Real-Time Synchronized Measurement of Preferred Listening Levels: Effects of Ambient Noise and Music Genre in Everyday Environments

Abstract

Background:

Portable listening devices are widely used in daily activities, exposing users to music under highly variable ambient noise, and potentially increasing the risk of music-induced hearing loss. Previous studies show preferred listening levels (PLLs) increase in noise and vary by genre; however, these factors have typically been examined separately under laboratory conditions without synchronized real-world measurement.

Methods:

Fifty-one normal-hearing young adults used a custom Android application with standardized earphones. Participants listened to four music genres (ballad, dance, pop, and new age) across five everyday environments (on-campus, off-campus, library, gym, and cafeteria). The application recorded an average listening level as Z-weighted equivalent continuous sound pressure level (LZeq), peak listening level as C-weighted peak level, and simultaneous ambient noise as LZeq. Listening-level data were sampled every 1 second and transmitted to a secure server at 5-second intervals, allowing cross-validation with participant worksheets. Average and peak outcomes were analyzed using two-way repeated-measures analyses of variance (ANOVAs).

Results:

Ambient noise varied across environments, with the lowest level in the library (50.40 dB SPL) and the highest in the cafeteria and gym (73.03 and 72.05 dB SPL). Across all observations, PLLs ranged from 43.0 to 99.8 dB SPL, with the lowest mean in the library–new age condition (55.14 dB SPL) and the highest mean in the gym–dance condition (74.85 dB SPL), demonstrating substantial inter-individual variability. Repeated-measures ANOVAs showed significant main effects of environment and genre on average PLLs, with no environment × genre interaction, indicating that these effects were independent. For peak levels, both main effects were significant, and an environment × genre interaction emerged, suggesting context-dependent modulation of transient peaks.

Conclusion:

In everyday listening, average listening levels increased in noisier environments and differed across music genres, whereas peak levels reflected a context-dependent interaction between environment and genre. Real-time synchronized monitoring may help characterize everyday listening exposure and inform context-aware safe-listening strategies.

KEY MESSAGES

• (1)Preferred listening levels (PLLs) tended to increase with ambient noise across everyday listening environments.
• (2)Average listening levels varied across both listening environments and music genres, suggesting that both context and content contribute to listening behavior.
• (3)Real-time synchronized monitoring may offer a useful way to characterize context-dependent listening exposure in real-world settings.

INTRODUCTION

Portable listening devices (PLDs) have become an integral part of daily life, enabling music listening during commuting, studying, exercising, and other activities.[1,2] While this accessibility offers clear benefits, it also raises concerns that repeated exposure to elevated listening levels may contribute to noise-induced hearing loss (NIHL), particularly among young adults who frequently use earphones for extended periods.[3,4] Accordingly, recent public-health initiatives have emphasized the need for practical safe-listening guidance and device-supported exposure management for leisure listening, including concepts of a weekly “sound allowance.”[5]

A consistent finding in the PLD literature is that users increase their listening levels in the presence of background noise, a phenomenon commonly referred to as a noise-compensation effect.[6] This behavior creates a “double exposure” scenario in which environmental noise and recreational audio co-occur, potentially increasing cumulative auditory dose even when each source alone may appear acceptable.[1,7]

Laboratory studies have been essential for isolating the causal effects of background noise, earphone characteristics, and music-related factors on preferred listening levels (PLLs). However, these studies often rely on fixed noise recordings and short experimental blocks that may not reflect the dynamic nature of real-world sound environments.[8,9] In addition to ambient noise, music content itself can influence self-selected listening levels. Previous research suggests that PLLs vary across music genres, likely reflecting differences in acoustic and perceptual characteristics such as spectral distribution, rhythmic salience, and vocal prominence, as well as differences in listener engagement and preference.[10,11,12,13]

Despite these advances, environmental noise and music-related factors have often been examined separately or under tightly controlled conditions, leaving uncertainty about how they interact during real-world listening. A key methodological limitation is the lack of synchronized measurement of (i) user-selected earphone output and (ii) the ambient noise that motivates volume adjustments.[14] When these variables are measured separately—or when ambient noise is represented only by a nominal “location” label—important within-location fluctuations and moment-to-moment behavioral adjustments may be overlooked.

