Bernard et al. ask their survey subjects how many people they know in certain specific categories—people named Nicole, people with pilot's licenses, people who gave birth last year, etc. These are groups whose size is known. Using a subject's answers—three people named Nicole, no one with a pilot's license, and so forth—they can extrapolate how big the subject's circle is. According to Bernard, networks vary from "almost zero to a couple thousand," but the average, consistently, is 290.

That's probably a low estimate because the survey is conducted without "aided recall." In other surveys, subjects are prodded with cues to aid their memory. They are asked, say, "Is there anyone in your immediate family named Nicole? Your office?" etc. They remember more people, enough to boost their network above 400. But because these aided studies are not as numerous or consistent as the non-aided surveys, Bernard and his team stuck with the reliable, experience-tested figure of 290.