Speech perception Flashcards
language organization helps us to deal with
variability in the speech signal and in listening conditions
we can put language into a
Hierarchy of linguistic structure < check onenote
models of speech perception =
serial, parallel, interactive
all models have the following hard problems >
variability between speakers, variability due to context (co-articulation) and we can solve this via categorical perception
2) another hard problem is no consistent boundary between words ( not clear where a word starts or ends) and we can solve this via segmentation
categorical perception differs in
voice onset times
categorical perception =
we perceive them as one category only
segmentation
a way to extract where words end and where they start
hierarchical approach to segmentation
check notes
optimal way of segmentation = tier 1 > lexical
sentential context (pragmatics, syntax and semantics) lead to lexical knowledge < we use this lexical knowledge to segment words
if we have poor lexical information, we rely on
phonotactics (acoustic-phonetics, allophony, coarticulation)
if we have poor segmental information, we rely on
word stress
2 context effects in speech perception =
phonemic restoration effect (warren and warren, 1970) > listeners use context to predict the incoming speech signal which can repair phonemes replaced by a meaningless sound , listeners rapidly engage in appropriate phonemic processing in auditory coryexy < consistent with interactionist account
2) ganong effect > an ambiguous phoneme perceived as one completing a word rather than non word
alex fastened the button
constraining context
alex chose the button
neutral context
TRACE model > mclelland and elman 1986
processing units or nodes at 3 levels, connections between levels always facilitatory , connections at same level always inhibitory , competitive process = words with greatest activation “wins”
