No one loves wearing a mask, at least no one I've met. Yet present regulations demand it, and community-mindedness warrants it. So, what mask to choose? Let's look at what's clear from scientific research on cloth masks, and where the information is muddy.

Fit is key. If you'll remember only one thing from this article, remember that fit is key. If the air flows around the mask, through gaps next to your nose, out the sides... that air isn't being filtered. A leaky mask filters about half as much as a well-fitting mask [1,6].

Fabric matters. Different fabrics filter differently. The data suggests that home-made masks can be nearly as good as N95s, though most available for purchase are not. Several scientific labs have measured the filtration efficiencies for various fabrics, that is, what fraction of particles are blocked. Also important is how easily the air flows through the fabric, that is the pressure drop. If the mask is not breathable, then when you cough the air pressure will lift the mask and the air will flow around, rather than through, the mask.

• Cotton woven is a staple of home sewn masks. Two layers of cotton woven filter 49% to 61% in three studies [1,3,8], and 57% to 68% in a fourth study [10]. The results are fairly consistent across studies.
• Cotton knits vary in their filtering properties, with reported filtration rates as poor as 27% for a 3-layer mask, and as good as 88% for a 2-layer heavy t-shirt [1,3,7,8,10]. A multi-layer knit cotton mask is a gamble, might be great, might be awful. Single-layer masks are awful, filtering 10% to 25% if fitted perfectly [11].
• Synthetic knits: Like cotton knits, polyester and nylon knits vary, and generally but not always perform less well than cotton. Filtration efficiencies of 9.5%, 18.5%, 23.6%, 38%, 39%, 41.2%, 47%, 53%, 56% are reported for eight different masks [1,7,10,5]. Doubling a neck gaiter (95% polyester, 5% spandex) improved its filtration from 47% to 60%.
• Non-woven synthetic filters: Some home sewers add non-woven polyester interfacing or non-woven polypropylene. Since N95 masks are made with spunbond non-woven polypropylene, it can be tempting to incorporate OlyFun (a spunbond non-woven polypropylene used for crafts) or a non-woven polyester interfacing (a similar material). However, medical grade masks rely on the electret effect of electrostatically charged fibers, nicely explained in this YouTube video. Data is lacking on how well non-charged interfacing substitutes for charged polypropylene. One study showed that one cotton woven layer plus one non-woven polyester layer (Pellon 931TC midweight) filtered equally well as three cotton woven layers [8], about 65%.
• Procedure masks: The filtration efficiency of procedure mask fabric is better than most fabrics, though rarely as good as N95s. The Achilles heel of procedure mask is their fit, with many leaking significantly out the sides. Studies of procedure mask fabric show filtration efficiencies of 65% to 98% with most on the higher end of that range [1, 3, 8, 10, 4], but when worn the effectiveness drops significantly due to air leaking through gaps. In [1], a medical procedure mask worn normally filtered at 39%, but when a nylon stocking was used to force a perfect fit the filtration was twice as good, at 80%.

What to make of all that?

• If you're wearing a mask because the rules say so but are not personally worried about contracting COVID-19, wear a mask that fits well and has 2 or more layers. Fit seems to affect overall performance more than the choice of cloth, and a well-fitting mask won't distract you by requiring adjustments so you're more likely to wear it. A nose wire helps seal at the top, reducing fogging of glasses. Even mediocre masks help public health. Imagine that my mask 50% of the virus that I exhale, and I meet 10 colleagues who also wear masks that filter 50% of what they inhale. Imagine that I was infected with COVID-19 four days ago, so while I'm emitting virus today my symptoms will not start until tomorrow. My mask halves my emissions, and thanks to my colleagues' masks their exposure is further halved. Our masking is reducing exposure by a factor of 4. For public health, that's a big win. Thanks to masks, I am likely to infect 4 times fewer people, or perhaps that 4 times lower exposure will result in my colleagues having mild, rather than severe, cases. The same scenario with 25% effective masks roughly halves the risk of transmission (0.75 x 0.75 = 56%).
• If you're wearing a mask to protect yourself, a cloth mask alone may not suffice. One alternative is to use a KN95 or KF94 mask, certified to the Chinese KN95 or Korean KF945 standards rather than the American N95 standard. Unfortunately the market is a mess and many KN95 masks don't actually perform as advertised ([12] found that 9 of 15 did not filter >95%), but most do out-perform cloth masks (12 filtered >70%). A second alternative is to find a cloth mask that fits you without leaks, and wear a procedure mask underneath. You can rely on the procedure mask for filtration while the cloth mask ensures a good fit.

Reusing N95 and procedure masks: The electrostatic charging so important to filtering particles in the 0.3um to 1um range fades with mask use. Yet medical professionals have been asked to reuse their N95 masks. How much worse is a used mask from a fresh mask? One study [4] looking to investigate a method for recharging used masks found that washing a KN95 in a washing machine resulted in filtration dropping from 95% (new) to 75% (washed). That's a significant drop, but the used KN95 is still better than a two-layer cotton mask.

The messy icky mask market: Most of what an American can readily buy is a gamble. N95s are still in scarce supply, with no centralized federal effort to increase production. As already mentioned, many KN95s and procedure masks vary in quality, some failing to meet specifications. Cloth masks are unregulated and for the most part untested. Non-medical non-woven masks are similar, with many providing inadequate information to even guess at their filtration. The U.S. Food & Drug Administration (FDA) put out a Personal Protective Equipment (PPE) Emergency Use Authorization (EUA) for N95 respirators which lists approved and unapproved N95 manufacturers. That helps, but figuring out the manufacturer of any given mask is tricky, and oftentimes the Amazon photos don't match the manufacturer photos resulting in concerns about fakes.

Particle sizes: There's a gap between public health policy and science regarding what size droplets are of most concern for COVID-19 transmission. If public health is right and the key job of a mask is to block >10um particles, then cloth masks may actually do far more good than the studies I summarized suggest. Cloth masks are better at filtering large particles than submicron particles. In one study, masks made of polyester and other synthetics filtered 5-10um particles far better than 0.3-0.5um particles, three to six times better [5]. Most of the data summarized here tested with sub-micron particles. If the public health assumption of >10um particles is correct, the practical performance of cloth masks may far exceed the data summarized herein.

Alternately, public health guidance may be inadequate. The ubiquitous 6ft rule, for example, assumes large droplets which quickly fall thanks to gravity, with lesser concern about tiny droplets ("aerosols") which may float in the air for a long time. The scientific publications I've read, in contrast, suggest things like: "Droplets of the order of 10um exists for up to a few tens of a second, while very large droplets 100um in diameter surive for up to almost a minute" before they evaporate leaving viral particles to float in the air [9]. In comparison, the virus itself is around 0.1um, and N95 masks are specified for 0.3um particles.

Dig deeper: For a more thorough discussion of which study said what, and links to the underlying studies, go to:
cloth-mask-supplement.html