Real-time synchronized monitoring can address this gap by capturing listening behavior as it unfolds and directly pairing the user’s selected level with contemporaneous acoustic context. Smartphones provide a practical platform for such monitoring because they are ubiquitous and can support continuous sampling, time stamping, and secure data transfer during typical daily activities.[15,16] With standardized earphones and an explicit calibration strategy, a smartphone-based system can enable scalable field measurement of both average listening level and peak exposure while simultaneously recording ambient noise.[17,18,19] This approach may improve ecological validity, characterize inter-individual variability in naturalistic settings, and inform context-aware safe-listening interventions. Therefore, field studies are needed to measure listening level and ambient noise synchronously in everyday environments while also accounting for music content.

Accordingly, the present study employed a real-time synchronized measurement approach to quantify PLLs across everyday environments and music genres. Specifically, we examined the following: (1) whether ambient noise differs across common listening environments, (2) how listening environment and music genre influence average preferred listening level (Z-weighted equivalent continuous sound pressure level [LZeq], and (3) whether peak listening levels (C-weighted peak level [LCpeak] are modulated by the combined influence of environment and genre. We hypothesized that ambient noise and average PLLs would vary by environment, that average PLLs would also differ by genre, that environment and genre would contribute additively to LZeq (i.e., without a significant interaction), and that LCpeak would show context-dependent modulation reflected in an environment × genre interaction.

MATERIALS AND METHODS

Study Design and Setting

This study used a within-subject, repeated-measures field design to examine PLLs on PLDs across everyday environments and music genres. Each participant listened to four music genres (ballad, dance, pop, and new age) in five everyday environments (on-campus walkway, off-campus walkway, library, gym, and campus cafeteria). Measurements were conducted under naturalistic conditions without experimental manipulation of environmental sound, allowing ambient-noise fluctuations to occur as encountered during typical daily activities.

To improve operational clarity of listening environments, the on-campus and off-campus conditions were defined as outdoor walking areas with pedestrian traffic, with the off-campus walkway located approximately 200–300 m from campus. The library condition was defined as an indoor quiet study space, the gym condition as an indoor exercise facility, and the cafeteria condition as an indoor dining area. Data collection was scheduled between 06:00 and 24:00 to minimize extreme nighttime or early-morning noise profiles while preserving real-world variability.

Participants

Fifty-one young adults (26 males and 25 females; age range, 20–29 years; mean age = 22 years; standard deviation [SD] = 1.64) participated in the study. All participants were Android smartphone users and had normal hearing sensitivity (air-conduction thresholds ≤20 dB HL at octave frequencies from 0.25 to 8 kHz), normal middle-ear status (bilateral Type A tympanograms), and no self-reported history of otologic disease. Written informed consent was obtained from all participants, and the study protocol was approved by the Institutional Review Board of Hallym University (HIRB-2016-060).

Smartphone Application, Earphones, and Acoustic Measures

A custom Android application, developed in our previous work, was used to quantify listening behavior during music playback through earphones while simultaneously capturing ambient noise for contextual pairing. The application recorded (i) average earphone output level as a LZeq and (ii) maximum earphone output level as a LCpeak, while ambient noise was quantified as LZeq using the smartphone microphone input. Microphone input values were acquired using the Android MediaRecorder class, and measurements were initiated and terminated using a manual start/stop control in the user interface.

Listening level was defined using the Android system media volume index, which consists of 16 discrete steps (0–15). Earphone-output sound pressure levels corresponding to each volume step were determined through laboratory calibration using a sound level meter (Type 2250, Brüel & Kjær, Nærum, Denmark) and an artificial ear simulator (Type 4153, Brüel & Kjær) with a 2-cc coupler. For each volume step, both LZeq and LCpeak were measured at the earphone output and mapped to the application-reported values, enabling conversion from the digital volume index to ear-level sound pressure level (SPL).

