The advent of widespread community masking during the SARS-CoV-2 pandemic has spurred new research exploring the effects of face coverings on acoustics and speech comprehension. This post will begin with an overview of the studies by Corey et al and Atcherson et al which evaluate the acoustics of face coverings and face shields by looking at their effect on sound (Decibels) in the frequency (Hertz) range of human speech. We then put their results in the real-world context of how speech is produced (volume and frequency of vowels and consonants) which provides greater insight into how face coverings can muffle and distort speech and how this varies with the different types of face coverings selected. Finally, with this new knowledge we discuss tips on selecting fabrics and masks for different situations, as well as ways to adapt our communication strategies to the needs at hand.
The term “face covering” refers to masks (e.g. cloth, surgical) and respirators (e.g. N95) that can seal to the face. MakerMask does not include face shields in this definition as these allow for unimpeded airflow around the shield.
Long before the SARS-CoV-2 pandemic, scientists had already undertaken research on the effect of a variety of face coverings such as surgical masks, helmets, scarves and niqabs on speech intelligibility. Research centered around safety (e.g. whether essential commands are understood when someone is masked), inclusivity (e.g. services for the hearing impaired), the relative importance of visual versus auditory cues in communication, and the extent to which background noise impacts speech intelligibility.
These studies can be grossly categorized as either, 1) speech attenuation studies, or 2) speech intelligibility studies. In this post we look at speech attenuation studies.
Speech attenuation studies measure the sound reduction (decibel loss) caused by face coverings (versus no covering) across the frequency range important for the human voice (i.e., 125 Hz to 8,000 Hz). They may use a loudspeaker or a human speaker to emit sounds (or speech) and a microphone to pick up these sounds for analysis.
More specifically, we focus on two of the speech attenuation studies that came out in 2020/2021 that evaluated a wide range of face coverings and face shields:
The Corey et al and Atcherson et al studies are very similar in design.
Both looked at a wide range of transparent and non-transparent face coverings including surgical masks, respirators, cloth masks, window masks, and shields. (See images below.)
Corey et al analyzed sound attenuation from two sources: 1) a styrofoam head shape fitted with a loudspeaker, and 2) a human talker. The microphone was positioned 2 meters from these sources to reflect physical distancing.
Atcherson et al analyzed sound attenuation from a single source: a styrofoam head shape fitted with a loudspeaker. This study varied the position of their microphone: 3 feet and 6 feet from the source.
This post focuses on the set ups that are comparable across both studies:
Corey et al: loudspeaker analysis taken at 2 meters (~ 6 feet)
Atcherson et al: loudspeaker analysis taken at ~ 2 meters (6 feet)
Face coverings studied in Corey et al
1. Surgical
2. KN 95 respirator
3. N95 respirator
4. 2 layers jersey
5. 2 layers plain
6. 3 layers (cotton/spandex)
7. 2 layers (cotton/spandex)
8. 2 layers (plain/denim)
9. 2 layers cotton bedsheet
10. Cloth with vinyl window
11. Cloth with PVC window
12. Shield
See study for more details on each face covering.
Face coverings studied in Atcherson et al
I.1 Surgical
I.2 Surgical
I.3 KN95
I.4 N95
I.5 PM2.5
I.6 Cotton
I.7 Cotton
II.1 Shield
II.2 Window
II.3 Window
II.4 Window
II.5 Window
II.6 Partial shield
II.7 Shield
II.8 Shield
See study for more details on each face covering.
Study Results
The graphs below from each study show the decibel difference across the human voice frequency spectrum between the recordings without versus those with face coverings or face shields.
Results from Corey et al
Results from Atcherson et al
Figure A
Figure C
Figure B
Figure D
What is immediately apparent in both studies is that the difference in decibels (shown on the vertical axis) varies with frequency (measured in Hertz on the horizontal axis). In general there is relatively little decibel loss below 1000 Hz (frequently up to 2000 Hz). However, this is followed by precipitous losses between 2000 Hz and 8000 Hz. In other words, compared to being completely unmasked, there is greater volume loss caused by face coverings and face shields in the higher speech frequencies.
