Estimates of the threshold for SARS-CoV-2 range from 10% to 70% or even more5,6. But models that calculate numbers at the lower end of that range rely on assumptions about how people interact in social networks that might not hold true, Scarpino says. Low-end estimates imagine that people with many contacts will get infected first, and that because they have a large number of contacts, they will spread the virus to more people. As these ‘superspreaders’ gain immunity to the virus, the transmission chains among those who are still susceptible are greatly reduced. And “as a result of that, you very quickly get to the herd-immunity threshold”, Scarpino says. But if it turns out that anybody could become a superspreader, then “those assumptions that people are relying on to get the estimates down to around 20% or 30% are just not accurate”, Scarpino explains. The result is that the herd-immunity threshold will be closer to 60–70%, which is what most models show (see, for example, ref. 6).

Looking at known superspreader events in prisons and on cruise ships, it seems clear that COVID-19 spreads widely initially, before slowing down in a captive, unvaccinated population, Andersen says. At San Quentin State Prison in California, more than 60% of the population was ultimately infected before the outbreak was halted, so it wasn’t as if it magically stopped after 30% of people got the virus, Andersen says. “There’s no mysterious dark matter that protects people,” he says.

And although scientists can estimate herd-immunity thresholds, they won’t know the actual numbers in real time, says Caitlin Rivers, an epidemiologist at the Johns Hopkins Center for Health Security in Baltimore. Instead, herd immunity is something that can be observed with certainty only by analysing the data in retrospect, maybe as long as ten years afterwards, she says.