To our knowledge, this study is the first to examine more than two representatives of the same instrument, but a greater and thus more representative number of individuals for two wind instruments (oboe, flute). Moreover, the study design using a hermetically closed probe chamber and standardized playing condition allowed, for the first time, measurement of total aerosol emission rates. Other than in previous studies^26,28 the different musicians performed the same repertoire piece of music including a variety of dynamics and articulation techniques. Our experiment resembles a real performance with respect to both the scores and the playing time (20 min).

Our results show that typical wind instrument playing generates higher aerosol emissions than typical speaking or calm breathing. During realistic performance, when the musicians play Mozart Concerto as usual, we observe total emission rates in the range previously reported for singing, exceeding 1000 particles per second³³. This is in line with results by He et al. who found that playing wind instruments in general generates more aerosol than breathing and speaking, whereby the emission rate is dependent on parameters, such as dynamics, articulation, and breathing techniques²⁶. Some of the aerosol particle size distributions emitted by our probands during speaking show the laryngeal mode (L-mode) around 2 µm discovered previously³⁴ which is not observed in the emissions from instrument playing (s. fig/Histograms/Comparison Laryngeal Mode Histogram.gif in our repository³²). This is consistent with the fact that playing wind instruments does not involve vocal fold vibrations associated with voicing, thus obviating the physical mechanism underlying the generation of L-mode particles.

The breathing of the musician is the common source of both the loudness of a wind instrument and the aerosol emission. Specifically, exhalation determines sound generation^35,36,37 whereas inhalation determines aerosol generation¹¹. Higher loudness is produced by greater exhalation flow rate^35,38,39,40. A flutist can sustain a note in forte for 8 s without rebreathing, but for 40 s in piano⁴¹, so the exhalation flow rate increases by an approximate factor five from piano to forte. The aerosol concentration of exhaled air depends on the particle yield of bronchiole fluid film burst which is modulated by the inhalation flow rate¹¹. Since aerosol exhalation rate is the product of exhalation air flow rate and particle concentration of the exhaled air, the aerosol emission during playing a woodwind instrument depends on both the inhalation and the exhalation process.

A correlation between loudness and aerosol emission is observed²⁸ because the two quantities correlate, each, with exhalation flow rate as modulator. The median aerosol particle number concentrations reported in McCarthy et al.²⁸ increase by a factor five from piano to forte, which equals the expected increase of exhalation flow rate during flute playing. The increase of aerosol emission at increasing loudness is, thus, largely explainable by increasing exhalation flow rate. The particles emitted during instrument playing have a similar size distribution as for breathing while speaking and singing would differ thereof by the additional L-mode in the size distribution^11,28,34.

During musical performance a musician autonomously adapts both inhalation and exhalation to the artistic requirements⁴² so that variation of the inhalation process is ordinary part of wind instrument playing. It introduces an independent modulator of aerosol emission since quicker inhalation, which is typical for flute playing⁴¹, produces higher aerosol concentrations of the exhalate¹¹. Therefore, aerosol emissions from a flute or oboe depend on the playing style in a more complex way than straight correlation with exhalation flow rate or sound pressure. Our probands played a whole Mozart Concerto rather than single notes²⁸ and playing all the different phrases with the prescribed dynamics requires most of the instrumental and breathing techniques, whereas sustaining a single tone for 20 s is often feasible without re-breathing.

When comparing the emissions during instrument playing to those from speaking, we refer to the typical loudness associated with playing the Mozart Concerto or reading the Hesse novel aloud, respectively. The music performance was louder than the reading, and the aerosol emission during playing was higher than during reading aloud. It is possible, though, to raise the voice while speaking to produce similar levels of aerosol emission as by playing wind instruments²⁸.

