In 2003, I discovered that when nine-month-old infants are exposed to a new natural language for twelve sessions over a month's time, that interaction with another human being is essential for infants to learn. A beautiful DVD presentation of the exact same material to another group of babies produced no learning whatsoever, nor did an audio-only presentation.

The graduate students and post-doctoral fellows in my laboratory took bets after watching infants in the DVD condition – who stared at the monitor and even crawled up to it. Some touched it. They bet that the DVD would win over “live” because the DVD was so beautifully clear. There were fewer distractions and babies seemed riveted by the action on the screen.

But the results of brain and behavioral tests showed that only infants who experienced native Mandarin speakers “live” showed evidence of having learned. Those who watched the same native speakers on DVD, or heard them through loudspeakers, performed no differently on tests of Mandarin sounds than infants in a control group who heard only English from American native speakers in otherwise identical sessions. Learning was phenomenal: “live” infants matched Taiwanese infants who had ten months experience listening to the language.

The result was a huge surprise. It challenged existing theory, has affected educational practices, and jolted the baby DVD industry.

What led us to do the experiment? What were the theories and predictions at the time? And how did the finding lead me to a new hypothesis, the “social gating” hypothesis?

Two highly polarized theories shape the field of language acquisition, each elegantly argued by giants in the field. MIT professor Noam Chomsky favors a nativist position, positing a Language Acquisition Device (LAD) that triggers parameters of innate universal phonology and grammar through language input. Brown University professor Peter Eimas fleshed out this view, arguing that innate “phonetic feature detectors,” encompassing the sounds of all languages, atrophied when the language infants heard did not include a particular phonetic feature. Al Liberman of Haskins’ Laboratories at Yale argued that “speech is special” – that brain systems were “modularized” such that other cognitive processes could not influence them.

The alternative view, expressed by Harvard professor B. F. Skinner, held that language required only operant conditioning and reinforcement to reward and eventually shape the behavior of children as they uttered words. Infants were, in effect, “blank slates,” waiting for parental tutoring.