The participants of the experiment were mainly recruited through the public recruitment plan for experimental volunteers at Qingdao University of Technology. The recruitment targets were college students and postgraduate students currently enrolled. Students sign up voluntarily. The main reason for choosing college students as the research group is that young people at this stage are generally healthier. They are more sensitive to sounds and have a stronger understanding of the acoustic environment. The study was conducted in accordance with the Declaration of Helsinki and approved by the University Ethics Committee (IRB number: QUT-HEC-2024023). All participants received written and verbal information about the study and provided written informed consent before data collection. Participants were informed of their right to withdraw at any time without penalty, and all data were anonymized prior to analysis. We calculated the sample size through G.Power software (Araujo et al., 2016). Under the condition that the effect size was 0.5(medium), a=0.05, and the power was 0.8, the sample size was at least 24. Therefore, 26 participants were recruited for the experiment. The participants’ information is shown in Table 1.

All sound stimuli used in the experiment were recorded by the authors using professional audio recording equipment. The recorder was positioned approximately 1.5 m above ground level, and a representative 5-min segment was selected from each recording for use as a restorative sound stimulus. To maximize the purity of the natural sound sources, recordings were conducted at sparsely populated natural sites with minimal anthropogenic noise. Because recordings were obtained under different environmental conditions, the original sound pressure levels (SPLs) varied across stimuli. To minimize potential confounding effects of SPL on the experimental outcomes, all sound-type stimuli were presented at a standardized SPL. Table 2 summarizes the sound stimuli and reports the preset and measured SPL values for each condition.

The subjective assessment in this study was conducted using the Perceived Restorativeness Scale (PRS). Each item was rated on a 5-point Likert scale. To reduce participant fatigue, four representative items were selected for each PRS subscale. Details of the selected items are provided in Appendix A.

Physiological responses were assessed using electrocardiogram (ECG) recordings. ECG reflects the periodic bioelectrical activity generated over each cardiac cycle, primarily resulting from sequential excitation of the sinoatrial node, atria, and ventricles. ECG-based analyses are commonly performed in both the time and frequency domains. In the time domain, key indices include heart rate and heart rate variability (HRV). Heart rate is typically derived from the R–R interval, defined as the time between two consecutive R-wave peaks, which represents one complete cardiac excitation cycle. HRV characterizes the temporal variability between successive heartbeats and is computed from fluctuations in consecutive R–R intervals. A widely used time-domain metric is the standard deviation of normal-to-normal intervals (SDNN) over a defined period. Higher SDNN values indicate greater overall autonomic regulation and have been associated with better stress recovery and cardiovascular health.

Frequency-domain analysis provides complementary information on autonomic function. Using a Fourier transform, the time-series ECG signal is decomposed into its frequency components, and the power spectral density is calculated within predefined bands, typically including high frequency (HF), low frequency (LF), very low frequency (VLF), and ultra-low frequency (ULF). Among these, LF and HF power, as well as the LF/HF ratio, are often used as indicators of autonomic modulation and are sensitive to environmental and affective influences. HF power predominantly reflects parasympathetic (vagal) activity; higher HF values are generally interpreted as indicating a more relaxed physiological state. (Sinha et al., 1992); (Murakami & Ohira, 2007). The LF/HF ratio is frequently used to describe sympathovagal balance: higher values are commonly interpreted as relatively greater sympathetic dominance and/or reduced vagal influence on sinoatrial node regulation. An LF/HF ratio close to 1 suggests comparable contributions of sympathetic and parasympathetic activity, whereas values >1 are often interpreted as reflecting heightened arousal and values <1 as reflecting parasympathetic predominance.