While the above general statement is true across all the face coverings and face shields tested, there are important differences in the amount of sound attenuation observed between the various mask and shield options.
Face Shields and Window Masks (Transparent Options)
Face shields showed the greatest variation in both decibel gains and losses. As shown in the graph from Atcherson et al there is an increase of roughly 5 decibels in the 500 to 1000 Hz range followed by a precipitous drop ranging between 9 and 17 decibels beyond 1000 Hz (Figure D). The results from Corey et al (only one shield option) are consistent with Atcherson et al(Figure A).
Window masks showed slightly less amplification between 500 and 1000 Hz followed by a significant loss beginning at a slightly higher frequency than the shields in the Atcherson et al study (Figure D). In Corey et al the plot lines for the window masks were similar to the shield (Figure A).
Both face shields and window masks show a significant “resonance-like” peak at high frequencies (above 5000 Hz). In Atcherson et al two window masks (II.2, II.3) demonstrated significant peaks, one of which (II.3) resulted in sound amplification. This variability in sound attenuation and amplification could result in perceived sound distortions.
Masks and Respirators (Non-Transparent Options)
The graph from Atcherson et al provides a direct comparison between surgical masks, respirators and a cotton mask (Figure C). The surgical masks produced the least amount of sound attenuation followed closely by the cotton mask. The respirators resulted in the greatest attenuation, but still lesser attenuation that that observed for face shields and window masks.
Corey et al looked at a greater variety of cloth masks (Figure B) and evaluated different fabric combinations. While they do not show a direct comparison of the surgical masks (Figure A) to the cloth masks, they state that “the cloth masks varied widely depending on material and weave. The 100% cotton masks in jersey (4) and plain (5) weaves had the best acoustic performance and were comparable to the surgical mask.” The worst performer was the two-layer bed sheet mask with a roughly 15 decibel loss at 8000 Hz compared to the second-worst choice of denim with an approximate 10 decibel loss at 8000 Hz.
It is worth noting that the fabric masks in Corey et al are all 2-layer cloth masks (Atcherson et al did not specify the fabrics used for the cotton options). However, at the time they conducted their work the CDC was advising 2-layer masks. Current recommendations from the WHO and Health Canada are for 3-layer masks and MakerMask recommends 3-layer masks.
Acoustic Implications of Masking
The results from Corey et al and Atcherson et al demonstrate the important role acoustics play when masking in addition to the usual considerations of filtration, breathability, and fit. In order to understand how this plays out in the real world it is worth taking a deeper dive into the acoustics of speech, in particular the difference between vowels and consonants and how that is impacted by masking. This knowledge, coupled with the study results will provide greater understanding and context of the decibel loss (aka sound attenuation) associated with masks and shields. This can then help you choose the right mask for your needs and select appropriate fabrics. Finally, this knowledge can help you adapt your communication strategies.
Acoustic factors to consider
Acoustics of speech and masking
Speech naturally varies in volume (decibel), pitch (frequency), and speed. In addition to the modulation that conveys emotion and emphasis, there are Intrinsic differences in volume and pitch related to how we produce vowels and consonants. This section looks at how these intrinsic factors will be affected by face coverings.
Within any given conversation, vowels are louder than consonants. In addition, vowels and voiced consonants (those that vibrate in the throat, e.g. B, Z, V, D, M, R) are typically produced at lower frequencies (400 to 2000 Hz) than unvoiced consonants (those without throat vibration e.g. P, S, F, T, at 2000 to 8000 Hz). The image below situates these sounds along the volume (Y axis) and frequency (X axis) continuum in the context of human speech and hearing. While ranges shown are average adult ranges, the relative position of each sound remains the same regardless of age and gender. In other words the sound bubble for a child would be overall at a higher pitch and lower volume.
