When babies listen, what they’re doing is taking statistics on the language that they hear.

Patricia K. Kuhl, a leading developmental psychologist and speech-language researcher at the University of Washington and co-director of the Institute for Learning & Brain Sciences (I-LABS), has often explained infant language learning in terms of “statistical learning.” In public lectures and interviews about early phonetic development, she describes how infants, long before they can speak, track patterns in the speech stream—such as which sounds and syllables tend to occur together and how frequently they appear. The remark is typically used to translate laboratory findings (e.g., infants’ sensitivity to distributional regularities in speech) into an accessible metaphor: listening as data-gathering that helps babies build the sound categories of their native language.

The quote frames infant listening as an active, computational process rather than passive exposure. “Taking statistics” suggests that babies implicitly tally frequencies and probabilities in what they hear—learning which sound distinctions matter in their environment and which can be ignored. This helps explain why early experience is so powerful: the language surrounding a child supplies the dataset from which the brain infers phonetic categories and word boundaries. The line also underscores a broader scientific point: human language acquisition can be studied as learning from input, where social interaction supplies rich, structured speech that makes those statistical patterns learnable and meaningful.