All measurements were conducted using a standardized earphone model (Xiaomi PISTON 3, Xiaomi, China) to reduce transducer-related variability. Application-derived LZeq and LCpeak values were validated against laboratory reference measurements across all volume steps, with agreement within ±5 dB. Ambient noise levels recorded via the smartphone microphone were also validated against a calibrated free-field microphone (Type 4189, Brüel & Kjær), showing discrepancies within ±5 dB across multiple environments; detailed calibration procedures have been described previously.[16]

To facilitate interpretation relative to hearing conservation guidelines expressed in A-weighted metrics, representative calibration comparisons were conducted across music genres and output steps. Rather than applying a single universal conversion factor, the relationship between LZeq and LAeq was interpreted cautiously across the output range. Discrepancies were largest at the lowest output steps, whereas relatively more stable relationships were observed at moderate output levels; however, this relationship was not constant across the full range.

Because music playback was delivered via participants’ personal Android smartphones, device and operating system heterogeneity could influence raw digital output behavior. To reduce cross-device variability in absolute output estimation, the present study restricted participation to Android smartphone users, standardized the earphone transducer across participants, and relied on volume-step–to–SPL mapping obtained through laboratory calibration of the standardized earphones.

Ambient-Noise Measurement Validation

To evaluate the accuracy of smartphone-based ambient-noise measurements, app-derived microphone LZeq values were compared with those obtained using a reference free-field ½-inch microphone (Type 4189, Brüel & Kjær). Across comparisons, discrepancies between the reference and application outputs were within 3 dB SPL, supporting acceptable agreement for field measurement of environmental noise in this study.

Music Stimuli and Listening Environments

Four music genres (ballad, dance, pop, and new age) were included, and two songs were selected for each genre based on the most downloaded tracks on www.melon.com at the time of selection, resulting in a total of eight songs [Table 1]. Songs were presented in randomized order across participants to minimize order effects. Full-length tracks were used rather than fixed-duration excerpts.

Table 1.

Music stimuli selected for the study

Genre	Artist	Title
Ballad	Han Dong-geun	Rewrite the End of This Novel
	Lim Chang-jung	The Love I Committed
Dance	Yoo Jae-suk, EXO	Dancing King
	TWICE	CHEER UP
Pop	Sam Smith	I’m Not the Only One
	Justin Bieber	Love Yourself
New Age	Yiruma	Kiss the Rain
	Yiruma	River Flows in You

Note: Two popular songs per genre (ballad, dance, pop, and new age) were selected from Melon’s most-downloaded tracks at the time of selection, presented in randomized order across participants to minimize order effects.

To characterize acoustic differences among the selected music genres, long-term average spectra of representative tracks were also examined using the same smartphone–earphone–2cc coupler system [Figure 1]. The spectra showed genre-related differences in frequency distribution, with relatively greater mid-to-high frequency energy in dance and pop tracks than in new age material.

Figure 1.

Long-term average spectra of music stimuli by genre. Sound pressure level (dB SPL) as a function of frequency (Hz) for representative tracks from each of the four music genres (ballad, dance, pop, and new age) used in the study. Spectra were computed using long-term averaging across the full duration of each track with 1/3-octave smoothing. Dance and pop genres exhibit higher energy in the mid-to-high frequency range (1–8 kHz), whereas new age shows relatively lower levels across the spectrum, potentially contributing to genre-related differences in preferred listening levels.

Audio files were distributed in their original MP3 format and were not subjected to additional loudness normalization prior to playback. This approach was intended to preserve ecologically valid listening conditions, as commercially available tracks and genres often differ in overall level and dynamic range, and real-world listeners typically encounter such variability during everyday music use. Music was played through participants’ personal Android devices using standard media playback functions. Equalization and sound-enhancement settings were not experimentally controlled and remained in their default user-configured state.

Procedures and Data Handling

In each environment, participants listened to the assigned music and adjusted playback to their PLL using smartphone controls while wearing the standardized earphones. Listening-level and ambient-noise data were sampled every 1 second and transmitted to a secure server at 5-second intervals, enabling cross-checking with participant worksheets. During data collection, the application transmitted each participant’s data to a secure server and linked the records using a participant-specific identifier for data management. For analysis, the dataset was de-identified prior to aggregation and statistical processing. Ambient noise measurements were obtained in each target environment using the app’s manual start/stop function and were recorded on a predefined data sheet for each location. During ambient-noise measurements, participants were instructed to keep the smartphone either in hand or placed on a nearby surface, with the built-in microphone unobstructed, to minimize obvious obstruction during recording.