Image from: Generating and understanding speech (ecophon.com)
Taking one graph from Corey et al (below) and roughly superimposing the frequencies of vowels and unvoiced consonants from the image above, it becomes apparent that face coverings will have a greater effect on unvoiced consonants (blue highlight) in the higher frequency range than on the vowels in the lower frequency range (yellow highlight). This is a double effect whereby the unvoiced consonants that are naturally produced at lower volume are the sounds that are the most affected by masking. This video by DPA Microphones also provides a clear explanation of this phenomenon.
Image from Corey et al adapted by MakerMask to include colored columns indicating the frequency ranges of vowels and unvoiced consonants. (Highlighting added by MakerMask).
Understanding Decibel Loss
Comparing the real world implication of the sound attenuation results from Corey et al and Atcherson et al to measures of hearing loss provides us with an understanding of the real world implications. For example, according to the American Speech-Language-Hearing Association (ASLHA),
A loss of 16 to 25 decibels is considered a “slight” hearing loss. This loss is only slightly greater than the loss associated with transparent face coverings and face shields in both Corey et al and Atcherson et al in the range above 1000 Hz.
A loss of under 15 decibels would be considered “normal”. This “normal” hearing loss range is comparable to the results for many of the non-transparent face coverings.
When considered alone, the sound attenuation attributed to face coverings and shields may not seem significant. However, most of the time we are communicating in less than ideal circumstances. Sound loss can be compounded by environmental factors such as background noise, plexiglass barriers, individual vocal characteristics, as well as any existing hearing loss. It is worth noting that according to the US National Institute of Health (NIH) “15% of American adults aged 18 and over report some trouble hearing” and “25 percent of those aged 65 to 74 and 50 percent of those who are 75 and older have disabling hearing loss”. (See: Quick Statistics About Hearing | NIDCD (nih.gov)) In addition, the sound attenuation created by face coverings is similar to age-related hearing loss (Presbycusis) with the loss occurring more prominently at higher frequencies compared to lower frequencies.
Selecting A Face Covering
Type Of Face Covering: Transparent or Non-Transparent
The findings from both Corey et al and Atcherson et al demonstrate that there is no perfect solution between transparent and non-transparent options. On the one hand, a transparent face covering or shield provides visual cues with decreased volume and clarity, while on the other hand many non-transparent options minimize sound attenuation with a loss of visual cues. The choice between a transparent and a non-transparent face covering will ultimately depend on context and the extent to which the listener relies on visual cues including lip reading.
Transparent face coverings may be essential when interacting with people with hearing loss who rely on lip reading and visual cues. However, the choice is less clear in situations where both auditory and visual cues are important such as for primary school teachers. A compromise may be to have both transparent and non-transparent options at hand. For example, the speaker can wear a window mask when visual cues are of primary concern and when that interaction is over the speaker can switch back to their non-transparent face covering. (The use of microphones is discussed at the end of this post.)
Finally, it is worth noting that many service providers (teachers, medical professionals, etc) are required to wear both a face shield and a mask. While Corey et al and Atcherson et al did not study this combination, it is likely that the mask with shield option results in a significant additional sound attenuation.
Guidance from the CDC and WHO regarding face shields and window masks.
Shield without mask. There is no doubt that shields provide the best view of a person’s face despite the acoustic drawbacks. However, in general the CDC does not recommend face shields in lieu of masks. The reason for this is that shields do not seal to the face and they allow aerosols to circulate freely.
Clear Masks (aka window masks). The CDC issued an update that stated the following:
“Clear masks or cloth masks with a clear plastic panel are an alternative type of mask that may be helpful when interacting with certain groups of people, such as:
People who are deaf or hard of hearing
Young children or students learning to read
Students learning a new language
People with disabilities
People who need to see the proper shape of the mouth for making appropriate vowel sounds (for example, when singing)”
“In the context of COVID-19, some children may not be able to wear a mask due to disabilities or specific situations such as speech classes where the teacher needs to see their mouths. In these cases, face shields may be considered an alternative to masks, but they do not provide the equivalent protection in keeping the virus from being transmitted to others. If a decision is made to use a face shield, it should cover the entire face, wrap around the sides of the face and extend to below the chin. Caution should be taken while wearing one to avoid injuries that could break it and harm the eyes or face.”
