Weeks 8-9 Ruth Ingram Flashcards
What is speech?
Translating an idea into a sentence structure and producing the appropriate word.
What are three steps that most speech models would agree on?
- Selecting a word based on the meaning one wants to convey.
- Retrieving and specifying the grammatical properties of the word.
- Retrieving the phonological patterns needed to articulate the word.
What is lexicalisation?
Selecting a word based on the meaning one wants to convey from the network of options.
(See spreading activation model).
What is the lemma?
The world level information that specifies the syntactic components of a word.
(It’s modality independent - whether you’re writing or speaking it’s the same).
What is the lexeme?
The phonological code that drives articulation.
What are the three steps of speech theories and their associated term: lexicalisation, lemma and lexeme?
- Selecting a word based on the meaning one wants to convey.
= Lexicalisation - Retrieving and specifying the grammatical properties of the word.
= Lemma - Retrieving the phonological patterns needed to articulate the word.
= Lexeme
Explain the speech errors that occur when you have issues with lexicalisation.
Freudian slips:
The substitution of one word for another that is sometimes thought to reflect the hidden intentions of the speaker.
Name the 3 speech errors that occur when you have issues with the lexeme.
Tip-of-the-tongue phenomenon.
Spoonerisms
Malapropisms
What are Tip-of-the-tongue phenomenon (ToT)?
Knowing conceptually what you want to say but being unable to retrieve the corresponding spoken form.
What are spoonerisms?
The initial consonants are swapped between words.
What are malapropisms?
Saying a word with a similar phonological form to the intended word.
(Sounds similar but incorrect words)
Outline Levelt’s model of word retrieval.
Word retrieval happens in 2 discrete, sequential stages.
E.g., Lexeme retrieval does not begin until lemma selection is complete.
What is some evidence from natural speech to support Levelt’s model?
Correct lemma selection even for homophones.
We rarely use homophones in the wrong way, suggesting that the lemma comes first.
What is a homophone?
One of two or more words, such as night and knight, that are pronounced the same but differ in meaning, origin, and sometimes spelling.
What is some evidence from speech errors to support Levelt’s model?
ToT Phenomenon:
Lexicalisation and retrieval of lemma without retrieval of lexeme (Stage 1 happened without stage 2)
Give two examples of evidence that suggest issues with Levelt’s model.
Mixed errors:
- Speech errors where the intended word and produced word are similar semantically and phonologically.
ToT Phenomenon:
- How does it occur when you know which phoneme the word starts with but not the word?
(Should be impossible if the lemma has to come first)
Outline Dell’s model of word retrieval.
It includes three stages:
- Lexicalisation (Semantic)
- Lemma (Syntax, lexical, word level)
- Lexeme (phonological, phonemes)
There is interactivity between these layers.
Multiple kinds of information (semantic, lexical, phonological) are active simultaneously with selection of the ‘winning’ word retrieved arising via a competitive process.
Give an example of what ‘competitive process’ means in Dell’s model of speech production.
Give the model some input (e.g., picture of dog)
Triggers activation of nodes at different levels.
How ‘active’ nodes become is based on the ‘weighting’ of their connections.
Activation spreads based on weighting until a pattern of nodes exceeds a threshold.
The word represented by the ‘winning’ pattern will be retrieved.
(Slide 11 of Speaking Brain 2 if confused)
How does interactivity in Dell’s model explain mixed errors?
Activation simultaneously at the semantic and phonological layers:
The produced word is both phonologically and semantically similar (“oyster” for “lobster”)
How does interactivity in Dell’s model explain parts of the ToT phenomenon?
Activation at the phonological layer but a word hasn’t “won” the competition yet.
Explains ToT whilst knowing it starts with a certain phoneme.
What is Logopenic progressive aphasia (LPA)?
Speech characterised by word-finding impairment and phonological speech errors (phonemic paraplegias).
BUT
Semantic memory is normal.
(effects writing & speaking abilities) - check living with LPA YT vid
How could Dell’s model explain Logopenic progressive aphasia (LPA)?
Could simulate this pattern of impairment with issues with the phonological-lexical layer, with intact semantic-lexical layer.
What is LPA associated with? - disease and location of atrophy.
Alzheimer’s disease
Atrophy in the left posterior temporal cortex.
What is the SLAM model of speech production?
Semantic -> Lexical -> Auditory -> Motor
Process is as follows:
Temporal lobes -> Auditory cortex -> inferior parietal cortex -> Frontal cortex
Based on evidence of dual stream model of speech.
Name the two streams in the Dual stream model of speech and explain them.
Dorsal stream for speech production (motor):
- Auditory information is passed through the parietal lobe to the inferior frontal gyrus and motor cortex.
Ventral stream for speech comprehensions (semantics):
- Auditory information is passed through temporal cortex to temporal pole for comprehension (semantic knowledge).
What does “stream” refer to in the dual stream model of speech?
A collection of brain areas working together.
Outline the timing of activation in brain regions (3) in the dorsal stream of the dual stream model during speech repitition.
Superior Temporal Gyrus first.
- When the patient first hears what they are being asked to repeat.
Brocas area.
- Process between hearing and repeating it.
Motor cortex.
- Articulation of speech they are being asked to repeat
What did Walker & Hickok propose in their SLAM model about speech production?
That it involves the ventral stream in reverse THEN the dorsal stream:
Semantic -> Lexical -> Auditory -> Motor.
Temporal lobes -> Auditory cortex -> Inferior parietal cortex -> frontal cortex.
Describe Fridriksson et al. (2018) patient evidence for the SLAM model.
(hint - philadelphia naming test)
METHODS:
- 159 individuals with post-stroke aphasia.
- Behavioural test batter: variety of language tasks.
- (E.g., picture naming or matching a picture to a heard word - recognition)
- MRI of grey and white matter.
RESULTS:
- Lesion-symptom mapping created a connectome (see bottom) in speech recognition, word recognition and Philadelphia naming test which was akin to the SLAM model.
(connectome = complete map of neural connections in a nervous system).
What is syntax?
The rules that specify how words can be combined into sentences.
What is parsing?
Assigning a syntactic structure to words.
Outline Paul Brocas patient ‘Tan’ study and findings.
Broca was a researcher interested in localising brain functions to brain areas.
Observed an individual (Tan) who could only produce single words not sentences (just the word ‘tan’)
Upon autopsy, saw that there was damage to left frontal lobe region.
Broca presumed this region was critical for production sentences.
This region became labelled Broca’s area.
What is the cardinal symptom of Broca’s aphasia?
Agrammatism (loss of grammar) in speech production.
Results in speech which is characterised by the absence of function words and verbs.
It’s often assessed through the cookie theft test.
Describe the Friederici et al. (2006) study into the neural correlates of Broca’s area.
METHODS:
- fMRI of 13 healthy participants.
- Task: judge validity of sentences with increasing grammatical complexity
RESULTS:
- Broca’s area (BA44) had greater activity with comprehension of sentences with increasing syntactic complexity.
What are the two debated answers to what feature drives parsing?
Structure-driven:
- Based only on syntactic properties (syntax separate from semantics)
Discourse-driven:
- Influenced by semantic properties of words.
(semantics integrated with syntactic processing)
