Aerosol emissions from wind instruments: effects of performer age, gender, sound pressure level and bells

Study subjects

Healthy adult performers, ages 18 and older, and underage performers, ages 12 to 18, were recruited to participate in the study. All participants provided written informed consent, or assent in the case of minors (accompanied by parent/guardian informed consent), in accordance with US regulatory guidelines for research involving human subjects (experimental protocols were approved by the Colorado State University Institutional Review Board, Approval #20-10174H). We recruited adults and minors (aged 12 to 17) in roughly equal proportion and in groups of 6 to 8 performers for each of the following instruments: clarinet, bassoon, flute, oboe, piccolo, saxophone (alto and tenor), French horn, trumpet, trombone and tuba. In addition, we recruited 16 singers (half professional/adult, half minor) to characterize the singing shows. These combinations (7 participants for 10 instruments and 16 singers) yielded a target panel size of n=86. Recruitment included both males and females to assess the effect of gender (male or female; assigned to birth) on emissions.

Participants brought their own instruments to the testing facility (described below) as well as a personal mask/face covering. For safety reasons, participants were excluded from the study if they actively exhibited symptoms of COVID-19, had a previous diagnosis of COVID-19 within the previous month, or had known exposure to someone with COVID-19. COVID-19 in the previous month. 14 days (as per national and local quarantine protocols at time of study).

Each participant performed a series of maneuvers specific to their specialty and level of ability, during a measurement session of approximately 2 hours. These maneuvers included the game of Balance, a prescribed directory which was provided to participants at least two weeks before the measurement session [selection]a self-selected directory at the choice of each participant [freestyle]and two generic vocal maneuvers [talking and singing]. This article focuses on aerosol emissions from trained instrumentalists and vocalists when performing scales, prescribed selections, and self-selected repertoires (generic vocal maneuver results have been published previously15). Each maneuver was repeated continuously over a period of four minutes, during which aerosol emissions were measured. Participants wore lint-free coveralls and lint-free hair nets (disposable polypropylene coveralls, McMaster-Carr, IL) in the facility to minimize particle loss from their clothing and hair, respectively. Between each maneuver, a background measurement was taken while participants wore their personal face coverings for at least two minutes and sat quietly, approximately 2 m from the aerosol collection instruments.


Participants performed the maneuvers inside a 3.45 m × 2.8 m × 2.45 m climatic chamber15.38 ventilated with HEPA filtered air. Chamber airflow was monitored (~8.5 air changes per hour) and environmental conditions (temperature, humidity) inside the chamber were recorded along with all measurement data. using LabVIEW (National Instruments, TX, version 21.0, instrument control and data acquisition software. A constant volume sampling device (10 L min−1 total flow) was used to capture aerosol emissions directly downstream of the instrument bell or the participant’s mouth (Figure S1). The sampling apparatus was mounted on a hinged frame and attached to a height-adjustable table. Participants performed standing maneuvers (except for the French horn, for which participants were seated in a chair), with the angle and height of the sampling inlet cone adjusted so that their instrument bell was positioned directly in front of and in approximate planar alignment with the center of the cone face (in the case of the singers, the entrance cone was adjusted to align with their mouth). An isokinetic sampling probe (inner diameter 0.05 m) was installed at the narrow end of the inlet cone (0.22 m from the front plane of the cone face). An optical particle counter (OPC; model 11D, GRIMM Tech.) was connected to the probe to quantify the number and size of particles between 0.25 and 35.15 μm in diameter (31 logarithmically spaced size bins at a resolution five seconds; inlet flow 1.2 L min−1). Carbon dioxide (CO2) mixing ratios were measured further downstream at 1 s resolution with a non-dispersive infrared spectrometer (LI-820, LI-COR Biosciences). The outlet of the sampling device was vented to the outside.

