Language III Flashcards
(10 cards)
segmentation problem
no pauses between word boundaries
shown on spectrograms/listening to foreign speech
how to english speakers get over segmentation problem
metrical segmentation strategy (MSS)
stress syllables taken as the likely start to word
3/4 of all syllables occur at beginning of content words
2/3 of unstressed syllables are grammatical words
listeners identified word mint within another word if it is stressed
exceptions: alert, need context and word knowledge
do we always use stress
Mattys et al. (2005) proposed hierarchy of segmentation cues
context and word knowledge - most strongly in optimal listening conditions
word/syllable stress being when the speech signal in impoverished.
Lexical selection stats as soon as the the word onset is identified (2)
shadowing paradigm - participants repeat passage, correct incorrect word, sometimes even before its been presented in full
Mispronunciation detection task - detected more frequently when mistake was at word onset
Cohort Model of spoken word recognition
gating paradigm - each cohort “activates” all relevant words in mental lexicon
more cohorts added until uniqueness point helps
But sometimes later point activates lexical entry
Participants instructed to grab “speaker” or “beaker”, and can get confused
Useful if word boundary is unclear or people don’t hear word onset
effect of context
“bug” or “spy” -> when ambiguous, priming showed both equally activated
phoneme restoration effect - during noise, word recognition is higher when words are in context than in isolation
words sliced out of conversation - word recognition at 50%, then increased dramatically with one or two neighbour words
problems with context studies + evidence
don’t show how and when context is important; could be
(1) ACTIVATION (initial perception of word)
(2) SELECTION (narrows down possible candidates)
(3) INTEGRATION (influences decisions about missing info)
gating study - not activation, but selection.
spoken word recognition (evidence in conjunction with Marinkov, 2003)
Hickok & Poeppel (2007) dual stream model
(1) spectro-temporal analysis (primary auditory cortices)
(2) phonological (superior temporal gyruses, slight left bias)
(3) ventral stream or dorsal stream
evolutionary basis of spoken words
fMRI on macaques - calls engage frontal-temporal regions of BOTH hemispheres
key to humans: grammar
(1) dorsal frontal-temporal regions in the left hemisphere
(2) connections in left are more pronounced in humans than chimps/macaques