The large number of probands playing oboe and flute in our study demonstrated the major individual variability within the two groups. Emission rates show uniform distribution within similar ranges for the two instruments. Unlike previous studies^22,23,26, no clear allocation of emission rates to the instrument type is possible. We conclude that individual factors dominate the variability of aerosol emission rather than the type of instrument. Outliers from the uniform distribution that might be interpreted as super-spreaders have not been observed, other than in a previous study that detected high aerosol emitting probands during speaking⁸.

In search for individual factors influencing the aerosol emission we found that emission rates do not correlate with body height or weight³². Hence, we assume that the breathing technique and the respiratory rate are probably the reason for individual variability of aerosol emission, as outlined in a recent study⁴³.

The humidification of exhaled air takes place in the upper respiratory tract^44,45 whereas aerosol formation is thought to originate deeper in the respiratory tract¹¹. Since the humidified air is saturated with water even at high flow rates⁴⁶, water emission likely correlates with pulmonary ventilation rate. Our results indicate higher aerosol-particles-per-water ratios for oboe playing than for speaking. Given that wind instrument playing requires higher pulmonary ventilation rate than speaking, our results are consistent with an increased air exchange in the respiratory tract during wind instrument playing. The required pulmonary volume apparently depends on individual factors, such as vital capacity or breathing technique, which explains the high variability of aerosol emission within the two instrument groups. We found noteworthy correlations between the water emission rates from wind instrument playing, speaking, and breathing indicating that the respiratory volume needed for the respective task might increase similarly for all the different individuals.

Regarding the particle size distribution, most of the particles are < 1 µm in diameter, as found previously for breathing and speaking probands³³. The SARS-CoV-2 virus has a diameter of 0.13 µm^47,48. An investigation of the load distribution of SARS-CoV-2 virions in airborne aerosol over different aerosol particle size bins revealed that aerosol particles smaller than 1 µm carried 67% of the total number of genome equivalents per cm³ in an air sample⁴⁹. This imposes great risk for long-range COVID-19 transmission since particles < 2 µm reach alveolar parenchyma. Consistently, particles with equilibrium diameters ≤ 1 µm emitted during breathing, speaking, and singing have been causing indoor airborne long-range COVID-19 transmission with attack rates as high as 89% (51 secondary infections among 57 susceptible exposed)⁵. Even particles emitted by infectious individuals during tidal breathing contain aerosolized SARS-CoV-2 RNA copies, 54% of which are contained in fine particles (diameters ≤ 5 μm) labelled here as “aerosol”⁵⁰. Therefore, the “aerosol” particles emitted during instrument playing ought to be considered efficient virus carriers. The emission rates measured here are the most important input parameter of disease transmission risk calculations for the assessment of indoor situations involving the presence of potentially infectious room occupants. Particles with diameters > 6.6 µm were rarely recorded, in agreement with McCarthy et al.²⁸, thus being negligible for long-range, airborne disease transmission.

Like other researchers before, we tried to reduce the aerosol emission by masking measures. We masked the bell with a surgical mask on the oboe, clarinet, and trumpet. Except for one oboist, all participants produced similar aerosol emission rates as without mask. A previously described reduction of 50–79%^22,25 was not observed at the measurement distances used in our study. We assume that aerosol emanated through keyholes and embouchure. Moreover, the most frequent particle class with diameters < 0.8 µm is not filtered efficiently by a surgical mask. Since 4 out of 11 oboists reported a flawed intonation, especially for the notes E5 and F5, while playing with mask we refrain from recommending surgical masks as emission filters for wind instruments.

Risk assessment of typical woodwind playing situations

Short-range exposure is difficult to model, but easy to mitigate (by social distancing following recommendations, e.g., in Gantner et al. and Hedworth et al.^21,51). The opposite applies for long-range exposure, in practice. The obvious countermeasures against aerosol transmission are ample fresh air and the wearing of FFP2 masks. However, the efficiency depends strongly on the specific setting. The sole simple rule available is the recommendation to do outdoor whatever can be done outdoor. For indoor occupation, COVID-19 transmission risk can be calculated as described in Reichert et al.⁵ and implemented online for free use: https://hri-pira.github.io¹⁹.

