An enormous randomized trial of communities in Bangladesh seems to provide the clearest evidence yet that regular mask-wearing can impede the spread of the covid-19 pandemic. The study found that villages where masks were highly promoted and became more popular experienced noticeably lower rates of covid-like symptoms and confirmed past infections than villages where mask-wearing remained low. These improvements were even more pronounced for villages given free surgical masks over cloth masks.

Plenty of data has emerged over the last year and a half to support the use of masks during the covid-19 pandemic, both in the real world and in the lab. But it’s less clear exactly how much of a benefit these masks can provide wearers (and their communities), and there are at least some studies that have been inconclusive in showing a noticeable benefit.

One problem in interpreting all this information is that we’ve largely relied on observational studies, which can only ever show a correlation between any two things, not establish a cause-and-effect relationship. There might be other factors that both explain why one city has a higher rate of mask-wearing and a lower rate of diagnosed cases than another city, for instance, rather than the former helping cause the latter.

Last late year, however, dozens of scientists teamed up with public health advocacy organizations and the Bangladesh government to conduct a massive randomized trial of masks—often seen as the gold standard of evidence. And on Wednesday, they released the results of their research in a working paper through the research nonprofit Innovations for Poverty Action.

The study involved 600 villages in a single region of the country with over 350,000 adult residents combined. Similarly matched villages were randomly assigned to two conditions (a pair of villages with similar population density, for instance, would go to one condition or the other). In one condition, the researchers and their partners promoted the use of masks through various incentives between November 2020 and January 2021. These incentives included free masks, endorsements by local leaders, and sometimes financial prizes for villages that achieved widespread mask usage. In two-thirds of the intervention villages, the free masks given were surgical, while one-third were given free cloth masks. In the second condition, the researchers simply observed the villages and did nothing to encourage masks during that time.

Residents in the villages where masks were encouraged did start wearing them more, though no individual nudge or incentive seemed to do better than the others. By the end, about 42% of residents in these villages wore masks regularly, compared to 13% of those in the control group. And in these communities, the odds of people reporting symptoms that may have been covid or testing positive for antibodies to the virus declined.

Overall, the average proportion of people who reported symptoms in the weeks following the mask promotions went down by 11% in these villages compared to the control group, and the average number of people having antibodies went down by over 9%. These differences were larger for surgical mask-wearing villages (12% vs 5% for reducing symptoms) and for residents over 60 (35% for reducing infections for older residents in surgical mask-wearing villages).

Some of this effect might not have come directly from the ability of masks to block transmission of the virus. Those who used masks, the study found, were also more likely to practice social distancing. That’s a relevant finding, the authors note, since some people who have argued against mask mandates do so by claiming that masks will only make people act more carelessly. This study suggests that the opposite is true—that masks make us more, not less, conscientious of others.

The findings are not in a peer-reviewed journal as of yet, an important step for validating any research. And they do carry some limitations, as any study does. The study began and ended before the emergence of the Delta variant, for instance, a much more transmissible version of the coronavirus that’s become widespread throughout the world (at the time, the Alpha variant was most prevalent).

Study author Jason Abaluck, a health and behavioral economist at Yale University, told Gizmodo in an email that his team has submitted the paper for publication in the journal Science. On Twitter, Abaluck has addressed other potential caveats of the study. Some have pointed out, for instance, that the authors only found a protective effect from masks for people under the age of 50 in experiencing covid-like symptoms, not in having antibodies (for older people, a reduction in symptoms and antibodies was seen across the board in mask-wearing villages). But Abaluck argues that this may simply be due to the fact that only 40% of people with symptoms opted to get tested, so any estimates from this group may be less precise. And even if masks somehow had no direct effect for people under 50, they may yet reduce the spread of the virus from younger people to older, so masking would still be a net positive on a population level.

The authors also say that masks could conceivably have a greater effect in slowing down the spread of the current pandemic on a population level than they did when the study concluded, given the higher transmissibility of Delta per case. And because they noticed a significant effect after only a modest increase in mask use, the benefits could be even greater with widespread masking.

“Our results should not be taken to imply that masks can prevent only 10% of covid-19 cases, let alone 10% of covid-19 mortality,” they wrote. “Our intervention induced 29 more people out of every 100 to wear masks, with 42% of people wearing masks in total. The total impact with near universal masking—perhaps achievable with alternative strategies or stricter enforcement—may be several times larger than our 10% estimate.”

If that assumption turns out to be true, it provides more support for models showing that universal mask coverage in places like the U.S. can still significantly blunt the impact of the pandemic. A recent forecast from researchers at the University of Washington, for instance, estimated that universal mask coverage could prevent up to 50,000 deaths by December 1 of this year.

Perhaps most importantly on the individual level, the study also suggests that cloth masks should be phased out as a recommended choice of mask and that surgical masks should be the default moving forward, the authors say.

“While cloth masks clearly reduce symptoms, we cannot reject that they have zero or only a small impact on symptomatic SARS-CoV-2 infections,” they wrote.