For statistical analyses, the primary analytic unit was the participant. Specifically, second-by-second data were aggregated within each participant to yield condition-level summaries (environment × genre) for (i) mean earphone output (LZeq) and (ii) peak earphone output (LCpeak) during playback, thereby preserving the repeated-measures structure and avoiding inflation of degrees of freedom due to high-frequency sampling.

To ensure data integrity, predefined data-quality criteria were applied before analysis. Segments with incomplete playback (<80% of total track duration) and periods during which the application was paused, or inactive, were excluded. LCpeak values were further screened for implausibility based on temporal discontinuity and deviation from the laboratory-calibrated measurement range. Values exceeding the calibrated maximum (120.1 dB SPL), taking observed measurement variability (±5 dB) into account, were classified as artifacts and excluded (LCpeak > 125 dB SPL).

Statistical Analysis

Normality was assessed using Shapiro–Wilk tests, and sphericity was evaluated using Mauchly’s test; Greenhouse–Geisser corrections were applied when the assumption of sphericity was violated. Preferred listening levels (LZeq) and peak listening levels (LCpeak) were analyzed using two-way repeated-measures ANOVAs with listening environment (library, on-campus, off-campus, cafeteria, and gym) and music genre (ballad, dance, pop, and new age) entered as within-subject factors. When significant main effects or interactions were observed, Bonferroni-adjusted pairwise comparisons were conducted using an alpha level of 0.05.

Exploratory analyses were also performed to examine potential sources of inter-individual variability in PLL, including Pearson correlation analyses, an independent-samples t-test, and a multiple linear regression model based on participant-level mean PLL values. In addition, to evaluate the direct association between continuously measured ambient noise and PLL, both a simple linear regression and a linear mixed-effects model with participant as a random intercept were fitted. Effect sizes are reported as generalized eta squared (η²_G) for ANOVA results and Cohen’s d for pairwise comparisons. All analyses were performed in R (version 4.5.2; R Core Team, Vienna, Austria).

RESULTS

Assumption Checks

Prior to inferential analyses, distributional assumptions were evaluated. Shapiro–Wilk tests indicated a statistically significant deviation from normality (W = 0.991, P < 0.001), although visual inspection of Q–Q plots suggested that the deviation was minor. Mauchly’s test indicated violations of sphericity for both location and genre; therefore, Greenhouse–Geisser corrections were applied, and all reported F-statistics reflect adjusted degrees of freedom.

Ambient Noise by Environment (LZeq)

Ambient noise levels (LZeq) differed across the five everyday listening environments [Table 2]. The library had the lowest mean ambient noise (50.40 ± 12.38 dB SPL), whereas higher mean levels were observed in the cafeteria (73.03 ± 7.00 dB SPL), gym (72.05 ± 9.38 dB SPL), and off-campus walkway (70.50 ± 11.57 dB SPL), with on-campus walkways showing an intermediate level (66.99 ± 10.10 dB SPL). Wide within-location ranges were observed across environments, indicating substantial real-world variability in background noise even within the same setting.

Table 2.

Ambient noise levels across listening environments

Listening environment	Ambient noise level(Mean ± SD, dB SPL)	Range(Min–Max, dB SPL)
Library	50.40 ± 12.38	18.86–99.64
On-campus	66.99 ± 10.10	35.73–87.83
Off-campus	70.50 ± 11.57	31.51–87.82
Cafeteria	73.03 ± 7.00	48.37–86.67
Gym	72.05 ± 9.38	30.26–96.90

Note: Abbreviations: LZeq, Z-weighted equivalent continuous sound pressure level; SD, standard deviation; SPL, sound pressure level. Z-weighted equivalent continuous sound pressure level (LZeq) of environmental background noise measured via smartphone microphone in each of the five everyday listening locations. Values are presented as mean ± standard deviation (SD) and observed range (minimum–maximum) in dB SPL.

