One thing that we’ve learned about covid-19 is that we still have a lot to learn. Any time we thought we knew everything we needed to about this virus, reality soon disabused us of the notion. Area after area has thought it was keeping the novel coronavirus at bay — that rich Western countries wouldn’t get it, or rural areas; that hard-hit places had herd immunity, or that Sweden had shown you could control the virus without lockdown . . . until, one by one, these theories have been falsified. Most recently, President Trump demonstrated that testing is not, by itself, a covid-19 control strategy, and resurgent outbreaks in Madrid and New York’s Orthodox Jewish community have made it look unlikely that any group has herd immunity yet.

Perhaps some undulation is inevitable: as caseloads rise, people get cautious, and as they fall, people take more risks, causing caseloads to rise again. But another possibility is that the undulation might not just be about policy, or even personal attitudes, but something as simple, and uncontrollable, as the weather.

Obviously, that echoes Trump’s famous speculation that the virus would “just disappear” over the summer. Just as obviously, it didn’t. But Trump’s theory may not have been wrong so much as insufficiently refined. After all, it was based on a real observation: back in March, colder places were having more outbreaks than warmer ones. In the United States, however, that relationship broke down this summer.

But don’t focus on the temperature; focus on how humans react to it. Because one thing we have learned in seven months is that most spread seems to take place over long exposures in enclosed spaces. So the right question may not be “how cold is it?” but “is this the kind of weather that drives gatherings indoors?”

The answer to that question probably varies from place to place, which seems to explain some puzzling patterns. Why did the European outbreak die down faster than America’s? Well, maybe it was all policy, but maybe it helped that Europeans rarely have air conditioning in their homes, so when it’s hot they tend to do their socializing outside as much as possible. The Sun Belt in the United States needs air conditioning to make its sizzling summers bearable — and perhaps unsurprisingly, had a lot of outbreaks this summer.

Now that Europe is getting cold and the Sun Belt is cooling off, that pattern seems to have once again reversed. As economist Michael Strain recently pointed out, “Cooler weather three weeks ago is strongly correlated with more covid cases this week.” Scott Gottlieb, the former head of the Food and Drug Administration, noted that 45 degrees Fahrenheit, “the temperature range where outdoor dining is no longer sought by consumers” seems to be a particularly important inflection point.

As yet unpublished work by researchers Sterling Haringa, Sean Trende and Christina Ramirez bolsters this case, with regression analysis suggesting that two factors account for much of the variation in cases between U.S. states: population density and temperature. But with an important caveat: When data was analyzed in March, places with lower temperatures generally had worse covid-19 outbreaks; in June, it was the places with higher temperatures that had the bigger problems, exactly as we’d expect if “indoor gatherings” are the main driver.

Trende told me that when Ohio reopened, he and his wife went to dinner for their anniversary — right about the time when he conducted the analysis for June. “I haven’t been back out since,” he added ruefully.

In other words, felicitous weather may have been impersonating herd immunity. As the weather changes, groups that thought they had herd immunity will discover they don’t, and policies that had looked highly successful will start to fail. It’s not that those policies are useless: Testing, tracing, quarantine, mask-wearing, hand-washing, ventilation and social distancing are all important. But the ever-changing weather can be a powerful tailwind behind those policies — or a headwind slowing them down.

In places where things are likely to get worse this winter, “policymakers need to be smart about when and where they shut down,” Trende told me. “Because they are going to shut down.” Especially, he says, they need fresh thinking about bars and restaurants. Instead of enduring repeated cycles of reopening, seeding new epidemics, and then shutting down again, we should probably just offer relief to help them stay solvent until indoor dining can safely resume. Meanwhile, other risky gatherings, from church services to birthday parties, will either have to downsize or defy the weather and stay outside, with the help of all the fire pits and patio heaters and snowsuits — or fan misters and pools — that we can scrounge up.

We will also, of course, need to nurture some humility, along with greater hardiness, in order to confront this virus successfully. No matter how clear a pattern may seem or how sure you are that you’re doing the right thing while others are getting it all wrong, before you pass final judgment, you might want to wait six months and see how covid-19 decides to surprise us.

Read more: