Building blocks of suspense: subjective and physiological effects of narrative content and film music

The study examines how narrative content and non-diegetic music contribute to the experience of suspense in film. Suspense is characterized by uncertainty and psychological tension about outcomes for valued protagonists and is influenced not only by plot information but also by formal features like sound and music. The authors adopt a broad framework in which tension and suspense are affective states associated with conflict and uncertainty, a yearning for resolution, and future-directed expectations, which can be evoked by both narrative cues and musical features. The work addresses gaps in understanding the dynamic, bodily (physiological) underpinnings of suspense and potential divergences between subjective judgments and physiological responses.

Prior research shows music can induce tension and shape expectations via features such as contour, harmony, tempo, pitch, and volume, paralleling narrative suspense mechanisms. Physiological correlates of suspense have been studied using heart rate, electrodermal activity (SCL), temperature, and pulse volume amplitude (PVA), with mixed findings. SCL often increases with uncertainty or its resolution; heart rate may rise with suspense and drop upon resolution, but HR is non-specific and also reflects attention and cognitive load. PVA has shown sensitivity to suspenseful scenes. In music perception, tension has been linked to vasoconstriction (lower PVA) and mixed SCL patterns; chill-inducing music can decrease PVA and sometimes SCL, with faster heart rate. Studies on film-music interactions indicate music can elevate arousal and shape interpretation, though the specific contributions of audio-only presentations remain underexplored. Overall, different physiological measures may be sensitive to different modalities and stimulus components, motivating integrated assessment of narrative content and non-diegetic music.

Design: Between-subjects with three conditions: audio-visual (AV; original film with sound and music), video-only (silent film), and audio-only (audio track with black screen). Stimulus: The suspenseful short film Love Field (~5:30 min), featuring distinct plot phases and contrasting musical moods (tense vs. cheerful). A 2:04 min nature relaxation video with relaxing music preceded the film in all conditions to establish baseline. Stimulus annotation: Four trained annotators performed frame-accurate coding of onsets of objects, people, and sounds, resolving discrepancies jointly. Based on events, the team defined four plot phases: (1) diffuse tension (ambiguous visuals + tense music), (2) story framing (appearance of clues—phone, money, blood, knife—+ tense music, suggesting violent crime), (3) anticipatory suspense (immersion in the narrative framework with tense music), and (4) resolution/relief (reveal of childbirth, positive outcome + cheerful music). Recurrent motifs included a crow and the farmer. Measures: Subjective suspense via continuous response measurement (CRM) on a 9-point slider from −4 (low) to +4 (high), updated with keyboard inputs. Physiological measures recorded with a Lightstone iom1 device: photoplethysmography (PPG) on the middle finger for inter-beat interval (IBI; inverse of heart rate) and pulse volume amplitude (PVA), and skin conductance level (SCL) via electrodes on adjacent fingers. Data acquisition: Physiological data sampled at 30 Hz via custom C++ software; a parallel Python program synchronized CRM and event markers. All channels saved with shared timestamps. HeartPy was used for PPG peak detection; custom tools supported preprocessing and alignment. Procedure: Participants were randomly assigned to one of the three conditions, seated individually, instrumented with sensors on the non-dominant hand, trained on CRM, and completed the baseline video followed by the experimental stimulus. Total session ~7.5 minutes. Debrief and course credit provided. Analysis: Repeated-measures ANOVAs tested effects of condition (between-subjects) and plot phase (within-subjects) on CRM, SCL, PVA, and IBI. Frame-by-frame pairwise t-tests examined temporal dynamics across conditions. Event-related IBI analyses included (a) stimulus-based comparisons of mean IBI in 5 s windows before vs. after five critical story-framing events; and (b) response-based identification of local IBI differences across groups around salient narrative cues.

