Please enable JavaScript.
Coggle requires JavaScript to display documents.
SENTENCES AND MEANINGS :PENCIL2: D+L2 LECTURE 1 - Coggle Diagram
SENTENCES AND MEANINGS
:PENCIL2:
D+L2 LECTURE 1
Word Recognition and Lexical Access
Lexical access
refers to the process of finding and retrieving all the stored knowledge we have about a word.
Principally - its semantic representation (its meaning)
It can be tempting to assume that these processes are serial and autonomous.
In other words, we might assume that one follows on from the other and that recognition of the form isn’t influenced by whatever information about that word is stored.
Word recognition
This is about identifying forms whereas lexical access is about retrieving meaning.
What do we already know about accessing meaning?
Grosjean (1980)
Gating Task
Where you play bits of the word in gradually increasing chunks and ask people to try and guess what the word might be.
Context influences recognition point (RP)
E.g. RP of “camel”: 333ms in isolation
But... RP is 199ms in “At the zoo, the kids rode on the camel”
Suggest that sentence context influences the word recognition process.
Zwisterlood (1989)
Cross-modal priming
Briefly activate numerous meanings of potential candidates as the speech unfolds.
‘Captain’ primes ‘ship’
‘Capital’ primes ‘money’
‘Cap’ primes both ‘ship’ and ‘money’
Activating lots of different meanings at the same time.
So there’s no hard division between word recognition and meaning activation.
We are activating meanings before we know what the word is, before we’ve completed the word recognition process.
It can’t be serial, it must be that we’re doing some lexical access before we’ve completed the word recognition.
Processes should be cascaded.
Ambiguous Words
Captain/Capital is a transient ambiguity.
At some point you’re going to get enough information to discriminate between the two, you just have to wait for a little while.
Information at the end of the spoken word resolves the ambiguity, just have to wait for the discriminating information to arrive.
But in other cases,
the discriminating form doesn’t arrive…
E.g. ‘Bank’ - has many meanings (OED - 5 different meanings (3 noun, 2 verb), 47 separate senses)
Different meanings of a word are semantically unrelated.
E.g. ‘bank’ as in High Street Bank, and ‘bank’ as in River Bank are completely unrelated semantically, they just share the same form.
Different senses of a word are semantically related, but are still separable.
E.g. ‘bank’ as in High Street Bank and ‘bank’ as in Blood Bank are related in the sense that both definitions have the common semantic representation to do with being a repository, a base for storage, yet are still separate.
Terminological Ambiguity
Homophone
Different meanings, same sound
E.g. Night/knight
Homograph
Different meanings, same written form
E.g. Lead [dog/metal]
Homonym
Different meanings, same written and spoken form
E.g. Bank [money/river]
Heterographic homophone
Different meanings and written forms, same sounds
E.g. Night/Knight
Heterographic homograph
Different meanings and sounds, same written form
E.g. Lead [dog/metal]
Polysemous
Often used interchangeably with homonym. Stricter definition {cf. Harley} is multiple senses within same dictionary definition.
E.g. Twist [dance, action]
Do these terms matter?
Rodd et al. (2002)
She looked at different kinds of ambiguity for words.
She wanted to know how quickly people could recognise these words.
‘Does ambiguity influence single word identification?’
Lexical decision task.
The idea is that people make quick decisions about whether something is a word or not.
Examples of the types of words that were used (varied in the number of senses and the number of meanings):
Few senses, Few meanings = snake
Few senses, Many meanings = cricket
Many senses, Few meanings = hook
Many senses, Many meanings = seal
Asked: 'Is this a word?' (Yes/No)
You would assume that if there was no involvement of meaning in the recognition process, then none of this would matter, because all you need to do is identify a particular word - don’t need to access its meaning.
Found that these variables do matter.
Recognition is faster for words with:
Many senses
Few meanings
E.g. hook
Word recognition is not independent of meaning.
The two somehow interact
It's helpful to have few meanings if you want to recognise a word quickly.
What does this tell us about word recognition?
Can be explained in terms of semantic feature models of lexical representation
Meanings represented as collections of semantic features
Bank (river) = [side, water, edge, slope, grassy, etc.]
Bank (money) = [brick, money, cashier, cashpoint, building, etc.]
Multiple meanings are incoherent sets of features, conflict needs to be resolved, which takes time.
Suggests that having lots of meanings is bad for recognition, because we’ve got to resolve this ambiguity before you can say ‘this is a word’ or ‘this isn’t a word’
Multiple senses are more coherent sets of features, leading to more flexible representation?
Word recognition involves a meaning resolution process.
NOT serial and independent.
Influence of Sentence Context
‘The cashier had to hurry because he was late for work. He ran along the High Street, past the opticians and burst through the doors of the bank.’
Recognition of ambiguous words without any contextual cues is atypical.
How does sentence context help?
All theoretical models assume that sentence context has some role to play in meaning selection.
They differ on:
When the context is brought into play (early in the lexical access process or later)
How - the extent to which higher level processes influence lower level processes (interactive, autonomous)
Role of meaning frequency
Multiple Access Model (MAM)
Swinney (1979)
When an ambiguous word is encountered, all meanings are accessed autonomously in a context-independent way.
Without any regard to the sentence context.
All meanings are activated at the same time in your lexical access process.
Following meaning activation, the contextual appropriate meaning is selected.
Autonomous/modular activation of meaning
Late timing of context effects
Low influence of frequency
Context-guided single-reading lexical access
Schvanevelt et al. (1976)
Context used to restrict access to only the appropriate meaning.
Don’t bother to activate or hypothesise alternative meanings that don’t fit the context.
Inappropriate meanings never accessed.
Interactive model - the context is directing the lexical access process.
As soon as lexical access begins, context comes into play.
