Joe Biden has counted on face masks to control the Covid-19 pandemic. On his first day in office, the president issued an executive order requiring masks on federal property and called for their widespread use for 100 days. More recently, the Centers for Disease Control has mandated masks on airplanes and mass transit, though most airlines and transit systems already require them.

These recommendations mark the latest step in a thorough reversal of the pre-pandemic public-health consensus that masks are useless or counterproductive. Yet the evidence underlying this reversal turns out to be quite weak. Widespread use of masks throughout the United States and Europe has failed to stop massive pandemic waves in the fall and winter. Preliminary signs also suggest that mask mandates may be causing considerable harm. Several state governments have recently rescinded their mask mandates, and a look at the science suggests that more should consider following.

Researchers have investigated masks for disease protection since the discovery, in the late nineteenth century, that many respiratory pathogens spread in droplets of exhaled moisture that follow ballistic paths extending a few yards. It seemed logical that a device to block relatively large droplets could also block smaller pathogens, both protecting wearers and serving as “source control,” stopping them from infecting others.

Later research has shown, however, that respiratory droplets obey a more complicated set of physical laws. In particular, once exhaled, they shrink rapidly through evaporation. One widely cited model estimates that droplets with diameters smaller than about 100 microns (a micron is a thousandth of a millimeter) evaporate before reaching the ground, leaving their contents as long-lasting aerosols; particles smaller than about five microns can stay aloft indefinitely and travel beyond droplet range. Individual coronaviruses are about 0.1 microns across, smaller than the pores of surgical masks (0.3 microns to 10 microns) and minuscule compared to those of cloth masks (80 microns to 500 microns).

In practice, other factors can inhibit complete droplet evaporation. But evaporation still reduces relatively large droplets to sizes small enough to stay airborne. And numerous case studies, such as a choir practice in which one singer infected dozens of members beyond droplet range, show that airborne Covid-19 transmission takes place—as does rare outdoor transmission, as outdoor air currents rapidly clear away aerosols but have little effect on droplets. So while masks may stop short-range, face-to-face spread from large droplets, they are likely less effective—and perhaps completely ineffective—at stopping airborne spread from aerosols.

It’s instructive to compare the first two epidemics involving wide usage of masks. To combat the Manchurian Plague, a bacterial epidemic in northern China in the winter of 1910-11, the doctor Wu Lien-Teh devised a cloth mask that tightly covered the nose and mouth. A report found that “the careful use of the mask appeared to confer absolute protection.” But masks proved far less useful in the subsequent 1918 Spanish flu, a viral disease spread by pathogens smaller than bacteria. California’s Department of Health, for instance, reported that the cities of Stockton, which required masks, and Boston, which did not, had scarcely different death rates, and so advised against mask mandates except for a few high-risk professions such as barbers. The Surgeon General of the United States Navy warned that masks were “designed only to afford protection against a direct spray from the mouth” and could even spread disease if used improperly. The epidemiologist Warren Vaughan used even blunter terms: “Certainly the face mask as extensively used during the 1918 epidemic was of little benefit and in many cases was, without doubt, a decided detriment.”

Nevertheless, doctors have continued to study the use of masks to protect against viral diseases. The studies vary widely, starting with the populations they consider—many studies use hospital workers as subjects, and their results may not generalize perfectly to nonhospital environments with different balances of aerosol and droplet transmission. Studies also differ in the type of masks they consider: some consider surgical masks, others makeshift cloth masks, still others N95 masks, which seal tightly around the mouth and nose. The latter protect the wearer well if worn correctly, but they are more expensive and limited in supply, are uncomfortable and difficult to wear correctly, and generally do not filter exhalations, thus providing no source-control benefit. Most discussions about mask mandates for the public have thus focused on cloth or surgical masks. Cloth masks, for their part, have proved far inferior to surgical masks in tests of filtration efficacy. One large randomized trial found that hospital workers who wore cloth masks were 6.6 times as likely to contract flu-like illnesses as those who wore surgical masks—a difference that the study authors speculated may be because cloth masks confer greater infection risk than no mask at all.

