Soundscapes: recordings, synthesis and what research studies
Ambient soundscapes (such as rain, surf, wind, and streams) can be captured with field microphones, assembled from loops, or synthesized procedurally from filtered noise. Understanding how each method works helps clarify what published studies actually test.
1. How the sound works: Recordings, loops, and procedural synthesis
Ambient nature soundscapes, such as rain, ocean surf, wind through trees, or flowing water, can be produced through four technical approaches, each with distinct acoustic and practical trade-offs (Farnell, 2010):
| Approach | How the audio is produced | Strengths | Typical limitations |
|---|---|---|---|
| Long field recordings | Microphones capture a real acoustic environment over tens of minutes or hours. | Authentic acoustic complexity, natural spatial cues, and organic environmental variation. | Large file storage or streaming bandwidth; may include unintended background transients (distant aircraft, birds, or wind buffeting). |
| Short looped or layered recordings | A recorded clip (10 to 120 s) is repeated continuously with a crossfade at the splice point, or several loops of different lengths are layered. | Real recorded timbre with modest file size. | During quiet, attentive listening, the human auditory system can learn to recognize a distinctive droplet, wave crest, or crossfade seam repeating at fixed intervals. |
| Procedural synthesis | Real-time mathematical noise generators, time-varying bandpass/resonant filters, and stochastic impulse models synthesize the texture continuously (Farnell, 2010). | No fixed audio-file splice seam, negligible storage footprint, and independent control over parameters (such as wave surge rate or droplet density). | Simpler physical models than a real forest or coastline; if driven by a simple sinusoidal LFO, the macro-envelope still repeats its periodic cycle even when the underlying pseudorandom noise generator uses a high-period state space (such as a 64-bit generator) whose mathematical period exceeds a human listening session (finite-state PRNGs remain deterministically periodic over their state space rather than literally infinite). |
| Hybrid engines | Combines procedural noise/filter beds with randomized short recorded one-shots or impulse responses. | Blends organic recorded transients with continuously varying procedural background beds. | Requires careful level matching and memory management. |
2. How it is perceived: Auditory scene analysis and masking vs. predictability
In psychoacoustics, listening to an environmental soundscape engages auditory scene analysis (Moore, 2012). Unlike stationary white or pink noise, whose statistical power spectral density and long-term RMS level remain time-invariant while instantaneous sample values fluctuate randomly, natural water and wind sounds are non-stationary acoustic processes whose short-time amplitude envelope and spectral centroid drift over seconds:
- Gentle spectral motion vs. stationary hiss: Slow amplitude swells and broadband water or wind textures are often chosen by listeners who find unmodulated high-frequency noise fatiguing, while still providing broadband masking whose effectiveness depends on how closely the soundscape spectrum overlaps the competing noise.
- Why abrupt transients or short loops can draw attention: The auditory system continuously monitors background sound for unexpected changes or repeating patterns. Sharp, loud transients (such as thunderclaps or close bird calls) or a noticeable loop splice can shift a soundscape from a receding background into an attention-capturing foreground event.
3. What research shows: Field recordings vs. decentralized sleep trials
Most published environmental psychology and clinical studies evaluate recorded nature sounds (birds, water, wind, or mixed natural environments) rather than purely synthesized procedural audio:
- Laboratory stress-recovery experiment (Alvarsson et al., 2010): Tested N = 40 university students exposed to nature sounds (50 dBA fountain and bird recording) vs. high-level road traffic noise (80 dBA), low-level traffic noise (50 dBA), or ambient noise (40 dBA) following a stressful mental arithmetic task. Mean skin conductance level (SCL, an index of sympathetic arousal) during recovery was significantly lower during nature sounds than during high traffic noise (p = 0.043), and polynomial curve fitting indicated 9% to 37% faster half-life SCL recovery during nature sound than during the noise conditions, whereas high-frequency heart rate variability (HF-HRV, an index of parasympathetic vagal activity) showed no detected difference between conditions.
- Syntheses and systematic reviews of natural sounds (Buxton et al., 2021; Zhu et al., 2024; Fan & Baharum, 2024): Buxton et al. (2021) identified 36 publications in its literature review and conducted meta-analyses on 18 of those publications, reporting decreased stress and annoyance (g = −0.60, 95% CI −0.97 to −0.23) and improved health and positive affective outcomes (g = 1.63, 95% CI 0.09 to 3.16) alongside soundscape monitoring across 68 U.S. national parks. Zhu et al. (2024) meta-analyzed 15 studies (N = 1,285) on natural sound exposure, reporting reductions in anxiety and physiological arousal measures (heart rate, systolic and diastolic blood pressure, and respiratory rate), substantial between-study heterogeneity, and inconsistent conclusions for cognitive performance restoration. In the indexed abstract of their systematic review and meta-analysis, Fan & Baharum (2024) reported that exposure to natural sounds compared with a quiet environment yielded statistically significant differences in heart rate (p = 0.006), blood pressure (p = 0.001), and respiratory rate (p = 0.032), but no statistically significant difference in mean arterial pressure (p = 0.407), perceived stress, or oxygen saturation SpO2 (p = 0.251). Importantly, these reviews evaluate recorded or live natural sounds rather than mathematical noise-band synthesis, so their findings cannot be transferred automatically to procedural generators.
- Four-week decentralized randomized trial of bedtime audio (Vazzaz et al., 2026, USleep study): Tested whether four weeks of app-based bedtime audio reduced self-reported sleep disturbance in working adults with poor sleep (N = 495 enrolled in the intention-to-treat sample after recruitment ended before the planned sample size because of budget constraints). Participants were randomized across Bedtime Stories, Sleep Sounds (music, white and pink noise, and nature sounds), Sleep Skills (voice-led behavioral and relaxation exercises), and an active digital control arm in the same app (Daily Boost). All four groups improved from baseline over four weeks, and none of the three audio intervention arms demonstrated statistically significant superiority over the active control on the primary week-4 PROMIS Sleep Disturbance outcome (Sleep Sounds vs. Daily Boost: g = 0.14, 95% CI −0.11 to 0.39, p = 0.281). Because the comparator was an active app routine rather than silence, this result shows non-detected superiority over an active control rather than zero change (comparison in What Research Shows; participant flow and arm details in References).