Skip to main content

Viral Load of SARS-CoV-2 in Respiratory Aerosols Emitted by COVID-19 Patients while Breathing, Talking, and Singing

Our take —

The study directly measured the viral shedding in aerosolized particles during talking, singing, and breathing. Among 22 participants from Singapore, they found that 41% had no viral RNA in their aerosols, but that singing released more viral RNA than talking or breathing, among those with viral RNA in any of their aerosol samples. Having detectable viral particles was more likely if the sample was collected earlier post symptom onset. SARS-CoV-2 antibody levels were not associated with viral RNA in aerosolized particles. The study was limited by the small number of participants, with most viral RNA copies recorded among just 2 individuals. Therefore, it is unclear how well these findings represent all people infected with SARS-CoV-2. However, the study provides evidence that singing has greater risk for transmitting viral particles compared to talking or breathing, regardless of underlying characteristics of a patient.

Study design


Study population and setting

The objective of the study was to characterize viral loads in coarse and fine respiratory aerosol particles among patients >=21 years of age and positive for SARS-CoV-2 infection by reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Participants were recruited from February to April 2021 at the National Centre for Infectious Diseases in Singapore, regardless of clinical status. Participants were admitted for inpatient isolation and evaluation before being transferred to isolation facilities. No severely ill patients were included in the study. Expiratory samples were collected from breath, with participants facing a cone which draws in air continually. Participants completed three activities: 30 minutes of breathing, 15 minutes of talking, 15 minutes of singing. Aerosols were collected and two sizes were defined (coarse and fine using a 5-micrometer threshold) for analysis. Participants also completed demographic questionnaires and answered symptom checklists on 7 pre-specified symptoms. The day of diagnosis was also recorded with previously conducted SARS-CoV-2 serology noted in the medical records as well. Day of illness sampling was determined among symptomatic patients as from the day symptoms began, while for asymptomatic and presymptomatic patients, it was defined as the day of PCR-positive clinical sample diagnosis. Virus genome sequences were also obtained from the National Public Health Laboratory. They used Fisher’s exact test to compare categorical variables, Mann-Whitney U tests for continuous variables, and Kruskal-Wallis tests to compare median viral loads between participants with positive SARS-CoV-2 aerosol detection vs. negative aerosol detection.

Summary of Main Findings

The study used data from 23 participants, and 1 withdrew before sample collection. Of the 22 participants, 13 (59%) had detectable SARS-CoV-2 RNA in their aerosol samples. Of these13, 3 participants were asymptomatic and 1 was presymptomatic at the time of collection. There was no evidence of significant associations with age, sex, type of virus variant, clinical symptoms, or presence of SARS-CoV-2 antibodies at time of diagnosis with detectable SARS-CoV-2 RNA in their aerosol samples. However, patients with detectable viral RNA in aerosols were sampled a median of 3 days post-illness onset, compared to a median of 5 days for individuals without detectable viral RNA in aerosols (P-value: 0.025). Two participants whose samples were collected on day 3 of their illness accounted for 52% of the total viral load captured by the study overall. Six participants (27% of the 13) had detectable levels of RNA from all three activities. Two (9%) had detectable levels only from fine particles collected from speaking, while another 2 only had detectable levels from fine particles collected from singing. The amount of detected viral RNA also differed by activities, with the median number of viral N gene copies in singing was 713.6, compared to talking with 477.9, and breathing with 63.5 (p 0.026). Among participants overall, 16 (73%) were infected with a variant of concern or a variant of interest for SARS-CoV-2.

Study Strengths

The major study strength was the direct testing of aerosol particles, which they collected through standardized activities. This is in contrast to other environmental studies which conduct ambient testing after a potential outbreak has occurred. This method is likely more accurate in measurement. They also were able to conduct quantitative PCR, which put a continuous number to the viral RNA copies, as opposed to less granular assessments. They also conducted testing for variants of concern or variants of interest to see if different variant strains were more highly associated. They also were able to compare both fine aerosols and coarse aerosols that may be blocked at different rates by masking or air filtering methods.


The primary limitation was the small sample size. There was not enough power in their statistical analyses to quantify the difference in the average amount of RNA between individuals given the high amount of person-to-person variation. Additionally, they could not disaggregate this further by symptom status (e.g., asymptomatic vs. symptomatic) to better describe potential viral shedding patterns between participants. Given the diversity of strains, and that only 4 study participants were infected with variants that were neither of concern nor interest, they were not able to determine differences between strains, or groups of strains, and associated aerosol particles. Additionally, 2 participants accounted for the majority of the viral load detected, which may mean these results are most reflective of those participants, rather than fully generalizable to the rest of the sample or the at-large population. The study also passively ascertained antibody serology testing, which was likely impacted by the time at which the antibody specimens were collected and may not reflect the antibody levels the day of the actual aerosol sample collection. It is not possible to predict how this may bias potential associations with viral shedding in aerosol specimens.

Value added

This is the first study to directly measure the viral shedding in aerosolized particles during common activities.

This review was posted on: 18 October 2021