In a subset of performers, we also measured sound pressure levels (n = 32 participants) and the effect of bell covers (n = 67 participants) on reducing aerosol emissions from instruments. Sound pressure levels (i.e. the volume of the instrument above the ambient background) were recorded during the maneuvers at a fixed location approximately 0.3 m above the face of the cone sampling using a prepolarized free-field condenser microphone with a preamplifier (model 378B02l + 426E01, PCB Piezotronics Inc.). The bell covers were constructed from two layers of spandex and an inner layer of Halyard H600 medical wrap and sized to fit a variety of instrument bells. For particles larger than 1 μm in aerodynamic diameter, the efficiency of the bell lid was 95–99.9% (see supplement for details on the aerosol collection efficiency of bell lids). The efficiency decreased to approximately 80% at 0.5 μm (the lower size limit for this protocol). Not all participants chose to use the bell covers due to poor sizing/fitting (i.e. the covers available did not fit all instruments); a few participants chose not to use the bell cover due to perceived airflow restrictions and/or impeded playability of the instrument. Bell’s coverage results were therefore limited to instruments with 3 or more participant measurements to provide a measure of statistical confidence.

To establish background levels, participants were asked to sit in a corner of the chamber, approximately 2 m from the sampling device, while wearing a face mask. For each background measure, the participants waited, before and after each maneuver, until the total number of particles approaches 50 L−1, as determined by the OPC. During this time, basic data was also recorded for CO2 and ambient sound pressure levels. Further details on instrumentation, measurement system and background corrections are provided in the online supplement.

Data analytics

All data analyzes were performed in R (R Core Team, version 4.1.2, Time series data for each participant were averaged throughout the maneuver (each duration approximately 4 minutes). Results are reported in terms of near-field concentration (i.e. the concentration of particles measured in the sampling device immediately downstream of the instrument [particles L−1 of sampled air]), particle emission rate (i.e., the particles s−1), or CO2-standardized emission factor (i.e. near-field concentrations divided by measured CO2 mixing ratio, corrected for background noise).

We developed linear mixed models to explore how the following variables affect aerosol emissions: type and/or class of instrument (i.e. voice, woodwind, brass), type of maneuver (ranges, selection , freestyle), instrument sound pressure level (A-weighted decibels), participant’s gender (assigned at birth; male versus female), participant’s age (minor versus adult), and use of a bell cover. Models assessed maneuver, gender, age, sound pressure level, and type (or class) of instrument as fixed effects, including a random intercept term to account for correlation in the repeated measures of each participant. The measured emissions data have been transformed into natural logarithm to reduce the asymmetry of their distribution. Results are presented as the percent change in geometric mean emissions from a given fixed effect, along with 95% confidence intervals (CIs). Linear mixed models take the following general form:

$$Y_{i,j} = ln left( {emission rate_{i,j} } right) = {upbeta }_{0} + {{varvec{upbeta}}}^{ T} {varvec{X}}_{i,j} + alpha_{i} + epsilon_{i,j}$$


where ({Y}_{i,j}) represents the log-transformed emissions for Ie participant and Ith maneuver, ({{varvec{X}}}_{i,j}) represents a set of fixed-effect variables (e.g., age, gender, device class, etc.) for each measure, ({varvec{upbeta}}) is the vector of the coefficients for the fixed effects, and ({alpha}_{i}) represents a random intercept term for the participant I. The last term, ({epsilon}_{i,j}), represents the residual model error (i.e. unexplained variation) that is assumed to have a mean of zero and be normally distributed with constant error variance. We used a likelihood ratio test to assess whether there was a significant difference in geometric mean aerosol concentration between different maneuvers, genders, or ages. Statistical significance was assessed at the 0.05 level. The percentage of variation explained by the model was calculated using conditional and marginal R239.

The relationship between sound pressure level and emissions was assessed with Pearson correlations and a linear mixed model that included fixed effects for sound pressure level and its interaction with instrument type. The effect of the bell covers was assessed by using an interaction term with each instrument in a model that also controlled for age, gender, and participant. For bell cover analyses, we normalize emissions data to background-corrected CO2 mixing ratio measured downstream of the instrument, as previous work suggests that ill-fitting barriers and/or control technologies can redirect airflow orthogonally to the direction of exhaled air8,40,41. Thus, normalizing to CO2 we account for sampling losses (i.e. air that did not enter the sampling device) in the measured emissions that could differentially bias comparisons with or without a bell. We tested the effects of CO2 normalization in subsequent sensitivity analyses.

Comments are closed.