Finally, many commenters on social media have noted that window masks frequently fog up or accumulate moisture. While some anti-fog strategies may yield results, it is important to choose wisely as most anti-fog products were not intended for use in front of the mouth and nose and they may pose hazards when used in this way. MakerMask’s post 5 Key Considerations For Window Masks provides more information for those interested in window masks.
Fabric Choices
In general, the most acoustic-friendly fabrics are the most breathable fabrics. MakerMask has demonstrated that spunbond non-woven polypropylene (NWPP) at approx 70 gsm (2 oz per square yard) is much more breathable than quilter’s cotton. (Links in the box below.) Substituting at least one layer of NWPP in place of cotton should improve the acoustic profile of a mask. Whatever combination of quilter’s cotton and NWPP you choose, MakerMask recommends 3-layer masks as per the WHO and Health Canada. See box below for many of MakerMask’s blog posts on NWPP and breathability.
Corey et al and Atcherson et al did not include cloth masks with NWPP, however based on test results by MakerMask (links in box above), it is expected that the acoustical performance of a mask that includes spunbond NWPP will yield results somewhere between the surgical mask and the cloth masks.
You can listen to an informal comparison of singer’s masks with different fabric combinations including NWPP HERE.
Breathability data as a proxy for acoustics, an example
Currently, mask standards, such as the ASTM 2100 and F3502, do not include acoustic test results. However, these standards do include breathing resistance testing (usually referred to as “pressure drop”). Since breathable fabrics generally have better acoustics, breathing resistance scores can be used as a proxy to evaluate the acoustics of a mask.
For example, the data on breathing resistance from the Colorado State University (CSU) lab (below) are consistent with the findings from Corey et al. Both Corey et al and CSU tested a mask made with cotton bed sheet fabric. In Corey et al the bed sheet mask performed poorly for sound attenuation. Similarly, in the CSU lab the bed sheet mask (percale high thread cotton, highlighted in brown) performed poorly for breathing resistance. (A higher breathing resistance score indicates a less breathable mask.)
Breathing resistance data from Colorado State University
Source: Comparison of Cloth Mask Performance (colostate.edu)
As demonstrated by Atcherson et al and Corey et al, speech attenuation is a concern that will always be present with masks, albeit to varying degrees. Some compensating behaviors have negative consequences, while other strategies improve communication.
Compensating Behaviors
When speaking with a mask, the natural reaction to muffled sound is to simply increase volume, also known as the Lombard effect. Unfortunately, this compensating strategy can have negative implications.
Vocal fatigue and damage. Sustained loud talking is hard on the voice and, over time, can affect those who depend significantly on their voice such as teachers, singers, and call center operators.
Increased viral aerosol production. Increased aerosol production, which may carry SARS-CoV-2, has been linked to increased volume. (See Gregson et al)
Non-compliance with protocols. In the effort to understand and be understood individuals may remove their mask or compromise physical distancing.
Communication Strategies With Masks
A number of strategies for communicating with masks do not require increased vocal pressure. A simple Google search for “communicating with masks” yielded a long list of useful advice that can serve us well both with and without masks. However, advice such as “speak clearly” is vague and thus hard to put into practice. As a professional singer and choir director, I have compiled and expanded upon this advice in the box below.
Communication Strategies With Masks
1) Vocal strategies
Emphasize consonants clearly. In this blog post we discussed that consonants are produced at lower decibels than vowels, with a greater decibel loss at the higher frequencies where unvoiced consonants are produced. Adding more air pressure to support consonants without increasing vowel volume results in a clearer message. Be mindful of throat tension which might accompany this greater pressure. Produce crisp consonants with a relaxed throat.