Preferred Listening Levels (LZeq)

Descriptive preferred listening levels (LZeq) by environment and genre are presented in Table 3, and estimated marginal means are shown in Figure 2. Across all observations, preferred listening levels ranged from 43.0 to 99.8 dB SPL (mean = 68.09, SD = 9.69), indicating substantial inter-individual variability. The lowest condition mean was observed in the library–new age condition (55.14 dB SPL), whereas the highest was observed in the gym–dance condition (74.85 dB SPL), yielding a 19.71 dB difference across condition means.

Table 3.

Preferred listening levels (LZeq) and peak listening levels (LCpeak) across listening environments and music genres

Conditions	LZeq (dB SPL)				LCpeak (dB SPL)
	Ballad	Dance	Pop	New age	Ballad	Dance	Pop	New age
On-campus	70.22 (7.03)	71.81 (7.02)	69.19 (8.93)	59.61 (9.65)	92.45 (6.91)	93.07 (7.25)	92.69 (7.82)	85.28 (15.02)
Off-campus	72.19 (6.90)	73.34 (7.27)	71.59 (7.75)	61.27 (10.80)	94.02 (7.71)	94.20 (7.58)	93.34 (9.13)	88.58 (14.51)
Cafeteria	71.85 (7.33)	73.28 (7.66)	70.56 (7.49)	61.21(9.94)	94.56 (7.08)	93.59 (6.42)	93.04 (6.41)	87.38 (14.21)
Library	65.48 (7.78)	66.88 (7.75)	64.68 (7.33)	55.14 (6.63)	89.99 (6.72)	90.85 (7.80)	88.81 (8.70)	74.83 (13.53)
Gym	74.65 (7.01)	74.85 (6.47)	72.54 (6.70)	61.45 (10.11)	93.42 (8.85)	93.50 (7.95)	93.64 (8.60)	88.10 (13.68)

Note: Values are presented as mean (SD) for each environment × genre condition. LCpeak, C-weighted peak sound pressure level; LZeq, Z-weighted equivalent continuous sound pressure level; SD, standard deviation; SPL, sound pressure level.

Figure 2.

Preferred listening levels (LZeq) as a function of listening environment and music genre. Data points represent estimated marginal means with 95% confidence intervals computed from within-subject standard errors. Error bars reflect within-subject variability across repeated measures. Lower listening levels were consistently observed in the library and for new age music, whereas higher levels were observed in the gym and for dance music.

To examine potential sources of inter-individual variability in PLL, exploratory analyses were conducted using participant-level mean PLL values. Pearson correlation analyses showed that PLL was not significantly associated with age (r = −0.083, P = 0.563) or pure-tone average (PTA) hearing threshold (r = 0.131, P = 0.358). An independent-samples t-test also showed no significant gender difference in PLL between male participants (mean = 68.43 dB SPL, SD = 5.70) and female participants (mean = 67.78 dB SPL, SD = 6.14) (t (49) = 0.392, P = 0.697). In addition, a multiple regression model, including age, PTA, and gender accounted for only 2.4% of the variance in PLL and was not statistically significant (R² = 0.024, F(3, 47) = 0.389, P = 0.761).

The relationship between continuously measured ambient noise levels and PLL values was also examined. An initial simple linear regression showed a significant positive association between ambient noise and PLL (β = 0.160, 95% CI [0.116, 0.204], R² = 0.048, P < 0.001). To account for repeated measurements within participants, a linear mixed-effects model with participant as a random intercept was then fitted. The fixed effect of ambient noise remained significant (β = 0.225, 95% CI [0.184, 0.266], P < 0.001), indicating that PLL increased by approximately 0.23 dB SPL on average for every 1 dB SPL increase in ambient noise. In addition, within-subject regression slopes were positive for 48 of 51 participants (94.1%).

For average listening level (LZeq), a two-way repeated-measures ANOVA showed significant main effects of location (F(3.26, 163.73) = 31.57, P < 0.001, η²_G = 0.106) and genre (F(2.24, 112.07) = 153.63, P < 0.001, η²_G = 0.277). Post hoc comparisons indicated that the library yielded the lowest overall LZeq (mean = 63.05 dB SPL, averaged across genres) and was significantly lower than all other environments (all P < 0.001). The gym yielded the highest overall LZeq (mean = 70.87 dB SPL) and was significantly higher than the on-campus condition (P = 0.0008) and the library condition (P < 0.001).

