Please enable JavaScript.
Coggle requires JavaScript to display documents.
Speech perception - Y2 - Coggle Diagram
Speech perception - Y2
Defining language
Language is a signalling system -
- Arbitrary - there must be a connection between the sound or word and its meaning
- Productive - there is an infinite number of different expressions that can be communicated
- Displaced
- Dynamic
- Communicative
- Mulit-level
- Regular - has structure
- For most of us, listening to and understanding speech is almost effortless, but speech recognition is arguably one of the most complex auditory processes
-
Phonology - the most basic auditory unit of spoken, natural language:
- English has 34-44 based on accent - critical to understand language
- /r/ and /l/ are the only thing that distinguishes rake and lake etc
-
Is language special?
Special mechanism - linguistic relativism link:
- Motor theory (Trout, 2001; Liberman and Mattingly, 1989)
- We are born with a special neural mechanism that allows us to decode speech stimuli (Samuel, 2011)
-> This allows linguistic information to be privileged in processing, allowing it to be processed more quickly and efficiently than other acoustic information
-> Phonetic module - specialised neural mechanism that singularly processes speech (i.e. both segmenting auditory information and allowing access to word meanings - Liberman, 1996)
Categorical perception - Mann & Liberman, 1983 - evidence for a specialist mechanism:
- Physical space and the perceptual space of letters
- Rather than perceiving as a continuum, listeners either hear a clear cut 'ba' or 'ga'
- This is not the case when the sound presented does not sound like speech
- Suggests a specialist mechanism is used in speech, one which is not used in non-speech processing
- Miyawaki et al, 1975 - found that adult English speakers show categorical perception for /r/ v /l/, and Japanese speakers do not show this same boundary because they are the same phoneme - perceptual space
-
General mechanism - universality link:
- Direct realist theory (Cleary and Pisoni, 2001; Massaro and Cole, 2000)
- Humans use the same neural mechanisms to process speech sounds and non-speech sounds
- Speech perception is a learned ability
-> Our perception of speech relies on familiar cognitive processes such as feature recognition, learning and decision-making (Foley and Matlin, 2010)
-> As it is a learned ability, we learn to distinguish speech sounds in the same way that we learn other cognitive skills
Language familiarity effect - Werker and Tees;
- 6-8 month olds are very good at distinguishing foreign phoneme changes
- But this diminishes over time, suggesting experience affects speech perception - speech is not entirely reliant on an innate special mechanism, also dependent on learning
- Perceptual narrowing
Cross-language speech perception - Werker and Tees, 2002:
- Infants can discriminate non-native speech contrasts without relevant experience, and there is a decline in this ability with age - decline happens in the first year of life, and is a function of specific language experience
Moon, Lagercrantz & Kuhl, 2013:
- Exposed 20-40 hour old babies to vowels unique to their mother's language and another language
- Mothers were either Swedish or English speakers
- Babies sucked their dummies harder when listening to the foreign language
- Suggests babies listen to speech in the womb and this affects their phonetic perception
Choi, Broersma and Cutler, 2017:
- Compared Dutch adults who had been adopted from South Korea in the first 5 months of their life and had no knowledge of the Korean language
- Found those who had been adopted were significantly faster to learn Korean consonant distinctions faster than Dutch controls
- Indicates that phonological knowledge is in place before 6 months of age
-
Summary -
- There has been debate as to whether there is a special neural mechanism for speech perception or whether it relies on more general cognitive processes
- There is some evidence to suggest privileged processing of speech over other acoustic signals
- There is developmental evidence to show the speech perception develops over time
- There are also studies showing speech is processed in a similar manner to other auditory signals
- Speech perception relies on an interaction between multimodal information
- Models of word recognition try to accurately represent the word identification system in some way that can then be manipulated in a controlled manner, often trying to find the right balance between top-down and bottom-up processing
- Cohort model is based on selecting potential lexical candidates which unfolds more as the word is heard
-> As it has been shown to not account for mispronunciations and weak effects of context it was adjusted
- The TRACE model is connectionist and based on activation levels which are affected by top-down and bottom-up processing
-> TRACE is impenetrable and does not accurately predict top-down effects