On February 18, 2026, the JCU Department of Psychological and Social Sciences was delighted to host Professor Jenny Saffran, from the University of Wisconsin–Madison. In her talk, “How Babies Learn Language: The Role of Statistical Learning,” Professor Saffran invited the audience to reconsider what it really means to “know” a language. As adults, we can effortlessly recognize speech sounds, understand the meaning of words, identify where words begin and end in continuous speech, and intuitively follow grammatical patterns. Yet none of this knowledge is explicitly taught. So where does it come from?
From Womb to World
Drawing on decades of research, Professor Saffran compellingly demonstrated that this knowledge is available much earlier than we might imagine, long before babies can even produce their first word. Professor Saffran presented research suggesting that even fetuses can detect rhythmic patterns in speech, and that newborns can recognize familiar passages heard repeatedly during pregnancy, demonstrating memory for language experience before birth. After birth, infants quickly show preferences for the language spoken by their caregivers and for infant-directed speech, the expressive, melodic style that adults naturally use when interacting with babies.
By just four months of age, infants recognize their own names. By six months, they already show evidence of understanding familiar words by looking at their corresponding image. For example, looking at an apple rather than a different object when hearing phrases such as “Where’s the apple?”.
The Role of Statistical Learning
One of the central themes of the lecture was the concept of statistical learning. Professor Saffran explained that when adults speak, words do not come with clear pauses between them. To a baby, speech can sound like one continuous stream of sound. So, how do babies figure out where one word ends and the next begins?
The answer lies in patterns. Some sound combinations occur together very frequently, while others do not. In English, for example, the syllables in the word “pretty” reliably go together, with “pre” almost always followed by “tty.” But many different words could come after “pretty”: baby, dog, flower, and so on. Babies appear to be very sensitive to these patterns. They track how frequently certain sounds occur together, use this information to identify likely word boundaries, and when certain sounds appear reliably together, babies treat them as belonging to the same word.
Through laboratory experiments using “artificial languages,” Professor Saffran demonstrated that even eight-month-old infants can detect these patterns after only brief exposure.
Crucially, babies do not simply notice patterns — they use them to learn meaning. Professor Saffran presented evidence that when sounds form a word-like unit, babies learn them as object names more easily than when the sounds are just mixed-up pieces of different words. In other words, statistical learning is not just something that happens in laboratory experiments; it is a powerful tool that helps babies learn their language in the real world.
A Research with a Broader Impact
The lecture also highlighted the broader impact of this research. Methods originally developed to study infants, such as measuring where and how quickly children look in response to spoken words, are now being applied to clinically underserved populations, including children with severe motor impairments who cannot speak or point. These approaches offer new possibilities for assessing language comprehension and improving early intervention.
Throughout the talk, Professor Saffran combined scientific rigor with clarity and humor, engaging the audience. Her visit provided a vivid example of how fundamental research in cognitive development can help us understand the extraordinary capacities of the human brain while also providing practical pathways for supporting children’s learning.