For genre, new age was associated with lower LZeq values than ballad, dance, and pop (all P < 0.001; Cohen’s d = 1.16–1.44). Dance elicited the highest LZeq (mean = 72.03 dB SPL), exceeding ballad (P = 0.001) and pop (P = 0.006), whereas no significant difference was observed between ballad and pop. The location × genre interaction for LZeq was not significant, indicating that genre-related differences in average listening level were consistent across environments. Distributional patterns and post hoc comparisons are shown in Figure 3A–B.

Figure 3.

Main effects of listening environment and music genre on preferred (LZeq) and peak (LCpeak) listening levels. Box plots show the distribution of (A) LZeq by environment, (B) LZeq by genre, (C) LCpeak by environment, and (D) LCpeak by genre. Asterisks denote statistically significant pairwise differences identified by Bonferroni-corrected post hoc comparisons: **P < 0.05, *** P < 0.001. Boxes represent interquartile range (IQR), horizontal lines represent medians, and whiskers extend to 1.5 × IQR.

Peak Listening Level (LCpeak)

Descriptive peak listening levels (LCpeak) by environment and genre are presented in Table 3, and distributional patterns and post hoc comparisons are shown in Figure 3C–D. For peak listening level (LCpeak), significant main effects of location (F(3.03, 151.70) = 17.27, P < 0.001, η²_G = 0.058) and genre (F(1.39, 69.65) = 30.19, P < 0.001, η²_G = 0.113) were observed. Peak levels in the library were significantly lower than those in all other environments (all P < 0.001), whereas no significant differences were observed among the on-campus, off-campus, cafeteria, and gym conditions.

Unlike LZeq, LCpeak showed a significant location × genre interaction (F(7.16, 357.78) = 6.48, P < 0.001, ηG² = 0.027), indicating that environment-dependent changes in peak levels differed across genres Figure 3C–D).

DISCUSSION

The present study investigated how PLLs vary with real-world ambient noise and music genre using a real-time synchronized measurement approach. Ambient noise differed substantially across environments, with the lowest levels observed in the library and the highest in the cafeteria and gym. Across all observations, average listening levels (LZeq) showed wide inter-individual variability (43.0–99.8 dB SPL). Two key patterns emerged: (i) average listening level (LZeq) was independently influenced by both environment and genre, with no interaction, and (ii) peak listening level (LCpeak) showed a significant environment × genre interaction, indicating that transient peak exposure depends on the combined influence of listening context and music characteristics.

Influence of Ambient Noise on Preferred Listening Levels

Our findings indicate that listeners increase their average PLLs as ambient noise rises. Participants selected the lowest overall LZeq in the library and the highest in the gym, corresponding to an approximately 7.82 dB shift between the quietest and noisiest environments. This magnitude is consistent with the noise-compensation effect reported in controlled studies, in which PLLs typically increase by about 7–10 dB in the presence of background noise.[8,20] For example, Hodgetts et al.[8] reported higher preferred listening levels under multitalker babble and street-noise conditions than in quiet. A key strength of the present study is the synchronized pairing of earphone output with contemporaneous ambient noise, allowing the ambient noise–PLL relationship to be examined in the joint distribution that listeners actually experience.

As shown in Figure 4, our field measurements span a broader and denser region of ambient-noise × listening-level combinations than those typically represented in prior work.[8,20,21] The data also reveal substantial overlap among environments and marked dispersion of listening levels across the ambient-noise continuum, suggesting that location labels are imperfect proxies for acoustic demand and that within-location variability is a defining feature of everyday listening. Because prior studies used heterogeneous measurement methods and weighting schemes, the prior-study rectangles in Figure 4 should be interpreted as contextual rather than metrologically equivalent comparisons.

Figure 4.

