Dialogue Flashcards
Dialogue - Three components
- Evidence that we align to who we’re talking to
- express ourselves in a similar way - Do we take into account what we know about our partner when we are talking to them? YES - AUDIENCE DESIGN - we formulate our language output
- Evidence either way - Interactive-alignment model
How do we study dialogue?
Not often studied - difficult to control what’s happening in a dialogue
One way - use confederates - makes it artificial/scripted
Really time-consuming
Difficult to measure
- what’s the dependent variable?
– reaction time
– content of speech
Evidence on how we align to the person we’re speaking to
Picks up accent and mannerism - we have a tendency to align with the person we’re talking to - we start expressing ourselves in the same way
Research into SPEECH ACCOMMODATION THEORY
How we speak influences the way people perceive us - idea was that we change the way we speak to make people like us
So our speech partners have a better impression of us
Extent we use speech accommodation depends on
who we are interacting with - we match gestures more if we like the person
“Response matching” Argyle (1969)
We match our conversational partners on a range of different factors e.g. length of utterances, words used
e.g. this effect when you realise you’re both whispering but unsure why/talking really rapidly
How does this speech accommodation/response matching happen?
Conceptualisation
Word repetition
Structure repetition
Conceptualisation
The way we represent situation we are in (at a high-order level)
Garrod & Anderson (1987)
Two participants had to navigate their way through maze on screen - in separate rooms
Same maze layout
There were different gates in the maze - could only get through gate if your partner got to that point on their maze to and switched a switch specific to their maze to let you through - complex
Have to explain to partner where you were in this maze
Garrod & Anderson (1987) conceptualisation example
Participants tended to converge to one of 4 description schemas
- Path type
- Co-ordinate description type (MATRIX GRID)
- Line type description (top line, 4th box along)
- Figural description - pictorial/images
Results of G&A (1987)
Ppts would converge/adapt to similar forms of description over course of dialogue and DIDN’T appear to explicitly negotiate this
Principle of output/input co-ordination (G&A propose this)
You re use the same rules as needed to understand your partner
You formulate your output (your speech) according to same principles of interpretation as those are needed to interpret the most relevant input (your partner’s speech)
Garrod & Doherty (94) tested effects of relationship between speakers
Same maze task
3 conditions
1. Speakers changed partners regularly, but pairs were always drawn from same group of people - A with B, C with D, A with D - all were paired together at some point (Community condition)
- Speakers changed regularly NOT from same group of people - random
- Same partner for every game
In community condition - massive CONVERGENCE - by game 5, everyone using same description scheme
Results of G&D (94) - effects of relationship between speakers
- convergence to same description scheme
- never really settling on 1 description scheme
- real mix - different pairs of speakers converging to different description schemes
G&D - what does that tell us about how language works?
DIALECT
Communities converge to pronounce certain words
In a community of people = powerful convergence effect
Causing DIALECT, JARGON
Word repetition in dialogue (Brennan & Clark, 1996)
Lexical entrainment .
Picture description task e.g. shoe - fish - dog
or harder heels - trainer - smart shoe
This study looked at way people converged on words when asked to describe pictures
Ppts established a CONCEPTUAL PACT - conversation specific - pact about how they were going to refer to one thing a certain way
Lexical entrainment (Word repetition) - Brennan & Clark (1996)
Repeated use of same or related terms to refer to an object
Established jointly between ppts
become MORE ECONOMICAL with PHRASES - use less variety
Two exceptions to lexical entrainment
Court and politics
Where people may refuse to use same terminology
One article
Overhearers don’t seem to have same understanding as people who are part of it
Structure repetition (same as in L10)
You get repetition at every level of conversation
Branigan, Pickering & Cleland (2000) - sentence structure
If confed uses 1 sentence structure, ppt likely to reuse same sentence structure
ROBUST finding that participants do this in dialogue settings
They don’t seem to be aware they are doing it
less clear we use this as a strategy - more an automatic process
Audience design - how far do speakers take partners into account?
Do we change how we’re talking based on our knowledge of the audience?
