Then, around 2008, multiple research groups found something surprising in human blood samples: antibodies that could defend against two HA subtypes. “They were H1 antibodies that recognized H5 bird flu, and everyone said, ‘Wait, that’s not supposed to happen,’” James Crowe, the director of the Vanderbilt Vaccine Center, recalls. It turned out that those antibodies were bound to the stalk of the HA proteins, as opposed to the head — and the stalk appears to mutate less than the head between HA-NA strains. “That started the whole revolution of the universal flu vaccine,” Crowe says. “We thought the surface of the flu virus was infinitely variable. But actually, there are some places that rarely change drastically.” Multiple groups now have universal vaccines in various stages of clinical trials. Some of them, including one that the National Institutes of Health is testing for safety in people this flu season, produce more stalk antibodies than traditional vaccines.

Even if a vaccine is broadly effective against different flu strains, it’s unclear how long that protection might last. (A shot that lasts just two or three years would have a big public-health benefit.) There is evidence that, in fact, the immune system can generate a lifelong defense against influenza — but there’s a catch. People appear to be “imprinted” somehow with the HA proteins they encounter in childhood, says Aubree Gordon, an associate professor of epidemiology at the University of Michigan School of Public Health. People respond to flu viruses and vaccines differently based on their past exposure. “It’s not only what you’ve recently been exposed to which is important,” she says. “But we think your very first exposures are incredibly important for determining your response for the rest of your life.”

People imprinted with one HA subtype have a response ready for similar viruses. But when they are infected with one that’s less alike, they may still make antibodies partly geared toward the earlier virus, rendering them less beneficial.