We apply the framework outlined in Supplement S4 to assess the criticality of a few, typical situations of playing woodwinds. It is assumed that appropriate social distancing excludes short-range exposure so that the infection risk entirely results from long-range exposure. As mentioned before, the hazard in a particular scenario depends on both the individual aerosol emission rate q and the infectiousness of the instrument player⁵². In a real situation the disease transmission probability may therefore be a factor 10 less than stated below or even negligible since we assume the worst case of viral load.

For our calculations we assume the maximal infectiousness (\({Z}_{50}\) = 833 particles, for the Delta variant) and an aerosol emission rate of q = 2500 particles per second while playing, to examine whether the setting is safe or not. This question remains important even when an antigen test has been carried out before playing since asymptomatic spreaders may pass at significant rates reported with a sensitivity of 58% to 95%⁵³. A safe setting provides the necessary, second line of defense. Vaccination is neglected in the following, thus assuming susceptibility for infection.

Lesson at the music school

The teacher and an infectious student have a 60 min lesson in a 200 m³ classroom. The student listens 50%, plays 40%, and talks 10% of the time (average aerosol emission rate q = 4∙10⁶ /h). Neither wears a mask and the windows remain closed. The resulting long-range infection probability for the susceptible teacher is p = 96%. To reduce p to 10% by ventilation only, unrealistic 80 air changes per hour (ACH) sustained were necessary. If, instead, the teacher wears a tight FFP2 mask with a filter efficiency of 95% (\(\vartheta =0.05\))^54,55 then p = 15%. They may open the door and windows widely for 10 min after half an hour to clear the air from aerosols. Then, p = 79% without wearing a mask, or p = 7% wearing a mask. To conclude, acceptable safety levels can be reached even at worst-case conditions by

1. (i)

   limiting the duration to one hour,

2. (ii)

   wearing FFP2 mask whenever suitable, and

3. (iii)

   obligatory, thorough airing around half time.

Infection probability with mask is expected around 14% when the space volume of the room is half as large (100 m³).

Recital

An infectious soloist plays a one-hour program (net playing time) accompanied by two musicians in a 2000 m³ ballroom. The audience leaves the room after 90 min, including the encores and applause. Automatic ventilation exhausts air through the ceiling at 2 ACH (4000 m³/h) with fresh air streaming inward near the floor. The CO₂ level stays below 1000 ppm (good air quality) for audiences up to 100 persons. The average aerosol emission rate is q = 6∙10⁶/h. The long-range infection risk for susceptible persons is p = 3% if they wear FFP2 masks and p = 45% otherwise. Given that social distancing prevents accommodation of more than 50 spectators in the ballroom, one secondary infection case is expected when FFP2 masks are worn throughout. The reproductive number in this setting is R ≈ 1.

Symphonic performance

A woodwind player in an orchestra is infectious. They play symphonic literature, i.e., the duty cycle of woodwinds is average. We assume 30 min net playing time evenly distributed over 90 min concert duration, resulting in an average aerosol emission rate q = 3∙10⁶/h. The concert hall has a space volume of 20,000 m³. Automatic ventilation exhausts air above stage and auditorium at 2.5 ACH. The long-range infection risk for susceptible persons is p = 0.14% when they wear FFP2 masks, and p = 3% otherwise, owing to the large air space and fresh air supply.

The basic assumption of full mixing, or perfect aerosol dilution, is questionable in the latter example. Concert stages may have air exhaustion ducts which remove part of the air on stage from the hall before it mixes into the air surrounding the audience. In case of the opposite flow direction, fresh air streaming down from the ceiling, problems may arise, such as local stagnation or recirculation regions with elevated aerosol concentrations⁵¹. Generally, large premises require consideration of actual flows and should not be assessed using the well-mixed room air assumption. Our example cannot be generalized to other concert halls, based only on their size and total fresh air supply.