CoChat

Using Wearable Sensors to Measure Interpersonal Synchrony in Actors and Audience Members During a Live Theatre Performance

Open access
Sun, Yanke, Greaves, Dwaynica A., Orgs, Guido, de C. Hamilton, Antonia F., Day, Sally, Ward, Jamie A. · 2023 · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies

The DOI matches its Crossref record; 71 of the 71 references it cites were found in a registry or an index; 19 of the 25 sources its claims cite were checked for support.

Abstract

Here we collaborated with Flute Theatre to investigate interpersonal synchrony between actors-actors, actors-audience and audience-audience within a live theatrical setting. Our 28 participants consisted of 6 actors and 22 audience members, with 5 of these audience members being audience participants in the show. The performance was a compilation of acting, popular science talks and demonstrations, and an audience participation period. Interpersonal synchrony was measured using inertial measurement unit (IMU) wearable accelerometers worn on the heads of participants, whilst audio-visual data recorded everything that occurred on the stage. Participants also completed post-show self-report questionnaires on their engagement with the overall scientists and actors performance. Cross Wavelet Transform (XWT) and Wavelet Coherence Transform (WCT) analysis were conducted to extract synchrony at different frequencies, pairing with audio-visual data. Findings revealed that XWT and WCT analysis are useful methods in extracting the multiple types of synchronous activity that occurs when people perform or watch a live performance together. We also found that audience members with higher ratings on questionnaire items such as the strength of their emotional response...

Overview

CoChat Summary

TL;DR

This paper presents a proof-of-principle live-theatre study showing that head-worn wearable accelerometers, video, and self-report can capture interpersonal synchrony among 6 actors and 22 audience members during a public performance. Using XWT and WCT, the authors found synchrony patterns that varied by performance section, and some audience ratings were positively correlated with actor-audience synchrony.

Key findings

  • The study analysed 28 participants: 6 actors and 22 audience members, including 5 audience participants and 17 audience spectators.
  • Sensor syncing was improved from raw mean errors of 1534.0 ms, 3772.5 ms, and 3107.6 ms at three sync events to 18.2 ms, 13.3 ms, and 16.9 ms after shift-and-interpolate preprocessing.
  • In the acting and science-talk sections, actor-actor WCT power was “dominated for most of the activity periods,” while audience-audience and actor-audience WCT powers were described as similar and relatively stable for the 0.5–60 s range.
  • In the post-show survey analysis on 15 audience members, long-period XWT synchrony with actors correlated positively with Emotional Response (r = 0.6284, p = 0.0121) and Empowered (r = 0.5887, p = 0.0209), with a marginal correlation for Inhabit (r = 0.4697, p = 0.0773).
  • Paired t-tests showed higher ratings for the scientists’ performance than the actors’ performance on Emotional Response (3.33 vs 4.07, p = 0.006), Uplifted (3.53 vs 4.2, p = 0.007), Empowered (2.8 vs 3.53, p = 0.028), and Social Issues (3.13 vs 4.27, p = 0.006).

Methods

  • Proof-of-principle live theatre event over two consecutive evenings in May 2019 at Bloomsbury Theatre, London; only the second night was analysed because the first night had missing video and labelling data.
  • Participants wore MetaMotionR headband accelerometers recording 3-axis acceleration at about 25 Hz; the dataset included actors, audience participants, and audience spectators.
  • The authors synchronised sensor streams using multiple shake events and interpolation, then computed XWT and WCT, including time-averaged and scale-averaged power, interaction matrices, and Pearson correlations with post-show survey responses.

Limitations

  • The paper states that only one night’s performance was analysed because technical problems caused much of the first night’s video and labelling data to be missing.
  • The paper states that there was no sensor data from the non-actor presenters, so correlations involving audiences and scientists / artistic director could not be computed.
  • Evident from the text: only 15 of 18 returned audience surveys were usable for the actor-audience correlation analysis due to missing data.
  • Evident from the text: only one camera faced the stage, and the authors note this left insufficient video information about the audience and limited discussion of audience-pair interactions.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1145/3580781