Figure 1 shows the detailed experimental process. The experiment comprised four conditions and followed a within-subject design, such that each participant completed all four acoustic conditions. Each condition consisted of two phases: a stimulation phase and a recovery phase. Upon arrival, participants were fitted with HRV monitoring equipment and rested quietly for 5 min to acclimatize to the laboratory environment. During the stimulation phase, participants completed the Maastricht Acute Stress Test (MAST) twice. The MAST is a well-established standardized paradigm for stress induction that is simple and cost-effective to administer and elicits physiological and psychological responses comparable to real-world stress reactions (Quaedflieg et al., 2015). It comprises two consecutive tasks: a cold pressor test (physiological stressor) and a mental arithmetic task (psychological stressor). In the cold pressor test, participants immersed both palms fully in ice-cold water for 30 s to induce pain-related physiological stress, followed immediately by a 90-s mental arithmetic task requiring serial subtraction from four-digit numbers, with responses reported verbally. During the recovery phase, participants were exposed to one of four auditory stimuli while HRV was recorded continuously. To minimize order effects, the four conditions were presented in a counterbalanced randomized sequence. A 10-min washout period was implemented between conditions, during which participants remained in a neutral, silent environment to reduce potential carry-over effects.

ECG signals were acquired using the HeaLing-R211B system at a sampling rate of 1000 Hz. A single-lead configuration was applied, with electrodes placed at the right infraclavicular region and the left lower thorax. Participants remained seated throughout the recording. Before analysis, HRV data were preprocessed. R–R intervals were screened for artifacts and ectopic beats using a threshold criterion (R–R intervals outside 300–2000 ms). Identified artifacts were corrected using the automatic correction algorithm implemented in Kubios. (Lipponen & Tarvainen, 2019).

All statistical analyses were performed using IBM SPSS Statistics (Version 28.0). Subjective outcomes (Perceived Restorativeness Scale, PRS) and physiological indices (SDNN, HF power, LF/HF ratio, and stress index) were analyzed to evaluate differences in stress recovery across the four auditory conditions. A one-way repeated-measures analysis of variance (ANOVA) was used as the primary inferential approach. Figures were generated using OriginPro 2021 and further refined for publication using Adobe Photoshop 2024.

The subjective evaluation of the restorative potential of different sound environments was conducted using PRS, which encompassed four key dimensions: being-away, fascination, compatibility, and an overall PRS score. The results are visually presented in Figure 2, Figure 3, Figure 4, & Figure 5. Overall, the pattern of results was clear and consistent: compared with silence and wind sound, the two natural sound conditions (water sound and birdsong) received significantly higher ratings across all PRS dimensions. For the being-away dimension (Figure 2), water sound (mean = 1.91) and birdsong (mean = 2.14) were both significantly higher than silence (mean = 0.07; p < 0.01). Birdsong was also significantly higher than wind sound (p < 0.01), whereas wind sound remained close to baseline and below the two natural sound conditions. The distributional patterns further indicate higher medians for water sound and birdsong, with relatively narrower interquartile ranges, suggesting greater agreement among participants regarding their perceived sense of “being away.”

A similar pattern was observed for fascination (Figure 3). Water sound and birdsong again yielded relatively high ratings (means = 1.45 and 1.66, respectively), exceeding the silent control condition (mean = −0.28). In contrast, wind sound ratings clustered around zero to slightly negative values, indicating a comparatively limited capacity to capture attention relative to water sound and birdsong.

For compatibility (Figure 4), water sound and birdsong maintained high scores (means = 1.86 and 2.20, respectively). Wind sound produced the lowest compatibility rating (mean = −0.13), which was lower than the silent condition (mean = 0.47), suggesting that participants perceived the windy sound environment as less compatible with their preferred recovery experience.

When these dimensions were aggregated, the overall PRS under underwater sound and birdsong was significantly higher than under silence (p < 0.01; Figure 5). Specifically, the overall PRS under flowing water and birdsong was approximately 2.11 and 2.27 times that of the silent control condition, respectively, whereas the wind sound condition remained near baseline. Taken together, these subjective findings indicate that water sound and birdsong were perceived as the most restorative auditory environments, substantially outperforming both silence and wind sound. This provides a psychological basis for examining whether these perceived restorative benefits are accompanied by enhanced physiological stress recovery, as assessed via HRV in the following section.