- Subjective vs physiological dissociation: Continuous self-reports (CRM) and physiological indices often diverged, indicating different processing pathways. - Music induces suspense without visuals: Tense music alone (audio-only) elevated CRM suspense even without story content; video-only CRM stayed low during diffuse tension until visual clues appeared. - Modality-driven arousal differences: Audio-only condition showed the highest tonic arousal (higher SCL, PVA changes), while video-only showed the lowest arousal; AV typically aligned with audio patterns in arousal measures. - SCL patterns: Strong onset response in all conditions; SCL decreased during diffuse tension in video-only and AV (more in video-only), remained relatively unchanged in audio-only; during story framing, AV SCL rose and peaked around the bloody knife cue, with AV and audio exceeding video thereafter; limited change during resolution/relief. - PVA patterns: Video-only PVA stayed slightly above baseline (vasodilation). In AV and audio-only, PVA correlated positively with CRM. Around the knife event (~4.5 min), PVA showed a short-lived decrease (vasoconstriction) indicating local tension, followed by swift increase (relief), with audio-only peaking slightly earlier (driven by musical climax and brief silence preceding the knife sound). - IBI (heart rate) dynamics: Short-term HR decelerations occurred immediately before salient narrative clues (e.g., dramatic pauses, close-ups, faces), interpreted as increased attention/resource allocation, followed by acceleration after emotionally salient events. - Statistics (selected): • CRM: Segment main effect F≈196.25, p<0.001, η²≈0.641; Condition main effect F≈6.15, p<0.003, η²≈0.111; interaction significant. • SCL: Segment main effect F≈8.24, p≈0.014; Segment×Condition interaction p≈0.010; audio vs video showed most pronounced differences. • PVA: Segment×Condition interaction F≈4.06, p≈0.005; modality differences driven by presence of audio. • IBI: Segment main effect F≈10.82, p<0.001; event-related ANOVA for story-framing events F(3.57,335.38)=4.16, p=0.002, η²=0.04; condition effect F(2,99)=4.55, p=0.011, η²=0.08. • Event-related one-sample t-tests (pre vs post, 5 s windows): AV condition: event #3 (bloody fabric) t(33)=3.62, p<0.001; event #4 (twitching foot) t(33)=2.54, p=0.02. Audio-only: event #4 (woman’s scream) t(34)=2.54, p=0.02.

Findings show that non-diegetic music is a strong driver of both subjective suspense and physiological arousal, sometimes independent of visual narrative cues. Audio-only presentations elevated tonic arousal compared to video-only, suggesting a modality effect whereby sound/music globally increases arousal onto which narrative events are layered. Short-term IBI decelerations preceding salient cues support an attentional allocation mechanism, while SCL and PVA indexed more sustained arousal levels with limited short-lived fluctuations except around the knife cue. The divergence between CRM and physiological measures implies parallel processing streams: deliberative, feature-based subjective judgments influenced by genre expectations and instruction to rate suspense, and more automatic physiological responses to formal features (music) and salient narrative events. The results support a multidimensional perspective differentiating suspense (higher-level, narrative-bound, expectation-laden) from tension (lower-level, automatic arousal), with music capable of inducing tension that can bolster feelings of suspense or operate in parallel with narrative cues.

The study demonstrates that non-diegetic music can independently elicit feelings of suspense and elevate arousal, and that audio modality generally produces higher physiological arousal than video-only presentations. Objective physiological measures and subjective CRM ratings capture distinct facets of the suspense experience, indicating partially dissociable processing pathways. IBI deceleration tracks momentary attention to critical narrative clues, while SCL and PVA reflect broader arousal modulations with limited short-term reactivity. Contributions include: (1) an integrated multimodal measurement approach (CRM, IBI, SCL, PVA) coupled with fine-grained stimulus annotation; (2) evidence for strong music-driven suspense effects; and (3) clarification of measure-specific sensitivities to narrative versus formal features. Future research should test additional narrative structures (e.g., multiple micro-resolution arcs), disentangle general auditory modality effects from music-specific effects, examine mental imagery contributions, and further refine the conceptual distinction between suspense and tension and their physiological signatures.

- Single short film stimulus with a largely linear, revelatory suspense structure limits generalizability; results may differ for narratives with multiple micro-tension/relief cycles. - Modality effects (audio vs video) may reflect broader auditory processing differences not isolated to music; the study did not include non-musical auditory controls. - Mixed and sometimes contradictory sensitivities across physiological measures complicate interpretation; PVA showed clear local effects primarily around one key event (knife), while SCL reflected tonic differences. - Potential influence of instruction to continuously rate suspense (CRM) on deliberative judgments; prior media exposure/genre knowledge may bias self-reports. - Equipment and sampling constraints (PPG-derived IBI/PVA, commercial biofeedback device) may affect signal specificity and resolution; sample size details are not reported in the excerpt. - Frame-by-frame multiple comparisons were uncorrected in visualization steps, which may increase Type I error risk for local effects.