Colors of noise: spectral slopes, perception and research evidence
In acoustics, 'colors' of noise describe how random sound energy is distributed across frequency. Distinguishing power per hertz from power per octave explains why white, pink, brown, and grey noise sound so different.
1. How the sound works: Power spectral density vs. power per octave
In acoustics and signal processing, broadband random noise is classified by its power spectral density (PSD) S(f), which specifies how signal power is distributed per 1 Hz of bandwidth across frequency f (Voss & Clarke, 1978; McCartney, 1999). Idealized power-law noise families follow:
S(f) ∝ 1 / fα
When reading charts or audio specifications, it is essential to distinguish between two valid ways of expressing spectral slope:
- Power spectral density per 1 Hz bandwidth (S(f)): Measures the power inside a fixed 1 Hz-wide slice at frequency f. Doubling the frequency by one octave changes S(f) by 10·log10(2−α) ≈ −3.01α dB/octave (or −10α dB/decade).
- Integrated power per octave band (or 1/3-octave band): Because each octave spans twice as many hertz as the octave below it (for example, 1,000 to 2,000 Hz is 1,000 Hz wide, whereas 100 to 200 Hz is only 100 Hz wide), integrating S(f) across proportional octave bands adds +10·log10(2) ≈ +3.01 dB/octave to the slope:
Slope per octave band = Slope of S(f) per Hz + 3.01 dB/octave
| Noise family | PSD model S(f) (per 1 Hz) | PSD slope per 1 Hz | Power slope per octave band | Typical character & definition notes |
|---|---|---|---|---|
| White noise | S(f) ∝ constant (α = 0) | 0 dB/oct (0 dB/dec) | +3.01 dB/oct (+10 dB/dec) | Equal power per 1 Hz; sounds bright and hissy because higher octaves contain progressively more total bandwidth energy. |
| Pink (1/f) noise | S(f) ∝ 1/f (α = 1) | -3.01 dB/oct (-10 dB/dec) | 0 dB/oct (equal power per octave) | Equal energy in every octave band; sounds more balanced across bass, midrange, and treble than white noise (Voss & Clarke, 1978). |
| Brown (red / Brownian) noise | S(f) ∝ 1/f2 (α = 2, over a bounded passband) | -6.02 dB/oct (-20 dB/dec) | -3.01 dB/oct (-10 dB/dec) | Named after Brownian motion (random-walk integration of white noise); strongly bass-heavy with a deep, low-frequency rumble. Practical generators apply a low-frequency high-pass cutoff to prevent DC drift. |
| Grey noise | Shaped to an inverse equal-loudness contour | Non-linear (boosted bass & treble relative to 1 to 4 kHz) | Depends on target contour & listening level | Designed to approximate equal perceived loudness across frequency bands. Because standard pure-tone equal-loudness contours (ISO 226:2023) apply to pure tones under specified conditions and change shape with level (phons), there is no single universal grey-noise spectrum. |
| "Green noise" (informal label) | No standardized definition in acoustics | Varies by creator (typically mid-band filtered noise) | Varies by implementation | In spatial signal processing, Lau et al. (1999) defined "green-noise masks" as mid-spatial-frequency bandpass halftoning patterns. In consumer audio, "green noise" is an informal, non-standardized marketing label for midrange-filtered ambient noise with no canonical cutoff frequencies in acoustic standards. |
2. How it is perceived: Why white noise sounds bright and masking principles
Listeners are often surprised that mathematically "flat" white noise sounds like a sharp, high-pitched hiss rather than an even broadband wash. Three factors contribute to this perception (Moore, 2012; ISO 226:2023):
- Widening auditory filters and octave bandwidth scaling: Cochlear auditory-filter bandwidth (equivalent rectangular bandwidth, ERBN) increases with center frequency (Moore, 2012). Viewed across proportional octave bands, the octave from 4 to 8 kHz spans 4,000 Hz while the octave from 125 to 250 Hz spans only 125 Hz, so flat-PSD white noise delivers 32 times (+15 dB) more integrated power into the 4 to 8 kHz octave than into the 125 to 250 Hz octave.
- Human mid-to-high-frequency sensitivity (2 to 5 kHz): Outer- and middle-ear acoustics (including ear-canal resonance) make human hearing most sensitive in the 2 to 5 kHz region (Moore, 2012), a pattern also reflected in pure-tone equal-loudness contours (ISO 226:2023).
- Transducer and acoustic coupling: Perceived low- and high-frequency balance also depends on the frequency response, fit, signal processing, and room coupling of the headphones or loudspeaker used for playback.
In everyday environments, steady broadband noise works primarily through energetic auditory masking: raising the steady background floor so that sudden transient sounds (such as a door closing or distant traffic) have a smaller peak-to-background ratio and are less likely to capture attention or trigger an arousal response.
3. What research shows: Sleep, attention, physiology, and clinical settings
Peer-reviewed research on broadband noise spans several distinct domains with markedly different findings across populations and study designs:
A. Sleep and overnight sleep architecture
- Qualitative systematic review of continuous white noise (Riedy et al., 2021): Reviewed 38 studies of continuous noise as a sleep aid and rated overall evidence quality as very low under GRADE, noting mixed findings on sleep latency and fragmentation and cautioning that continuous overnight sound adds to cumulative noise exposure if played at elevated levels.