Isaacs & Clark (1987)
Isaacs & Clark (1987)
In NYC - ppts (native New Yorkers) had to describe pictures to other NYers or “out of towners”
about NYC land marks
Ppts weren’t told which was which but quickly established which category they were in and MODIFIED DESCRIPTIONS OF LANDMARK PHOTOS accordingly
Adjusted how they spoke, taking into account the knowledge of the person they are talking to and adjusting the way they spoke to them
Obvious manipulation
Ferreira et al (2005)
Subtle manipulation
Distinction between linguistic and non-linguistic ambiguity
Do we avoid ambiguity when we’re speaking to someone?
Ferreira et al (2005) - avoiding ambiguity
Subtle manipulation
Distinction between linguistic and non-linguistic ambiguity
Do we avoid ambiguity when we’re speaking to someone?
Picture description task
a) had big bat, little bat, duck - NON-linguistic ambuiguity
b) animal bat, cricket bat, duck - LINGUISTIC ambiguity - two types of bats
c) control trial - only 1 animal bat present
Ferreira et al (2005) results - avoiding ambiguity
Target was always to describe the (animal/small) bat
Participants were better at avoiding non-linguistic than linguistic
More likely to specify little bat than describing animal/flying bat
Not very good at avoiding linguistic ambiguity - didn’t realise it was ambiguous
Once their partner said “but which bat” - ppts noticed and avoided ambiguity
Limits to audience design (Ferreira & Dell, 2000)
The coach knew [that] you missed practice
The vampire knew [that] you hated blood
Inclusion of “that” makes sentence easier to understand - avoids garden path sentence
Do speakers take this into account?
No - even though including “that” helps avoid ambiguity, speakers don’t tend to do this
If inclusion of “that” –> stops sentences taking 2 paths:
The coach knew you.
The coach knew you missed…
Keysar & Henley (Ambiguous sentencing) - when we don’t help our audience
Got people to produce sentences that were ambiguous
Two types of ambiguity
- Syntactically ambiguous: Boris shot the man with the gun (listener could understand the sentence differently to what you mean)
- Lexically ambiguous: Currency is no longer exchanged at the banks (banks could refer to financial banks or river banks)
So do speakers have a good sense of whether the listener understood or not?
46% listeners did NOT have a good insight
Another study (2013) looked at whether our ability to avoid ambiguity when talking to someone is dependent on our emotional state
Role of “just being human”
Induced mood before task
Record speech that would be played back to others
Have to describe where to move objects based on instructions in a picture - a couple of instructions could be given ambiguously - “move donkey above the cat” - two cats in image - which cat? thus ambiguous
- the ambiguous instructions require more description
Happy vs sad speakers - vary in ambiguity?
Happy speakers were less likely to modify object description
- Less good at task
- Happy, carefree? hinders communication
- Produces more ambiguous appearances
- More mistakes
Why are happy people more likely to be ambiguous?
Happy mood - watch bambi skating scene
Sad move - watch Mufasa death scene in Lion King
Possible explanation to why when happy, more ambiguous - worry less about if someone’s going to understand you or not
Paper is agnostic - don’t give explanation why
Interesting result
Audience design conclusion
Sometimes we avoid audience design
Sometimes we don’t
Limits to audience design
Harley book
Interactive-Alignment model (Pickering & Garrod, 2004)
Mechanistic theory of dialogue
- produce a model of how dialogue works
- what are people doing when talking to someone else
Situational model
shared understanding of what their conversation is about
The repetition found multiple levels of speech is what drives this model
Repetition of what in interactive alignment model?
Conceptualisation (semantics) Syntactic repetition (reuse syntax of your partner) Phonological level (put on accent of partner)
Automatic, unconscious process - LINGUISTIC PRIMING
- not aware that we do it
They argue that it’s really important as if you start to express yourself in the same way at different levels - helps you to build up a model of conversation between you that is SIMILAR at all different levels
Similar model of conversation facilitates
Good communication which relies on same understanding as your conversational partner
Evaluation of this IA model
Easily explains extent of repetition in dialogue
and the fact this appears at 3 distinct levels (meaning, word, structure)
BUT assumes little explicit reasoning about person we speak to (AUDIENCE DESIGN) is it complex enough?
Doesn’t say anything about how we reason about someone’s knowledge and adjust accordingly
or modify how we talk depending on who we talk to
Good model of repetition in dialogue
Building blocks to speech
Pickering paper - what is dialogue?