Speak slowly and practice word economy. The sound attenuation caused by face coverings requires the listener to work harder to make out individual words. Speaking more slowly allows the listener more time to process the message. However, speaking more slowly is not about slowing everything down proportionally. Rather the focus is on greater separation of words to avoid words bleeding into each other. Take the time to mindfully emphasize key words and consonants. Formulate a succinct message and omit superfluous words.
Modulate your sound using pitch and volume to your advantage. In the best of times monotone speech is harder to comprehend than modulated speech. This includes emphasis on key words that convey meaning as well as changes in pitch that indicate the end of a phrase (lowered pitch) or a question (raised pitch). Modulation adds depth to communication beyond the words spoken.
Use body language and visual cues. Hand gestures and overall body language assist communication by conveying instructions or communicating a mood. Even facial expressions that depend on the mouth (smiles) can be seen in the eyes and cheekbones. Body energy often translates into the voice. Voice-over actors often put their whole body into acting to make sure emotion and emphasis is heard in the voice.
2) Conversation strategies
Grab the listener’s attention and plan ahead. Before you begin any interaction ensure that you have the listener’s undivided attention and minimize distractions. Verbalize your expectations at the beginning of the conversation (e.g. “I have three topics to discuss…”) to retain the listener’s attention. A bit of preparation can go a long way such as providing a doctor with a written list of concerns to be discussed.
Consider your audience. In addition to the sound attenuation caused by face coverings, many other factors affect communication and require attention to one’s audience. Atcherson et al note: “Compounding the problem further are other health-related and communication influences, such as illness, cognitive decline, speech impairments, voice disorders, foreign accents, and speech dialects.” Worth adding to that list are considerations such as stress/distress and safety concerns where misunderstanding commands could be dangerous and even life threatening.
Rephrase rather than repeat. Face coverings and shields can exacerbate misunderstandings. Some words may simply be more difficult to project through a mask, particularly industry jargon unfamiliar to the listener. Rephrasing may avoid problematic words and can provide additional context. Simply repeating a message often causes more frustration to both speaker and listener.
Request confirmation. In some situations, particularly where critical information is being conveyed, it is useful or even required that the listener confirm their understanding of the information received.
3) Environment
Avoid background noise. Competing with a noisy environment will put a strain on communication.Taking the time to move to a more quiet location or reschedule a conversation can avoid significant frustration and strain on the voice. If within your control, wait for a room to be silent before addressing the crowd.
4) Technology
Use a microphone. Amplification can go a long way to mitigate vocal strain. While it cannot eliminate the muffling caused by a mask, it can capture the softer and higher pitched sounds such as the unvoiced consonants. Corey et al studied the use and placement of a lapel microphone and even suggested that the placement of the microphone can make a difference. It is worth testing various options.
Conclusion
The two studies reviewed here (Corey et al and Atcherson et al) validate the experience we’ve all had with voice muffling when wearing masks. Their findings, coupled with an understanding of speech volumes and frequencies, explain why conversation is challenging. Furthermore, we frequently communicate in environments that are far from ideal (background noise, etc) with people who may have hearing loss. In some cases this is a mere nuisance, while in other cases the consequences may be more serious such as when a healthcare professional communicates important treatment options. With this knowledge in hand we can make wiser mask choices and implement communication strategies and habits that will allow us to communicate more effectively while masked, and will serve us well in the future well beyond the current pandemic.
Joan Fearnley, economist, soprano, and choir director, is the creator of the Open Source Mask for Singers Project (summer 2020) and is the creator of the only freely available pattern of a mask for singers. You can find her on YouTube and in her international Facebook group Masks for Performers where she supports a community dedicated to innovation in masks for singers, performers and anyone that might benefit from a more structured mask. She has co-authored two other blogs for Maker Mask (Constructing Masks for Singers and Double-Masking and the Science of Fabric Layering) as well as blogs for Choral Canada on masking for choirs (here and here). She is a member of the Cloth Mask Knowledge Exchange advisory committee (clothmasks.ca) led by Dr Catherine Clase from McMaster University. CBC recently profiled Joan’s open-source mask adventure in this web report and interview.
One thought on “Mask Acoustics”
Comments are closed.