Language & memory Flashcards
Morpho-syntax
The rules of syntax are stored in long-term memory
- Contains word combinations as well as rules that you apply
- Contains the exceptions
Stream of speech and WM
Listening/reading involves a stream of language input, the WM is needed to buffer this stream.
- Same goes for speech production, the planning of production
Segmentation with transition probability
Statistical learning is used to find speech segmentation in a flow of syllables. Transition probabilities are used;
Transition P is 1 from syllable to syllable within 1 word
Lower when it transitions to another (either .. or .. * No repetition of same word)
8 mo were able to make the distinction (preferred transP =1)
This strategy does not require knowledge, its domain general (also w tones, shapes etc)
New paradigm of transition prob.
Tested it with words embedded in equally-frequent but very low transition probability noise syllables.
- Also here the words were extracted more than non-words, but to lesser extent
Possible underlying mechanism TP
The entire sequence is too long for the phonological loop . If repetition is too far apart it sharder to find the words within noise. Tested this with:
C-words; placed in clumped distribution (close together)
F- words: placed more evenly w at least 24 syllables inbetw.
The occurrence, average repeat distance and transition p was identical for both groups BUT only C words were preferred over non-sense. Meaning clumpiness matters due to the limits of WM
Straddled words
By adding prosody to a string of syllables, it carves the word detection. Those words that are straddled are at the cut off of the prosody, = controu-straddling. Those within the string are called contour-internal
Now the word detection compared to non words is better for contour-internal words, when compared to a string without prosody
Remembering sequences
Beginnings and ends are better remembered than middle parts.
Prosodic edges are preferred over prosodic middles