Ecological validity and study coverage: real-time synchronized measurements spanning prior laboratory ranges. Scatterplot of ambient noise level (x-axis; LZeq, dB SPL) versus preferred listening level (y-axis; earphone output LZeq, dB SPL) obtained from real-time synchronized field measurements across five everyday environments (on-campus, off-campus, library, and cafeteria). Present-study observations are color-coded by environment. Shaded rectangles indicate the coverage (filled areas) and boundaries (outlined boxes) reported in prior laboratory and field studies (Hodgetts et al.[8]; Portnuff et al.[20]; Muchnik et al.[21], illustrating the extent to which the current dataset spans and extends previously tested ambient-noise and listening-level ranges. Note that prior studies used varied measurement methods and weighting schemes; rectangles are provided for contextual comparison of study scope.

Nevertheless, the figure provides a concise field-based rationale for context-aware monitoring. These findings are also consistent with field evidence that portable-listening levels can be high and widely distributed in public settings.[16,22] In the present dataset, ambient noise likewise showed broad within-environment ranges, including occasional high values even in nominally quiet settings such as the library. Together, these observations highlight the value of synchronized monitoring for interpreting moment-to-moment volume adjustment.

Influence of Music Genre on Listening Behavior

Beyond environment, music genre accounted for a substantial proportion of variance in average listening levels. Across environments, new age was associated with lower LZeq values than ballad, dance, and pop, whereas dance elicited the highest mean LZeq. These findings suggest that genre-related differences in listening level reflect not only listener volume selection but also acoustical characteristics of the selected music. In particular, differences in spectral distribution, rhythmic salience, vocal prominence, and overall recording characteristics may influence how listeners judge clarity and appropriate loudness in everyday environments. This interpretation is consistent with previous work showing that music-related perceptual demands can vary across genres and influence listening-related decisions.[10,11,13] At the same time, the present study used commercially available full-length tracks without additional loudness normalization, and the selected stimuli differed in their long-term spectral characteristics. Accordingly, the observed genre effect is best interpreted as reflecting the joint influence of music content and listener adjustment behavior rather than listener preference alone. Importantly, the absence of an environment × genre interaction for average LZeq indicates that genre-related differences were relatively stable across listening contexts. This pattern suggests that listeners may adopt a genre-dependent baseline listening level and then apply a broadly similar noise-driven adjustment across environments.

Context-Dependent Modulation of Peak Listening Levels

Peak listening levels showed a more complex dependence on context than average LZeq. For LCpeak, significant main effects of environment and genre were observed, and an environment × genre interaction emerged, indicating that changes in transient peaks across environments were not uniform across genres [Figure 3C–D]. Peak levels in the library were significantly lower than in all other environments, whereas peak levels did not differ significantly among the on-campus, off-campus, cafeteria, and gym conditions. The significant environment × genre interaction may reflect multiple, non-mutually exclusive mechanisms, including playback-system constraints such as amplifier saturation or dynamic compression. Genres with greater rhythmic salience, such as dance music, may encourage listeners to increase volume more aggressively in noisy environments. In addition, recordings with wider dynamic ranges may produce proportionally higher peak levels when listeners raise average listening level. From a hearing-health perspective, this pattern suggests that peak exposure may depend not only on environmental context but also on genre-specific acoustic characteristics. Because the present study did not directly quantify dynamic range characteristics of the music stimuli, future studies incorporating objective metrics such as crest factor or integrated loudness may help clarify this mechanism.

The dissociation between average and peak outcomes is important conceptually and clinically. Average LZeq likely reflects a sustained “set-and-maintain” volume strategy shaped by genre baseline and ambient-noise compensation, whereas LCpeak may be more sensitive to brief behaviors and stimulus-driven transients that do not scale linearly with average level. Accordingly, safe-listening guidance based solely on average level could miss situations in which the combined influence of environment and genre disproportionately elevates peak exposure.[13]

Hearing-Health Implications and Safe-Listening Guidance

The broad LZeq range observed here implies that some users may experience listening levels substantially above group means, increasing risk depending on duration and cumulative dose. World Health Organization (WHO) safe-listening guidance indicates that allowable listening time decreases as the level increases; for example, listening at 80 dB may be safe for up to 40 hours/week, whereas exposure at 90 dB may be safe for only about 4 hours/week.[5] Because the present study primarily recorded LZeq rather than LAeq-based metrics, direct quantitative comparison with WHO safe-listening thresholds should be made cautiously. Calibration comparisons using the same smartphone–earphone–2cc coupler system showed that the relationship between app-recorded levels and coupler-measured LAeq values varied by music genre and output step. Accordingly, the present findings are best interpreted as contextual evidence relevant to hearing-health risk rather than as direct equivalents of WHO A-weighted exposure criteria.