Some studies directly test whether masks filter air particles, either on volunteers connected to breath analyzers or on mannequin heads with breathing simulators. These studies show widely varying effectiveness. Other limitations on their real-world applicability exist, too: the fit of a mask in a supervised study or on a mannequin is likely better than in real-world use; experiments often last only a few minutes; and simulated breath likely degrades mask performance more slowly than real, humid breath. (One influential 2006 paper on influenza control claims that masks quickly become ineffective as “the pores in the mask become blocked by moisture from breathing.”)

Furthermore, few studies look specifically at the particles with diameters five microns or smaller that are responsible for airborne transmission. The studies that do examine those particles give equivocal results. One small study found that surgical masks filter exhaled particles smaller than five microns for volunteers sick with mild coronaviruses (based on a sample of only 17 patients), but not flu viruses or rhinoviruses. Another 2009 study tested filtration rates for five FDA-approved surgical masks fitted to mannequin heads with exhalation simulators. They provided widely varying and generally highly imperfect filtration, with the worst mask letting through more than 90 percent of particles in the 0.1–0.4-micron range in one test, even though masks were sealed to the mannequin’s face with silicone, providing far better filtration than real-world use. The authors concluded that “surgical masks should not be used for respiratory protection.”

Real-world studies, mostly focusing on protection rather than source control, are also equivocal. Some observational studies seem positive: for instance, one study of SARS in Beijing found that 94 SARS survivors were far less likely to have worn masks outside the home than 281 residents in their areas who had not contracted SARS. A study of Covid-19 among 124 families in Beijing found that wearing masks before the first household member showed symptoms was associated with a 79 percent lower transmission rate.

Observational studies such as these, however, have severe limitations. Mask-wearers are more likely to be conscientious and take other precautions whose effects can be confounded with those of masks. (One finding of the Beijing Covid-19 study suggests that something like this occurred: frequent surface disinfection supposedly reduced household transmission by 77 percent—almost as much as masks—though surface transmission of Covid-19 is now known to be insubstantial.) Observational studies also suffer from “recall bias”: inconsistent mask-wearers may report themselves as not wearing masks if they fell sick, or as wearing masks if they did not.

Randomized controlled trials (RCTs) on mask use, generally more reliable than observational studies, though not infallible, typically show that cloth and surgical masks offer little protection. A few RCTs suggest that perfect adherence to an exacting mask protocol may guard against influenza, but meta-analyses find little on the whole to suggest that masks offer meaningful protection. WHO guidelines from 2019 on influenza say that despite “mechanistic plausibility for the potential effectiveness” of masks, studies showed a benefit too small to be established with any certainty. Another literature review by researchers from the University of Hong Kong agrees. Its best estimate for the protective effect of surgical masks against influenza, based on ten RCTs published through 2018, was just 22 percent, and it could not rule out zero effect.

More recent studies back up these results. A November 2020 report by Cochrane, a nonprofit organization that produces authoritative reviews of medical research, notes: “There is low certainty evidence from nine trials (3507 participants) that wearing a mask may make little or no difference to the outcome of influenza-like illness . . . compared to not wearing a mask . . . . There is moderate certainty evidence that wearing a mask probably makes little or no difference to the outcome of laboratory-confirmed influenza compared to not wearing a mask.” The only complete RCT for Covid-19 was also disappointing: a spring 2020 trial in Denmark showed a protective effect statistically indistinguishable from zero.

Public-health authorities justified mask mandates not for personal protection, however, but as a means of stopping wearers from spreading Covid-19—a pressing matter once case studies showed that spreaders may not show symptoms. Since exhaled droplets are larger than inhaled droplets shrunk by evaporation, masks might be better at source control than at personal protection. Indeed, some trials support a source-control effect at least for brief direct exposures: one study shows that surgical masks cut bacterial spread from coughing cystic fibrosis patients by 94 percent.

Other studies have been less promising. First, many studies (summarized by a researcher for the National Institutes of Health) find that exhaled droplets can be small enough to become airborne and slip through mask pores: “most particles in exhaled breath are smaller than 4 [microns], with a median between 0.7 and 1.0 [microns].” These small particles are indeed infectious: one study on influenza viruses in exhaled droplets found that “the fine particles [smaller than 5 microns] contained 8.8 times more viral copies than did the coarse ones.” Another study found that most infectious particles were smaller than one micron.