Cites

71 of 71 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Four moments during the performance (left to right): Act 1 (with visualisation of an actor's on-stage live brain scan), Science talk 1 (actors not on stage), Act 2 (note sensor-embedded headbands worn by the actors), Science talk 2 (actors on stage, also shows visualisation altered by live audience movement).
Fig. 1 Four moments during the performance (left to right): Act 1 (with visualisation of an actor's on-stage live brain scan), Science talk 1 (actors not on stage), Act 2 (note sensor-embedded headbands worn by the actors), Science talk 2 (actors on stage, also shows visualisation altered by live audience movement).From the PDF
Image 2
From the PDF
Fig. 3 Synchronised acceleration magnitude of actors, participants and spectators. The black dotted lines present different sections of the show, and the red dotted lines show the synchronisation events.
Fig. 3 Synchronised acceleration magnitude of actors, participants and spectators. The black dotted lines present different sections of the show, and the red dotted lines show the synchronisation events.From the PDF
Fig. 4 Cross wavelet transform analysis. (a) and (b) are part of the head acceleration data from actor 0 and actor 1. (c) and (d) are the Continuous wavelet transform outputs of (a) and (b). Combing the outputs (c) and (d) gives the cross wavelet transform (e) and wavelet coherence (f) time-frequency spectral. Scale-averaged power (h) and time-averaged power (g) are calculated from cross wavelet transform power and wavelet coherence power.
Fig. 4 Cross wavelet transform analysis. (a) and (b) are part of the head acceleration data from actor 0 and actor 1. (c) and (d) are the Continuous wavelet transform outputs of (a) and (b). Combing the outputs (c) and (d) gives the cross wavelet transform (e) and wavelet coherence (f) time-frequency spectral. Scale-averaged power (h) and time-averaged power (g) are calculated from cross wavelet transform power and wavelet coherence power.From the PDF
Fig. 5 Interaction matrices using XWT power for different sections of the show.
Fig. 5 Interaction matrices using XWT power for different sections of the show.From the PDF
Image 6
From the PDF
Fig. 7 Time-averaged WCT power is shown for different sections of the show. The means of time-averaged WCT power of appropriate groups of people for the different sections are calculated, i.e. actors, audience as a whole, audience participants and audience spectators. For example, the game sections have the audience separated into participants on the stage and spectators, which is not the case for Acts and Science talks. Periods are plotted in log scale with the range from 0.5s to 60s.
Fig. 7 Time-averaged WCT power is shown for different sections of the show. The means of time-averaged WCT power of appropriate groups of people for the different sections are calculated, i.e. actors, audience as a whole, audience participants and audience spectators. For example, the game sections have the audience separated into participants on the stage and spectators, which is not the case for Acts and Science talks. Periods are plotted in log scale with the range from 0.5s to 60s.From the PDF
Fig. 8 The mean of scale-averaged power of groups during the games sections is shown in (a) using both XWT and (b) WCT power for comparison. The scale-averaged XWT power is calculated between actors and participants, while the scale-averaged WCT power is computed between participants, actors & spectators, participants & spectators, and spectators. The period range used for average is from 5s to 60s. Vertical lines show the important timestamps during game sections.
Fig. 8 The mean of scale-averaged power of groups during the games sections is shown in (a) using both XWT and (b) WCT power for comparison. The scale-averaged XWT power is calculated between actors and participants, while the scale-averaged WCT power is computed between participants, actors & spectators, participants & spectators, and spectators. The period range used for average is from 5s to 60s. Vertical lines show the important timestamps during game sections.From the PDF
Fig. 9 The mean of scale-averaged power between audiences during Act 2. The period range chosen for average is from 0.5s-5s in order to observe more details interactions.
Fig. 9 The mean of scale-averaged power between audiences during Act 2. The period range chosen for average is from 0.5s-5s in order to observe more details interactions.From the PDF

Verification

Claims the paper makes

12 Valid · 5 Not valid · 4 Not checked · 118 Own results and interpretations

48 more claims

Create a free account to see all 139 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 54
  • Probably not a substantial claim. 56
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentt(14) = -3.214reported p = .006 · recomputed p = .006

    “t(14) = -3.214, p = .006”

  • Consistentt(14) = -3.162reported p = .007 · recomputed p = .007

    “t(14) = -3.162, p = .007”

  • Consistentt(14) = -2.442reported p = .028 · recomputed p = .028

    “t(14) = -2.442, p = .028”

  • Consistentt(14) = -3.238reported p = .006 · recomputed p = .006

    “t (14) = -3.238, p = .006”

4 of 4

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Citations

Cites

Show all 71 references

Research works printed but not linked: 1

  1. Dani Karmakar. 2013. Theatre and Communication: Relation Between Actor and Audience. Global Media Journal, Indian edición (2249-5835) (2013).

Cited by

See all 35

Metadata

Authors
Sun, Y., Greaves, D. A., Orgs, G., de C. Hamilton, A. F., Day, S., & Ward, J. A.
Affiliations
  • University College London (GB)
  • Goldsmiths University of London (GB)
  • Institute of Cognitive Neuroscience, University College London, London, UK
Venue
Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies
Published
2023-03-27
Volume
7
Issue
1
Pages
1-29
ISSN
2474-9567 (ISSN-L)
OpenAlex
W4361758412
Cited by
35
References
53
Citations by year
2026: 10 · 2025: 8 · 2024: 13 · 2023: 4
FWCI
10.38
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies (publishedVersion)
  • Goldsmiths (University of London) (submittedVersion)
  • UCL Discovery (University College London) (submittedVersion)
Funders
  • Leverhulme Trust
  • European Commission
Topics
  • Social Sciences › Psychology › Social Psychology › Action Observation and Synchronization
  • Social Sciences › Arts and Humanities › Visual Arts and Performing Arts › Theatre and Performance Studies
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
Keywords
interpersonal synchrony, audience engagement, actor-audience interaction, wearable accelerometers, inertial measurement unit, cross-wavelet transform, wavelet coherence, public engagement with science, social interaction

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Using Wearable Sensors to Measure Interpersonal Synchrony in Actors and Audience Members During a Live Theatre Performance. https://app.cochat.ai/papers/10.1145/3580781
The paper
Sun, Y., Greaves, D. A., Orgs, G., de C. Hamilton, A. F., Day, S., & Ward, J. A. (2023). Using Wearable Sensors to Measure Interpersonal Synchrony in Actors and Audience Members During a Live Theatre Performance. Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies. https://doi.org/10.1145/3580781

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.