Nevertheless, preferred listening levels reached approximately 100 dB SPL in some cases, suggesting that certain real-world listening episodes may still represent elevated-risk situations, particularly when combined with prolonged listening duration. The WHO–International Telecommunication Union (ITU) standard on safe listening devices and systems emphasizes dose-based management and user-facing feedback, such as exposure tracking and warnings.[5] In this context, real-time synchronized monitoring is particularly relevant because it can identify when elevated listening levels co-occur with high ambient noise, enabling context-aware nudges such as active noise control, better-isolating earphones, or reduced volume in noisy settings.

Strengths, Limitations, and Future Directions

A major strength of the present study is the synchronized field pairing of earphone output and ambient noise across multiple everyday environments and genres, enabling direct tests of how context and content jointly shape listening behavior outside the laboratory. In addition, smartphone-based ambient-noise measurements showed agreement within 3 dB SPL compared with a reference microphone, supporting the feasibility of naturalistic monitoring.[16]

However, the sample was limited to young normal-hearing Android users, and cross-device variability in output estimation remains an important consideration for generalizability and absolute SPL interpretation. Another limitation is that participants used their personal Android smartphones, and device-specific calibration was not performed for each device. Although our previous work supported the feasibility of the Android-based measurement framework,[23] some residual hardware-related variability in absolute SPL estimation may have remained across smartphone models. Although exploratory analyses using the available participant-level variables (gender, age, and hearing threshold) were conducted, these factors did not fully explain the substantial inter-individual variability in PLL observed in the present study. This suggests that real-world differences in PLL may depend more strongly on unmeasured behavioral factors, such as music preference, habitual PLD usage duration, and listening habits. Variation by time of day may also reflect changes in everyday listening context, including studying, commuting, and social activity, further indicating that contextual factors may contribute to real-world variability in PLL. A related methodological point is that we reported Z-weighted averages and C-weighted peaks, which limits direct numeric comparison with studies and guidelines that use A-weighted exposure metrics.

Nevertheless, because A-, C-, and Z-weightings are formally defined in IEC 61672-1,[24] future work could improve comparability by dual-reporting A-weighted dose metrics alongside Z-weighted and peak measures or by validating conversion pipelines for specific device–earphone configurations. Future studies should also expand musical repertoires within genres, quantify subjective preference and enjoyment more explicitly, and extend monitoring to additional populations to determine how general these behavioral patterns are and how they relate to long-term outcomes.

CONCLUSIONS

This study used real-world, in situ, real-time synchronized monitoring to examine how preferred listening levels vary with ambient noise and music genre during everyday activities. Across environments, average listening levels increased in noisier settings, indicating that the noise-compensation pattern previously observed in laboratory studies is also evident under naturalistic, fluctuating conditions. Average listening levels also differed across music genres, likely reflecting the combined influence of listener volume adjustment and acoustic characteristics of the selected music stimuli. In contrast, peak listening levels showed a context-dependent environment × genre interaction, suggesting that transient exposure may be shaped by both listening context and music content. By capturing earphone output and ambient noise simultaneously and objectively, the present findings provide an ecologically valid view of everyday listening exposure and may help inform context-aware approaches to safe-listening guidance.

Availability of data and materials

Available from the corresponding author on reasonable request.

Author contributions

Conceptualization, G.K. and W.H.; Methodology, G.K. and W.H.; Validation, G.K.; Formal analysis, G.K. and W.H.; Investigation, W.H.; Data curation, G.K.; Writing—original draft preparation, G.K.; Writing—review and editing, W.H.; Visualization, G.K.; Supervision, W.H.; Project administration, W.H.; Funding acquisition, W.H. All authors have read and agreed to the published version of the manuscript.

Ethics approval and consent to participate

The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of Hallym University (HIRB-2016-060). Informed consent was obtained from all subjects involved in the study.

Conflicts of interest

The authors declare there was no conflict of interest.

Funding Statement

This work was supported by Hallym University Research Fund (HRF-202401-007).