Interactive activation of meaning
Early timing of context effect
Low influence of frequency
Reminder-priming studies
Cross-modal priming (e.g. Zwitserlood, 1989)
Involves hearing a spoken stimulus and seeing a visual target (hence, cross-modal)
Spoken prime word followed by semantically related visual target.
Participants make a speeded lexical decision judgement on the target (is it a word?)
“Captain-ship”: FAST
Quick when there is a semantic relationship between the two.
“Pheasant-ship”: SLOW
-
Provides a measure of meaning activation.
On-line evidence of meaning activation
Swinney (1979); Over 1000 citations!
Seminal study in terms of this area of psycholinguistics.
Primes were ambiguous words (e.g. bugs) embedded in biasing (or neutral) sentences.
Example of a biasing context
:
‘Rumour had it that, for years, the government building had been plagued with problems. The man was not surprised when he found several spiders, roaches and other
bugs
in the corner of his room.’
Cues in the sentence leading up to the ambiguous word that would favour one interpretation of ‘bugs’ (insect) over the other interpretation of ‘bugs’ (spyware).
Neutral sentences did not contain these cues.
‘When you hear these words, to what extent do you activate both of these meanings?’
Or is it just the meaning that’s appropriate for context which is activated?
Looked at the activation of meanings of ambiguous words using speed of lexical decision to visually-presented target words.
Visually presented target words:
ANT (contextually related)
SPY (contextually inappropriate)
SEW (unrelated)
Early effects of context
Target words presented at the offset of the ambiguous word.
Get priming for both meanings.
Suggests that the contextually appropriate and the contextually inappropriate meaning are both being activated.
Context has no effect.
Fits with the multiple access model (MAM) (Swinney, 1979)
Later effects of context
Target words presented three syllables later (~½ a second?).
Given a bit more time for context to have some role in identifying the appropriate meaning.
Get priming only for appropriate meaning, not the inappropriate meaning.
Context effect occurs after lexical access.
There’s this early stage at which you get no significant context effects.
In totality, this fits with the multiple access model (MAM) (Swinney, 1979).
Syntactic Ambiguity
Tanenhaus et al. (1979)
Syntactic ambiguity = For example, when a verb and a noun have the same form (e.g. WATCH)
How is noun/verb ambiguity resolved?
Syntactic ambiguity resolution may operate differently from semantic ambiguity.
E.g. ‘WATCH’: Noun or verb?
Prime sentences biased one or other interpretation:
E.g. ‘Boris began to watch.’ (verb-based)
E..g ‘Boris looked at his watch’ (noun-based)
Do you get priming for both the verb interpretation and the noun interpretation at the points where you’ve received this context?
Initial priming of both interpretations (e.g. LOOK for verb, TIME for noun)
You do get priming of both meanings, both the noun and the verb meaning.
Consistent with the Swinney’s Multiple Access Model.
Later (~200ms) priming of context-appropriate item only.
Similar pattern to test of semantic ambiguity.
Initially we seemed to be retrieving both the syntactically appropriate and the syntactically inappropriate interpretation.
Then very quickly selecting between those to find the one that works.
Meaning Frequency
Ambiguous words vary in the relative frequency of meanings.
Balanced:
Both meanings are equally common.
Unbalanced:
One meaning is dominant
Do we still access all meanings when the ambiguous words are unbalanced?
Context and unbalanced homonyms
Tabossi & Zardon (1993)
Context strongly constrained the core meaning of the homonym.
E.g.
Dominant meaning:
‘The violent hurricane did not damage the ships which were in the
port
, one of the best equipped along the coast.’
E.g.
Subordinate meaning:
‘Misled by the similar colour the landlord took a bottle of Barolo instead of
port
, and sold it to the guests.’
Visual targets:
safe, red, short
Cross-modal priming
We don't always get multiple access initially in those circumstances.
Some selectivity?
For
subordinate
constraining context
Multiple access.
Priming for both the contextually appropriate meaning and the contextually-inappropriate meaning.
For
dominant
constraining context
Selective priming
If you’ve got a context that favours the more common meaning, then obviously you get priming of the more common meaning.
You get priming of SAFE compared with SHORT.
You don’t get priming where the context nor the balance of frequencies favour it.
Evidence
against
the strong modular account (Swinney, 1979)
Meta-analysis
Lucas (1999)
Reviewed the literature on sentential context effects.
Weaker interactive models suggest that sentential contexts will not always rule out inappropriate meanings immediately, they may just bias activations.
Biasing effects may be difficult to spot - depends on experimental power.
Lucas analysed all experiments in the area, and found small systematic biasing effects across the experiments.
There was a small difference in the strength of priming for the contextually appropriate versus the contextually inappropriate mean.
Not significant in each individual experiment, but across lots of studies which an average is being taken from… may have an effect.
Favouring the contextually appropriate mean.
Persistent Vegetative State (PVS) patients
Rodd et al. (2005)
Brain regions involved in lexical ambiguity resolution.
Contrast: high ambiguity > low ambiguity
To what extent does the brain work a little bit harder for cases where there’s high ambiguity versus lower ambiguity?
What does this study tell us?
We know that some patients in persistent vegetative state are sensitive to high-level distinctions in language comprehension.
There are various parts of the brain that show a stronger BOLD response in cases where there’s lots of ambiguity in sentences that were presented.
Can see large parts of the
temporal lobe, superior, inferior temporal lobe and also some of the frontal lobe
as well.
Those are the bits of the brain that seem to be sensitive to lexical ambiguity in normally comprehending adults.
Tested persistent vegetative state (PVS) participants on the same contrast.
For selected individuals, you could see brain responses that were very similar to the controls in terms of this contrast.
Suggests that for those PVS patients, there is some activity that’s similar and sensitive to lexical ambiguity.