Statistical learning

Question 1

Regularities

Answer 1

-the world has regularities
-features (object/phonological
/categorical/sequential) tend to co-occur across space and time

Question 2

Statistical learning

Answer 2

• extraction of regularities in how features tend to co-occur over space and time
• with or without conscious awareness
• multiple time scales

Question 3

Segmenting auditory streams

Answer 3

• auditory input is a continuous stream
• infants have to learn where words start and end
• statistics of transition
• high vs low transition probability
• low transition is where word breaks are
• unconscious

Question 4

Saffran- segmenting audio streams

Answer 4

• listen to continuous stream
• some features co-occur mroe frequently
• show word and non-word
• babies pay more attention to novel (non) word

Question 5

Segmenting other temporal sequences

Answer 5

• visual shapes
• events in the world
  
  • people segment sequences into events consistent with transition probabilities
  • people tend to parse events similarly

Question 6

The brain learns temporal sequences

Answer 6

• monkeys view continuous stream of fractals in the same order
• then view in random order
• neurons that preferred one fractal also preferred ones that occurred close to it in time during learning
• temporal regularities are being coded into the brain by neural firing

Question 7

Brain learns category feature co-occurance

Answer 7

• computational model of cortical representation
• storage of probabilistic relation between concepts and their properties
• might not co-occur together but relate to similar features that co-occur
• similar perceptual categorization share similar representations in the brain (membership in category is graded)

Question 8

hippocampus and statistical learning

Answer 8

• damage to hippocampus leads to impaired semantic memory
• fMRI data:
  
  • hippocampus engaged by temporal regulation
  • responds to violations of temporal regulation