Studies in hospitals, likewise, have repeatedly shown that surgical masks worn by surgeons do not reduce bacterial wound infections, even though masks might be expected to work best against short-distance transmission of bacteria rather than smaller viruses. One study even found that when surgeons in one British hospital got rid of face masks, the rate of wound infections fell by half. Surgical masks, another study showed, did not reduce bacterial contamination of surfaces in an operating theater. There are several case studies, furthermore, of Covid-19 outbreaks in confined spaces despite good mask adherence, such as one outbreak in a Marine Corps barracks whose residents wore cloth masks almost constantly.

The most powerful evidence, though, comes from comparing Covid-19 in different regions: do areas with mask mandates or widespread mask use see fewer cases? Last spring, the answer seemed to be yes. Many Asian nations, prominently Taiwan, South Korea, Hong Kong, and Japan, had adopted public mask use during SARS in 2003 and readopted it for Covid-19. They saw outbreaks in February that quickly subsided. Japan’s success seemed particularly hard to attribute to anything but masks: the nation saw only weak government interventions, principally a brief, likely ineffective school closure. Some Western nations followed suit, with apparent success. The Czech Republic was the first, thanks to Petr Ludwig, an influential science popularizer, who posted a YouTube video on March 14 urging Czechs to use improvised cloth masks. Poland quickly followed; a social media campaign in late March exhorted Poles to kryj ryj, or “cover your snout.” Both nations escaped the spring wave nearly unscathed.

The subsequent record, though, is less encouraging. Ian Miller, a writer for the lockdown-skeptical website Rational Ground, has compiled dozens of examples of massive outbreaks in U.S. states and other nations with strict mask mandates. His most convincing examples are natural experiments: comparisons of similar jurisdictions, only one of which mandated masks. For instance, North Dakota and South Dakota both largely escaped Covid-19 in spring 2020 but saw large waves of infections in the fall. North Dakota introduced a mask mandate and business closures on November 17, but South Dakota never did. Nevertheless, both states saw nearly identical epidemic curves, with a peak in new cases in mid-November and quick falls thereafter. North and South Dakota now have low rates of new cases and nearly identical total cases and deaths: 133 cases and 1.96 deaths per 1,000 residents in North Dakota, versus 130 and 2.17 in South Dakota, according to data collated by the New York Times as of March 18. Similarly, in the summer and fall, Alabama, which imposed a mask mandate on July 16, had an epidemic curve nearly indistinguishable from its neighbors Georgia and Mississippi, which never did.

The fall coronavirus wave hit several of the mask success stories especially hard. Poland and the Czech Republic had uncontrolled outbreaks (the Czech wave was Europe’s worst for a time); Japan and South Korea, meanwhile, saw more muted surges.

So what accounts for some nations’ initial good fortune? The best explanation may just be luck. Much early Covid-19 spread was caused by a few superspreader events; nations without a superspreader event may simply have managed to avoid an outbreak until better weather reduced transmission. Denmark and Norway, for instance, had tiny first waves, despite loose lockdowns and little mask use.

Another possibility: SARS-CoV-2 is related to several widespread mild coronaviruses, and immune responses to one virus frequently confer some resistance to others. In some studies, more than half of people in Western nations show preexisting immune responses to SARS-CoV-2. Viruses that confer cross-immunity may circulate more strongly in Asia than elsewhere, explaining why not just East Asia but also Indochina—including some nations, such as Cambodia, with weak initial responses—had tiny outbreaks. Similar patterns, with lower death rates near the origin than elsewhere, have occurred in flu pandemics from Asia as well as the misnamed Spanish flu, which started in the United States but had higher mortality in Europe.

The remaining evidence that masks work for source control amounts largely to case studies that could simply be coincidence (Covid-19 patients vary greatly in their infectiousness) and questionable regional comparisons with poor controls for one important fact: the farther an epidemic has progressed, the slower it grows in percent per day, as each newly infected person encounters fewer susceptible others. One widely cited study finds that states that enacted mask mandates in the spring saw lower growth rates in the following weeks—but those states were mostly early hotspots already closer to herd immunity, where one would naturally expect to see decelerating transmission. The same problem affects an influential CDC study finding that, during the summer, Kansas counties with mask mandates had slower growth of Covid-19 cases than those without: mask-mandate counties had seen a large spike in cases just before the mandate went into effect and had consistently higher absolute infection rates.

