Sommers makes a good case that the one in five finding is inflated, and that the administration and the media have been careless in disseminating it as fact. It comes from a 2007 study conducted by the Justice Department’s National Institute of Justice, a web-based survey of two large public universities, one in the Midwest, and one in the South. Female undergraduates aged 18 to 25 took part in the anonymous random sample, and each received a $10 certificate to Amazon for participating.

Web surveys are automatically suspect, and the response rate was less than half, 42 percent, “which is miserable,” says Sommers. The researchers themselves cautioned against over-generalizing from the 5,446 respondents to the entire country’s college population, and Sommers cites the danger of selection bias, where people who feel most strongly respond.

“I always like to know what did they ask,” she says, reserving her strongest criticism for the expansive definition of sexual assault that is now part of the lexicon, particularly factoring in the role of alcohol. She notes that in some 70 percent of assaults categorized in the poll, alcohol was involved, and the woman was incapacitated.