Blue bar english speaking infants infants who were

Blue bar: english-speaking infants (infants who were exposed only to a video
tape of the interaction session or only to the audio)
■ They did not do very well
○ Red bar: control group; chinese infants who were exposed to chinese from birth
■ Even though they just heard or watched mandarin, they obviously still did
really well
● First two red bars have almost the same succession rate
How do infants learn words?
● Speech segmentation problem
○ When you hear speech, it is actually a whole running steam of sounds and
typically people don’t make big pauses between words
○ It is fluid speech stream
○ Sound wave: it’s hard to tell where one word begins and one words ends
Statistical learning
● Infants keep track of frequent patterns in speech and use these to help them distinguish
word boundaries
○ Probability -if syllables co-occur with high probability, they are likely words
(hap-py ba-by)

■ By extracting this kind of information, infants are able to discover which
sounds go together to make words and which sounds don’t really go
together
■ They are just at that word boundary and happen to co-occur
○ In English, the stress syllable is either a word on its own (cat) or the start of a
multi-syllable words
○ They can also phonotactic cues
■ So certain sounds rarely go together in a single word
■ S and t go together quite often whereas s and d don’t really go together
that often in english
Can infants track regularity in speech?
What do these studies show?