[Verse 1] Speech flows like a river with no breaks between the stones No silence marks the borders where each meaning finds its home Eight-month babies crack the code without a single pause to guide They're counting up probabilities, statistics are their guide When syllables stick together tight, the chances climb up high But cross a word boundary and watch those numbers die [Chorus] Transitional probabilities, the dips reveal the seams High inside, low outside, parsing language streams Hip-po-cam-pus learning what comes next in every chain Statistical segmentation, how the infant brain Carves the flow to find the words hidden in the sound Transitional probabilities, where word boundaries are found [Verse 2] Saffran played her nonsense streams for just two minutes straight No acoustic cues available, pure statistics sealed their fate The babies heard "tupiro" and "golabu" emerge From continuous gibberish, their pattern-seeking urge Distinguished real word units from the part-words at the edge Predictive links decoded at eight months on the ledge [Chorus] Transitional probabilities, the dips reveal the seams High inside, low outside, parsing language streams Hip-po-cam-pus learning what comes next in every chain Statistical segmentation, how the infant brain Carves the flow to find the words hidden in the sound Transitional probabilities, where word boundaries are found [Bridge] Not just for syllables alone, this machinery runs wide Visual sequences and tones, motor patterns on all sides Domain-general computation, not a language-only tool The same stats that parse our speech make other patterns cool MTL separates the items while CA fields connect What tends to follow what—the chains that we expect [Verse 3] Fifteen years have changed the map of where this process lives Cortical and frontal zones, but hippocampus gives The rapid online learning power, temporal structure's key Episodic memory systems set our word boundaries free Striatal habits catch the patterns practiced day by day While hippocampal prediction points the segmentation way [Chorus] Transitional probabilities, the dips reveal the seams High inside, low outside, parsing language streams Hip-po-cam-pus learning what comes next in every chain Statistical segmentation, how the infant brain Carves the flow to find the words hidden in the sound Transitional probabilities, where word boundaries are found [Outro] From continuous streams to meaning, statistics light the path Predictive associations, sequence-learning's math
← Module 2: Phonological learning and perceptual narrowing | Module 4: The hippocampus and fast mapping. The core of the course →