocket. The same walk was repeated again by injecting a third person between the original two participants (Scenario 2).

Bivariate correlations of the acceleration signals between the participants in Scenario 1 show a positive correlation (Pearson Correlation = 0.23, p < 0.01). Smaller correlations exist when comparing the signals in Scenario 2: 0.02 (p < 0.01) for P1 and P2, 0.11 (p < 0.01) for P1 and P3, and 0.03 (p < 0.01) for P2 and P3. Similar correlation values were identified when we performed a cross-correlation with a maximum lag of 1 second, indicating that the two participants in Scenario 1 have the highest correlation and with the minimum lag in their stepping (lag = 0.01s). Since both analyses P1 × P2 and P2 × P3 show much smaller correlation and high lag in the sync, we can assume that P2 is the least synchronised person of the group.

The same synchronisation was observed visually by the researcher in Scenario 1 during the experiment, but only in cases were two of the people were discussing together while the other was not paying attention to the conversation. A likely explanation of this correlation is the engagement of the conversation that affects the participants to synchronise their steps unconsciously. When three people are walking together, the conversation is taking place in turns between two people, changing their gaze so that each person looks at each other. This influences their walking, making their steps to correlate again in each turn and break the synchronisation between the participants of the past turn.


Eventual gait synchronisation between two individuals while walking and talking with each other has been shown to be an indicator of agreeableness and companionship. The inferred physical signal from this subconscious phenomenon can potentially be an indicator of cooperation or relation between two individuals.

In order to investigate this phenomenon, we analysed the accelerometer data from individuals walking in different formations, while having conversations in groups of two or three. We recorded and processed high frequency data from the gait motions and have cross-correlated these data. From our results, we have empirically confirmed previously published observatory studies: the non-verbal social signals such as the gaze, head orientation and gestures between individuals plays a significant factor in gait synchronisation between two individuals walking together. A third person being involved in this formation can distort the synchronisation, unless they are not actively engaged in the conversation between a pair.

Watch the 2nd part of the Christmas Lectures 2007 “The Language of Life: Silent Messages” featuring our work.

This work was presented at the 2nd Workshop on Physical Analytics (WPA) in conjunction with ACM MobiSys 2015 in May 2015.

Walking in Sync: Two is Company, Three’s a Crowd.
Kleomenis Katevas, Hamed Haddadi, Laurissa Tokarchuk, Richard G. Clegg.
2nd Workshop on Physical Analytics (WPA) in conjunction with ACM MobiSys 2015, Florence, Italy, May 2015.

Kleomenis Katevas

Written by

Research Associate at Imperial College London, UK.