PSY2002 W1 Language I Flashcards
An introduction to computational models of speech production
Language Definition
an exchange of information: language processing occurs under the level of conscious awareness. You are not aware of all information being processed to allow you to speak.
Methods to test speech prodcution
Timing of speech onset, hesitations and pauses
Tip-of-the-tongue state
Speech errors
Why do we hesitated
Word representations in the lexicon compete for selection due to spreading activation.
Competition between words slow down the speech
Tip-of-the-tongue state
Unable to retrieve the word you want
Tip of the tongue - grammatical gender (Vigliocco et al. 1997)
In language with grammatical gender, you are able to state the grammatical gender of the word but cannot access the phonological form of the word
Can retrieve the syntactic elements that go with the word. Cannot retrieve the sounds to expression the word.
What does the tip of the tongue state suggest
Sentence is planned, structured is planned, syntax in place but strugglin gto retrieve the sound, suggest specific levels of procssing required to produce speech
Speech errors - specific processes
Speech errors tend to occur within specific processes, such as the selection of syntatic units or form (syllables and phonemes)
Specific levels of processing language
Semantic processing (think about the concept) => Syntactic adn morphemic processing => Articulation (form processing)
Types of speech errors
Omission, Addition, Lexical substitution, Morpheme substitution, phonological substitution, lexical exchange, phonological exchange, anticipations, perseverations, blends, morpheme shift, morpheme standings
Speech errors - language processes
Errors tend to occur within the following process rather than corssing th eboundaries between them.
- Semantic processing
- Syntactic and morphemic processing
- Articulation (form processing)
Semantic processing - Speech errors
Semantic processing = conceptualisation semantic blend errors
Syntactic and morphemic processing - speech errors
formulation syntactic and morpheme exchange
Articulation (form processing) - speech errors
articulation word and phoneme exchange.
Model of utterance generation (1971): Victoria Fromkin
First models of speech prodcution. Serial processing (discrete).
1. Meaning
2. Syntactic structure [article + // + plural + possessive + article //]
3. Position of intonation
4. Lexicon look-up (finds words and generate phonological segments)
5.Morphophonemic constraints
Discrete Models
Each stage must be completed before the next begins.
Activation is feed forwards can only move from meaning to sound
Dell, G. S. (1986). A spreading-activation theory of retrieval in sentence production.
Assumes processes are fluid, information can go back and forth through the language production processes
Levelt, W. J., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production.
Assumes processes are serially organised and must complete 1 before the other begins – similar to Fromkin
Types of models (cognitive psychology)
Box and arrow model
Computational model
Box and arrow model
Fromkin’s model
Describe the relationship between different mentla prossess under the assumption that the mind opperates like multi stage information processing machines.
We can study this models by manipulating the input and observing the output.
Computational model
Implemented as functiong computational models = (Dell, G. S. (1986) and Levelt, W. J., Roelofs, A., & Meyer, A. S. (1999))
Statistical models
a mathematical relationship between variables, that hold under specific assumptions
Theoretical models
description of the relationship between different mental processes, that makes assumptions about the nature of these processes.
Behaviourism
Input => Brain => Output (behaviour)
Neural Networks
Nodes represent groups of neurons. Activation of nodes spreads activation across the network.
Dell (1986) Spreading activation theory
The theory is a cascaded interactive model. Information is active in parallel.
Spreading activation in both semantic level adn phonemic level.
All diffirent stages happen at once, we may also activated other semantically related concepts and could account to word errors.
The thoery and prediction - Dell 1986
Spreadin gactivation results in activation of related items at all levels of processing. Processing items that verlap in semantics adn or phonology will result in higher errors rates than processing items that do not overlap
Experiement Ferreira and Griffin 2003
Participants named target pictures (picture of a priest) after first reading sentences that primed a semantic competitor (‘nun’), a semantic and phonological competitor (‘none’) or an unrelated word (‘match’).
In an interactive model the combined influence of Semantic and Phonological similarity results in an error.
Participants were more likely to mistakenly produce the semantic or phonoligcal and sematic competitor than they were to produce an unrealated word.
The Evidence: Ferreira and Griffin (2003)
Participants can successfully inhibit ‘match’ to say ‘priest’ but fail to inhibit ‘nun/none’ and say ‘priest’. Semantically and phonologically related items are active at the same time and can be selected for output.
Is speech production based on interactive processes?
Evidence from Ferreira and Griffin (2003) demonstrates that semantic and phonological similarity can result in speech errors suggesting that processing is fluid and interactive
Speech Errors in Dell’s 1986 Model: How does this account for the speech errors observed by Fromkin?
- errors may occur due to more than one concept receiving the same amount of activation. Both concepts become active resultign in a cascade of processing to the level of phonology
Dell 1986 Interim Conclusion
The model provides a convincing account of speech errors. Evidence from lab induced speech errors supports the suggestion that processing is interactive.
However, some suggest that the parallel processing in the Dell model would predict that we make far more speech errors than we actually do.
Levelt et al., 1999 WEAVER++ = Word-form Encoding by Activation and VERification
Similar model to Dell but Discrete.
Activation can only go in one direction, process must be complet before another can begin, prevents errors in speech all the time.
LEMMA
The word form and the syntactic information that links with the word form
Competitive processes
Selection of items at each level is based on competition between items. Once word has been selected, this node sends inhibitory signal to other related concepts to stop activation.
WEAVER++ (levelt et al.) predicts that items which are semantically similar will inhibit processing.
Dell 1986 - Inhibition
No inhibition in dell’s model, due to the lack of inhibitory connections processing items which are semantically similar will not be inhibitive
The Experiment: Wheeldon & Monsell 1994
Participants were asked to answer a question out loud and then name a picture ( What is the largest creature that swims in the seam - whale - picture of a shark)
Once you activate the concept of whale you inhibit related concepts like shark so you will be slower in naming the image aka shark
The Experiment: Wheeldon & Monsell 1994
Results:
Mean reaction times for semantically related adn unrelated items. It takes longer to record the word of an image semantically related to the answer to the question
Is speech prodcution based on interactive processes
Evidence that semantically related items inhibit processing of each other suggest that speech production processes are discrete.
Speech errors in Levelt et al.’s WEAVER ++ Model
Errors may occur if the wrong concept is selected and processed to the phonological level. Two concepts that are both viable options for the utterance could be selected and processed to the phonological level
WEAVER ++ Interim conclusion
Levelt et al.’s model has to suggest that speech errors are ‘special cases’, but WEAVER++ accounts for inhibitory effects in speech production where Dell’s model appears ill equipped to accommodate these effects
Interactive processing
If processing of spoken words is interactive and parallel (Dell, 1986) phonological processing is likely to occur for all items that are related by semantics and/or phonology to the target item.
Discrete processing
If processing of spoken words is discrete and feedforwards (Levelt et al., 1999) phonological processing should only occur for items that have been selected and should not be affected by ‘non target’ information.
Semantic interference Meyer & Damian (2007)
When phonological items overlap (related items) are faster, because the DO of doll and dog will activate the concepts quick.
Phonological should not have an effect on our speech production then there must be some interaction even if they don’t match the studies