- Meta-analysis of white-noise RCTs across age and ICU settings (Ding et al., 2025): Synthesized 12 randomized controlled trials (N = 1,301: 356 infants/toddlers aged 0 to 3 years, 648 adults aged 18 to 64 years, and 297 older adults ≥ 65 years; including 544 ICU and 401 non-ICU adult patients) testing white noise only (not pink or brown noise). In the indexed abstract, white noise improved infant 24-hour total sleep time, 12-hour sleep efficiency, and 12-hour nocturnal awakenings, but had no statistically significant effect on 12-hour nocturnal total sleep time (p = 0.16), 24-hour sleep efficiency (p = 0.46), or 24-hour wake after sleep onset (p = 0.46). In adults, older adults, ICU, and non-ICU subgroups, white noise significantly lowered Pittsburgh Sleep Quality Index (PSQI) scores alongside substantial clinical and age-group heterogeneity (see What Research Shows and References for subgroup effect estimates). Because clinical masking trials mostly take place in noisy hospital wards, results cannot be automatically generalized to quiet home bedrooms.
- Narrative systematic review in hospitalized adults (Khandelwal et al., 2025): Reviewed 7 randomized trials (N = 496 hospitalized adults: 6 testing white noise at 40 to 60 dB and 1 testing USASI masking noise). Because substantial heterogeneity in interventions and outcome measures prevented a quantitative meta-analysis, the authors performed a narrative synthesis, reporting improvements in sleep efficiency and study-level subjective sleep questionnaires (such as PSQI, RCSQ, and VSH) in noisy hospital wards while noting 8 participants across reviewed trials who discontinued because of discomfort or found the intervention less acceptable.
- Laboratory polysomnography trial of continuous pink noise vs. environmental noise (Basner et al., 2026): Conducted a 7-night laboratory crossover polysomnography (PSG) study (1 noise-free adaptation night followed by 6 experimental condition nights) in n = 25 healthy adults using calibrated bedroom loudspeakers across quiet control, recorded environmental traffic/aircraft noise (EN), continuous pink noise at 50 dBA (PN50), EN+PN40, EN+PN50, and EN+earplugs (the manuscript Methods state that body-text p-values are adjusted for multiple testing using Benjamini–Hochberg FDR unless otherwise specified). Compared with the quiet control night, continuous pink noise alone at 50 dBA (PN50) significantly reduced REM sleep duration by 18.6 minutes (SE 4.7 min in Table 3, p = 0.0003). Under the tested environmental-noise contrast (EN+earplugs vs. EN), earplugs mitigated environmental-noise N3 disruption without the REM reduction observed under PN50; morning subjective ratings were also worse after EN and EN+PN nights relative to quiet control (partially mitigated by earplugs), whereas no significant between-condition differences were detected on next-morning cognitive, cardiovascular, or hearing-threshold tests (all p > 0.65; single-night laboratory tests do not establish long-term safety).
B. Attention and cognitive task performance
- Meta-analysis in ADHD vs. non-ADHD populations (Nigg et al., 2024): Analyzed 13 studies (N = 335) of white and pink noise during cognitive tasks in children and young adults with ADHD or elevated attentional symptoms, finding a small, statistically significant benefit (g = 0.249, 95% CI 0.135 to 0.363, p < 0.0001, graded by the authors as moderate certainty under GRADE). Conversely, across 11 studies (N = 335) in non-ADHD comparison participants, white/pink noise produced a small negative effect on task performance (reported as g = −0.212, 95% CI −0.355 to −0.069, p = 0.0036 in the abstract and g = −0.206, 95% CI −0.330 to −0.081, p = 0.001 in the main text results, without a separate GRADE certainty label stated for the non-ADHD contrast). Notably, the authors found zero qualifying experimental studies of brown noise within their literature search window.
- Systematic review of auditory stimulation on arousal and cognition (Chee et al., 2024): Reviewed 31 studies across music, ambient/white noise, and binaural beats examining autonomic arousal (heart-rate variability, electrodermal activity, and pupillometry) and cognitive/attentional task performance, reporting heterogeneous, context-dependent findings across stimulus categories and listener populations.
- Simulated car-cabin attention study (De Francesco et al., 2025): Tested N = 26 healthy adults behaviorally (with n = 21 analyzed for heart-rate recordings) on the Attention Network Test (ANT) inside a stationary car cabin across four acoustic conditions: slow-beat music (≈ 50 BPM), fast-beat music (≈ 155 BPM), brown noise, and silence. Slow-beat music significantly improved attentional orienting and executive control, lowered heart rate, and increased subjective pleasantness, relaxation, and perceived concentration relative to silence, whereas brown noise did not outperform silence or music on objective attentional network measures.
C. Physiological arousal measures
- Pupillometry comparison of white, pink, and brown noise (Erfanian et al., 2025): Enrolled 38 healthy adults (31 analyzed after exclusions) exposed to 10-second presentations of white, pink, and brown noise during continuous pupil-size recording (an index of locus coeruleus–noradrenergic arousal). The study detected no significant differences in pupil-size modulation between white, pink, and brown noise colors.
D. Pediatric dental procedural anxiety and pain
- Pediatric tooth extraction randomized trial (Kolhe et al., 2024): Randomized 40 children aged 8 to 12 years undergoing primary tooth extraction into four parallel groups (n = 10 per group: white noise, pink noise, brown noise, and cartoon music control) to evaluate acute dental anxiety and pain perception. That small pediatric dental trial measured procedural anxiety and pain during tooth extraction (not sleep or sustained attention) and cannot be extrapolated to general sleep or cognitive claims.