3. Learning Sounds 2

Continuous Stream of Speech

  1. use familiar words to break up stream (segmentation)

  2. track transitional probabilities

Familiar Words and Segmentation

evidence

  • head turn preference procedure (familiarize words to kids)

  • head turn preference procedure (using naturally familiar words)

development of ability

  • 6 months: babies can use familiar names (the words they hear in isolation) to learn/segment directly adjacent words

  • 7.5 months: babies are better at this and can learn words they hear in isolation much faster and segment them

Phonotactic Constraints and Stress Patterns

stress pattern - experiment design

  • paradigm: head turn preference procedure

  • age: ~7.5 months

  • stimuli example: nonce CVCV words, two conditions, words should use english phonotactics and come from limited pool of consonants/vowels

    • SW stress pattern words

    • WS stress pattern words

  • test phase: randomly play words from both conditions (within-subjects) for the babies, record looking time 

  • predictions: since it's younger infants, expect looking time for SW words to be statistically significantly longer than for WS words


phonotactic constraints - experiment design

  • paradigm: head turn preference procedure

  • age: ~7.5-9 months, too much older and words with non-English phonotactic constraints may be perceived as novel and exciting and influence results 

  • stimuli example: two sets of words (conditions), both should use the same consonants/vowels (shouldn't be any unique to one set), but follow different phonotactics

    • words that follow English phonotactics

    • words that don't follow English phonotactics

  • test phase: randomly play streams of words from both conditions (within-subjects) for the babies, record looking time 

  • predictions: since it's younger infants, if they prefer words with English phonotactics, expect a longer looking time for the stream of words with English phonotactics

Transitional Probabilities

what are TPs

  • a statistical cue

  • TPs are higher between syllables that are part of the same word

  • children keep track of how often syllables follow each other -- aid in segmentation

evidence

  • head turn preference procedure

  • 8 months: can represent the order of syllables in a word -> can use TPs to understand when one word ends and another begins

unique to humans? is stat learning unique to humans?

  • no it is not

  • tamarins (monkeys?) tested also distinguished words vs partwords

  • rats tested also distinguished words vs non-words