So what we're seeing here is changing our models of what the critical period is about. We're arguing from a mathematical standpoint that the learning of language material may slow down when our distributions stabilize. It's raising lots of questions about bilingual people. Bilinguals must keep two sets of statistics in mind at once and flip between them, one after the other, depending on who they're speaking to. So we asked ourselves, can the babies take statistics on a brand new language? And we tested this by exposing American babies who'd never heard a second language to Mandarin for the first time during the critical period. We knew that, when monolinguals were tested in Taipei and Seattle on the Mandarin sounds, they showed the same pattern. Six-eight months, they're totally equivalent. Two months later, something incredible happens. But the Taiwanese babies are getting better, not the American babies. What we did was expose American babies during this period to Mandarin. It was like having Mandarin relatives come and visit for a month and move into your house and talk to the babies for 12 sessions. Here's what it looked like in the laboratory.