Figure 6 presents the evaluation results of SDNN under four working conditions. The analysis of SDNN data revealed obvious physiological trends corresponding to subjective restorative perception. A repeated-measures ANOVA revealed a significant main effect of acoustic condition on SDNN, F (3, 90) = 4.62, p = 0.005, ηp² = 0.13, indicating a moderate effect size. As shown in the figure, the average SDNN value of water sound was the highest (34.9 ms±9.8), indicating that among all conditions, the enhancement of overall autonomous regulation and stress recovery was the most significant. Secondly, Birdsong (33.3 ms±9.3) and Wind Sound (32.2 ms±10.8). The average SDNN value under the silent condition was the lowest (31.6 ms±12.5), indicating a relatively low level of physiological recovery. In conclusion, the SDNN results indicated that exposure to natural sounds, especially water sounds, leads to a significant increase in SDNN compared with silence. This provided objective physiological evidence that natural sounds, especially water sounds, could enhance the function of the autonomic nervous system, promote stress recovery, and effectively confirm the subjective perceptual restorative discovery.

The frequency domain analysis of HRV provided deeper insights into the autonomic nervous system (ANS) dynamics during exposure to different sound environments. HF power, a marker of parasympathetic (vagal) activity and respiratory sinus arrhythmia, and the LF/HF ratio, an indicator of the sympathovagal balance, were analyzed in Figure 7 and Figure 8. As shown in Figure 7, repeated-measures ANOVA indicated a significant effect of acoustic condition on HF power, F (2.41, 72.30) = 8.57, p < 0.001, ηp² = 0.22. The high-frequency power data analysis indicates that the average high-frequency power caused by water sound was up to 57.2 ms², suggesting that the activity of the parasympathetic nervous system is significantly increased compared to other conditions. Although the average values of the still sound, birdsong, and wind sound were similar, being 52.0 ms², 52.8 ms², and 52.2 ms², respectively.

However, the increase in the average value of water sounds indicates a stronger vagus nerve-mediated relaxation response. The most striking result comes from the analysis of the LF/HF ratio (Figure 8). The average LF/HF ratio under water sound was the lowest (0.39±0.15), followed by birdsong (0.59±0.48), indicating a shift from autonomic nerve balance to parasympathetic nerve dominance. The LF/HF ratios of the silence group (0.71±0.52) and the wind sound group (0.82±0.58) were significantly higher than those of the water sound, reflecting a relatively higher sympathetic tone or a less relaxed state. In conclusion, the frequency-domain analysis of HRV provides compelling physiological evidence that the sounds of water and birds uniquely promote stress recovery by significantly enhancing parasympathetic nerve activity and shifting autonomic nerve balance towards a resting and digestive state. This effect is obvious and much stronger than that of silence and the sound of the wind. These results, in synergy with the subjective PRS scores and SDNN results, construct a coherent narrative. That is, the sound of water is the most effective auditory stimulus for triggering psychological and physiological recovery.

The stress index, a derived HRV parameter that reflects the level of sympathetic activation and overall stress load on the autonomic nervous system, was analyzed across the four experimental conditions. Lower values indicate reduced stress and a more relaxed physiological state. Figure 9 shows the results of the pressure index under the four working conditions. As shown in Figure 9, repeated-measures ANOVA demonstrated a significant effect of acoustic condition on stress index, F (3, 90) = 5.73, p = 0.001, ηp² = 0.16. The average stress index under the birdsong condition was the lowest (10.54±2.39), which was 14.48% lower than that under the silent condition. It indicates that among the four conditions, the reduction in sympathetic nervous system activity and overall physiological stress is the greatest. Secondly, there are wind sounds and water sounds, which are 10.54 and 11.09, respectively, 11.44% and 9.9% lower than the silent conditions. In conclusion, the results of the stress index provide convergent physiological evidence that exposure to natural sounds effectively reduces sympathetic nerve activation and promotes a less stressful physiological state compared to silence.

A part of this paper was presented at a conference named Urban Planning & Architectural Design for Sustainable Development (UPADSD).

This work was supported by JST SPRING, Japan Grant Number  JPMJSP2149.

The study was conducted in accordance with the Declaration of Helsinki and approved by the University Ethics Committee (IRB number: QUT-HEC-2024023).

The author(s) declare(s) that there is no competing interest.