Computational Models of Speech Perception

What are some challenges to lexical access?

Co-articulation

Accents

Invariance Problem

What is the invariance problem?

Problems with the definition of acoustic properties

How can we disambiguate the speech stream?

Categorical perception

Perceptual Learning

Top-down Perception

What is categorical perception?

Ability to distinguish between sounds on a continuum based on Voice Onset Times (VOT)

What is perceptual learning?

Adjust categorical perception based on sounds we hear

What lexical characteristics affect speed of lexical access?

World length

Neighbourhood density

Frequency

What is The Cohort Model (Marslen-Wilson, 1987)?

It predicts that we access words in the lexicon via activation of all words sharing initial features and gradually deactivate words that stop matching the features

What is The Trace Model (Elman & McClelland, 1999)?

It predicts that features activate phonemes that activate words with a gradual increase in activation of words that match all features, so that the word with the most activation wins

What is the uniqueness point?

The point where nothing else in your lexicon matches your input

What are neighbourhood effects in the Cohort Model?

Words that match the acoustic input compete for activation

What are frequency effects in the Cohort Model?

Words with high frequency have resting states- less activation is required to recognise high frequency words

Is the Cohort Model bottom-up or top-down?

Bottom-up

What are the 3 stages of the Cohort Model?

Access

Selection

Integration

What is the access stage in the Cohort Model?

Acoustic-Phoneme information is mapped onto lexical items

What is the selection stage in the Cohort Model?

Candidate words that mismatch the acoustic input are deselected

What is the integration stage in the Cohort Model?

Semantic and syntactic properties of the word are integrated and checked against the sentence

How does the Cohort Model process context?

Spreading Activation from a semantically related word primes the correct word

What does the Cohort Model predict?

Items that match acoustic input but do not match the sentence context are activated

Items that match acoustic input but do not match the sentence context are deactivated once the word is selected

What is the Revised Cohort Model?

Context influences selection/integration of word into sentence

What is the TRACE Model?

Words are recognised "incrementally by slowly ramping up the activation of the correct units at the phoneme and word levels"

What is the gradual activation of items that match the input?

Activation the phoneme regardless of where it is

What is lexical competitive inhibition?

Not removing words, words can increased activation

What do the units on the feature level represent?

Features

Phonemes

Words

What are facilitatory connections?

Connections between levels that travel up/down and across the feature level

Are connections between nodes inhibitory or excitatory?

Inhibitory

What is the Radical Activation Model?

Any consistency between input and representation may result in some degree of activation

How is the TRACE model affected by top-down processing?

Activation from word to feature level

How is the TRACE Model different from the Cohort Model?

Emphasises top-down processing

Accommodates the activation of rhyming competitors

Does not provide an account of how context might affect speech perception

How is the Cohort Model different from the TRACE Model?

Minimises the impact of top-down processing

Predicts that lexical access is biased towards the activation of words with shared onsets