Statistical learning

Statistical learning is the acquisition of patterns from the frequencies and co-occurrences of events or features. It can occur without deliberate instruction or conscious awareness.

Transitional probability: The conditional probability that one event follows another in a sequence. Learners can use low transition probabilities to detect boundaries between recurring sequences.
Probability: A mathematical measure of how likely an event is to occur. Statistical learning depends on sensitivity to differences in event likelihood.
Word segmentation: The identification of word boundaries in continuous speech or text. Infants can use transitional probabilities between syllables to infer word boundaries.
Jenny Saffran: A developmental psychologist known for research on statistical learning in infants and language acquisition. Her experiments showed that infants use transitional probabilities to segment speech.
Statistical learning debate: Scientific discussion about the mechanisms, scope, and interpretation of statistical-learning findings. Researchers disagree about whether one general mechanism explains results across tasks.
Distributional learning: Learning categories or structure from the frequency distribution of sensory or linguistic input. Frequency peaks can reveal category boundaries, such as distinctions between speech sounds.
Conditional probability: The probability of an event given that another event has occurred. Co-occurrence patterns are often expressed as conditional probabilities.
Phonetic category learning: Learning to group speech sounds into categories that distinguish meaningful contrasts. Listeners can infer speech-sound categories from the distribution of acoustic examples.
Richard Aslin: A developmental psychologist whose research examines infant perception, cognition, and learning. His work helped demonstrate statistical speech segmentation in infancy.
Domain-general learning: Learning mechanisms that operate across different kinds of information or subject matter. A central question is whether statistical learning uses shared mechanisms across modalities.

Linked from 12 pages