Understanding emotional and behavioural responses through physiological signal analysis recorded during theatrical performance

1 Introduction and Research Background

• Theatre and other cultural interventions are presented as contexts for emotional expression, social connection, and psychological well-being.Stories can engage audiences through narrative and, in live theatre, actors’ physical presence and real-time interaction.Source: §1 · pp. 1–2
• Physiological signals can track dynamic emotional responses, but HR and SCL indicate activation rather than uniquely identifying discrete emotions.Their interpretation depends on emotional intensity and context; responses to amusement or joy can vary with the stimulus.Source: §1 · p. 2
• Prior work has used physiological signals to study engagement across settings and to develop supervised and unsupervised emotion-recognition systems.Related approaches have also combined physiological signals with personality traits.Source: §1 · pp. 2–3
• The study examines physiological response patterns during a live performance without explicit labels, asking how they vary by scene, align across participants, and reflect overall intensity.It also asks how effectively z-score intensity analysis and visualization capture arousal and engagement patterns.Source: §1 · p. 3

2 Data Collection and Experimental Setup

• The study was designed to capture real-time audience reactions, and the experimental protocol received approval from the Human Sciences Ethics Committee of the Universities in Satakunta, Finland.The approval ID is 16.05.2025.Source: §2 · p. 3

2.1 Theatrical Performance Overview

• The performance, Isin tyttö (Daddy’s Girl), lasted about 125–130 minutes across two parts, with a 20-minute break, and included scenes intended to evoke amusement, empathy, and suspense.Selected key events included changes in tone, intensity, or expected engagement, with some involving flashing lights or loud sounds.Source: §2.1 · pp. 3–4

2.2 Participant Demographics

• The study recruited 17 adults, and all participants were Finnish; ages ranged from 20 to 68, with 65% female and 35% male.Participants had no diagnosed long-term illness affecting central nervous system functioning and provided informed consent.Source: §2.2 · p. 5

2.3 Data Collection Procedure

• Non-invasive ECG and GSR were recorded continuously during the performance and synchronized with video to align physiological responses with scenes and events.Sensors and video were paused during the 20-minute break; ECG measured heart activity, while GSR measured skin conductance.Source: §2.3 · pp. 5–6

3 Methodology

• ECG and GSR recordings were cleaned and processed into smoothed HR and tonic-component SCL measures for analysis of physiological activation.The methodology used NeuroKit2 for ECG cleaning and R-peak detection, and for EDA cleaning and decomposition.Source: §3 · p. 6
• Aligned HR and SCL plots were visually examined for peaks or sustained increases, while two researchers manually annotated key events using synchronized performance video.The annotations segmented the physiological data into event-specific intervals for subsequent analysis.Source: §3 · pp. 7–8
• For scene-level intensity, HR and SCL were standardized as participant-specific z-scores and averaged for each time segment.The index represents deviation from each participant’s baseline; higher positive values indicate greater physiological arousal.Source: §3 · p. 8
• Synchronized video and physiological signals were used together to identify peak arousal patterns in the context of narrative events.The framework supports interpretation of responses in both physiological and contextual dimensions.Source: §3 · p. 8

4 Visual Analysis and Key Inferences

• The analysis combines participant-level and scene-wise observations with mean and SD curves, heatmaps, and intensity plots to examine response consistency and peaks.Physiological signals are related to time-aligned performance scenes.Source: §4 · pp. 8–9, §4.3 · p. 11

4.1 Participant-level observation

• Eight of 17 participants had coinciding HR and SCL peaks during amusement scenes; eight also showed medium-to-high SCL peaks at sudden reactions and humorous moments.Three participants showed pronounced responses during narrative climaxes, while four had little or no substantial change in overall SCL.Source: §4.1 · p. 9
• Some participants showed frequent HR fluctuations during amusing, auditory-stimulating scenes, while others had HR drops at sudden stimuli or prolonged HR and SCL elevations.The authors suggest laughter-related breathing changes and dynamic arousal may contribute to HR fluctuations; they also suggest sound or lighting sensitivity may relate to sustained activation.Source: §4.1 · p. 10

4.2 Scene-wise physiological response

• In Part 2, surprise, intense, and sudden-reaction scenes produced the strongest activation, while empathy scenes showed moderate HR and low SCL.In Part 1, amusement scenes produced moderate-to-high HR with variable SCL; empathy responses were relatively low to moderate in HR with stable SCL.Source: Table 2 · p. 10, §4.2 · p. 11
• For one participant, sudden reaction elicited a concurrent HR and SCL increase, while empathy responses were comparatively stable and amusement SCL varied.The comparison illustrates scene-specific responses in Part 2.Source: §4.2 · p. 11, Figure 4 · p. 12

4.3 Physiological inference and intensity analysis

• Part 1 HR was generally higher than Part 2, while Part 2 SCL began higher before declining to levels comparable with Part 1.The initial Part 2 SCL increase coincided with loud sounds, bright lighting, and an intense scene.Source: §4.3 · p. 13

2. Normalized z-score based heatmaps

• Part 1 heatmaps show more consistent shared arousal across participants, whereas Part 2 shows greater inter-individual variability and synchronized HR/SCL increases at key moments.Part 2 synchronized arousal is prominent at the beginning, during an intense scene, and around the narrative climax.Source: §2 · p. 13, Figure 7 · p. 16, Figure 8 · p. 17
• In the heatmaps, warmer colors mark positive deviations from baseline and cooler colors mark lower-than-baseline activity; some participants show HR/SCL increases in the same time windows.Other Part 2 participants show elevated SCL with stable HR, or the opposite pattern.Source: §2 · p. 13, Figure 7 · p. 16, Figure 8 · p. 17
• The intensity index averages z-scored HR and SCL; Part 1 peaks most during empathy, while Part 2 has peaks during amusement and climax-associated empathy scenes.Part 2 intensity values were lower overall than Part 1, and emotional labels were used because scene boundaries did not align perfectly across participants.Source: §2 · p. 18

• The analysis combined ECG-derived HR and GSR-derived SCL from 17 participants with demographic, visual, participant-level, and scene-wise analyses.Source: §5 · p. 18
• Dramatic, surprising, and intense scenes produced greater physiological activation, often with concurrent HR and SCL increases, while empathy-driven scenes showed lower, steadier responses.Source: §5 · p. 20
• Amusement scenes elicited moderate-to-high responses, especially with loud sounds or sudden reactions, and most participants showed coinciding HR and SCL peaks in such moments.Source: §5 · p. 20
• Z-score intensity analysis and heatmaps concentrated activation around specific emotional and dramatic events; Fig. 9 presents intensity across scenes in both parts.Source: Figure 9 · p. 19, §5 · p. 20
• The authors identify HR and SCL analysis as potentially applicable to theatre design, audience research, affective analysis, and interactive performance evaluation [25].Source: §5 · p. 20
• The findings are preliminary; more comprehensive quantitative analysis is needed to validate and characterize these response patterns.Source: §5 · p. 20