But even if masks probably don’t help, one might reasonably ask, what’s the harm in requiring them if they might help a bit?

These harms, it turns out, are severely understudied. The Cochrane review of mask RCTs notes, “Harms were rarely measured and poorly reported”—but some are uncontroversial. The Washington Post reported in May, for instance, that masks could cause persistent face rashes. They can also cause headaches: a small Japanese study testing whether surgical masks could stop the spread of colds among health care workers found that mask wearers were almost four times as likely as controls to report headaches. In a self-selected survey of German schoolchildren, more than half of the participants reported headaches.

Unhygienically used masks can also trap moisture and thus promote bacterial and yeast growth. It’s natural to suspect that this promotes other infections: an article in the New York Post, for example, reported that many dentists had observed an increase in problems such as cavities and gingivitis that they attributed to mask wearing. Finally, there is the unknown health effect of fiber inhalation (not to mention environmental plastic pollution). One paper noted that most surgical masks have loose fibers.

Potential harms to children deserve special mention. Two Italian professors of plastic surgery, for instance, have hypothesized that the pressure of elastic ear straps may give children permanently protruding ears. Some child development researchers also worry that widespread mask-wearing may hamper children’s linguistic and emotional development.

There may even be ways by which masks might worsen Covid-19 itself. The basic reason is simple: germs caught by a mask do not simply disappear. The evidence for these is spotty or speculative but concerning enough to merit attention. In any case, the evidence justifying mask mandates is often equally speculative.

First, as Kevin McKernan, a microbiologist who worked on the Human Genome Project, has pointed out, forcing liquid through a porous membrane is a standard “nebulization” technique for producing emulsions of small droplets. McKernan hypothesizes that masks may likewise split large respiratory droplets into more dangerous airborne particles rather than trapping them. This hypothesis has some weak empirical support: multiple studies have found that some low-quality cloth masks increase emission of micron-scale particles, through nebulization or through detaching of particles from cloth threads. These studies, furthermore, last only for minutes, and the risk of pathogen dispersal from masks likely increases with prolonged use. One study of masks in surgery found that exhaled bacteria accumulate on the exterior of masks worn longer than about two hours.

Second, when droplets trapped in a mask dry out—according to speculation by the German doctor Zacharias Fögen—viruses might be re-inhaled as individual particles, penetrating deep into the lungs as well as infecting the olfactory nerve and, thence, possibly the brain, worsening existing Covid-19 cases. Fögen has analyzed Kansas epidemiological data to conclude that counties with mask mandates have a Covid-19 case fatality rate, or ratio of deaths to confirmed infections, 85 percent higher than those without. Fögen’s study is far from dispositive. My own attempt to apply his method to two other states has yielded different conclusions, and other evidence suggests a countervailing hypothesis that even imperfect masks may reduce Covid-19 severity by reducing the number of initially inhaled viruses. But Fögen’s argument is at least plausible enough to justify more attention and attempts at replication.

It would be an overstatement to say that cloth and surgical masks are unambiguously ineffective or harmful. But neither is there a firm case that they provide any meaningful benefit. Limited mask mandates may be justified in circumstances with unavoidable face-to-face contact within the range of droplet spread, such as public transport, and private businesses should be free to require masks if they like. Citizens at high risk should be free to wear effective N95 masks for their own protection, and federal regulators should clear away barriers to domestic production.

But mandates of cloth and surgical masks impose major inconveniences and potentially serious health risks on citizens, for no clear benefit either to themselves or to others. Leaders who pride themselves on following the science should consider ending them and letting citizens protect their health as they see fit.

Photo by Montinique Monroe/Getty Images


City Journal is a publication of the Manhattan Institute for Policy Research (MI), a leading free-market think tank. Are you interested in supporting the magazine? As a 501(c)(3) nonprofit, donations in support of MI and City Journal are fully tax-deductible as provided by law (EIN #13-2912529).

Further Reading

Up Next