A joint activity, where two people work together on a common task of producing conversation. They rapidly switch between comprehending and producing language
They do it with remarkable ease given the apparent ease - given difficulty of knowing how and when to respond
Psycholinguists may say dialogue is too hard to study - impossible to use experimental control
How can we do anything comparable in dialogue - speaker’s utterances depends on the interlocutor’s behaviour?
What makes dialogue successful?
Success occurs when interlocutors come to have the same understanding of the relevant aspect of the world
Align their mental states
Refer to the same entities and represent the same information about those entities
Assume people construct a situation model that represents salient aspects of the situation under discussions
Dialogue is extremely repetitive says Pickering
which is the principle cause of alignment of situation models
repeat sounds, words, grammar and abstract aspects of meaning
People use the same representations when producing and comprehending language
hear an utterance –> activate acoustic representations –> lexical representations –> retain activations –> more likely to be used –> imitation
Dialogue involves LINGUISTIC PRIMING - such imitation routinely takes place without the person being aware of it (automatic process)
Mechanism of alignment is largely automatic and unconscious
Garrod & Anderson (1987) maze game
Players in a mazer game tend to use the same terms to refer to their positions in a maze (e.g. row/level) and to use same description schemes - co-ordinates/line&row (paths)/matrix grid/pictorial descriptions
Alignment of grammar
People tend to repeat their own grammatical choices - therefore show syntactic priming
Could such priming occur between people in dialogue? - would indicate that comprehension and production share grammatical representations
Testing grammatical representation - Branigan et al (2000)
description card task - participants tended to repeat the confederate’s choice of grammatical form (sentence structure - DO/PO)
Even when debriefing - ppts hadn’t guessed what they were testing for
TENDENCY TO REPEAT LINGUISTIC FORM UNDERLIES ALIGNMENT OF DIALOGUE
Also manipulated whether prime and target used the same verb as each other (give-give) or used different verbs (show-give)
When used same verb, ppts were 55% more likely to use the same grammatical form as their confederate than use opposite form
when used different verb, same grammatical form dropped to 26%
Principle of interactive alignment model
Alignment at one level (in this case lexical alignment) ENHANCES alignment at other levels (grammatical alignment)
Interlocutors choice of sentence form can be affected by factors unrelated to alignment
Tendency to avoid ambiguity (by adding that’s) - small in comparison to a much stronger tendency to utter “that’s” if the addressee had just uttered “that’s”
The tendency to align can overwhelm other factors affecting language production
Grammatical alignment takes place between languages in bilinguals (Spanish & English)
confederate produce Spanish active or passive utterance, and the participant respond in English - ppt tended to use the same grammatical form as the confederate even though language differed
Conclusion of Pickering paper
- Interlocutors align grammatical representations, as well as words and meanings
- Suggests interlocutors have a very strong tendency to align their language, typically without being at all aware they are doing so
- Long way from natural, interactive language (lab tasks)
- BUT they do investigate language as a communicative device .
Conclusion of Pickering paper
- Interlocutors align grammatical representations, as well as words and meanings
- Suggests interlocutors have a very strong tendency to align their language, typically without being at all aware they are doing so
- Long way from natural, interactive language (lab tasks)
- BUT they do investigate language as a communicative device
Speakers do pay some attention to the needs of the listener (Audience design)
By trying to disambiguate their utterances
better at doing it with non-linguistic ambiguity
Sometimes don’t notice linguistic ambiguity to avoid it
Non-linguistic ambiguity
Multiple instances of similar meanings e.g. two apples in front of us, one red and one green - we are unlikely to just say “give me the apple”
BUT only sometimes avoid linguistic ambiguity (ambiguous words e.g. bat)
Ferreira et al (2005) bat study showed
Speakers monitor their speech and sometimes detect and avoid linguistic ambiguity before producing it but almost always avoid non-linguistic ambiguity
Limitations to audience design - Limits to how far a speaker will go to make the listener’s life easier - Ferreira & Dell (2000) examined extent to which ppts used optional complentizers e.g. that (the vampire knew [that] you hated blood
SHOULD use this optional word of “that” to make sure sentence is easy to understand and as unambiguous as possible
but choose structures that are easy to produce
Keysar and Henly (2002)
looked at 40 speakers producing syntactically ambiguous sentences such as “boris shot the man with the gun” and lexically ambiguous sentences such as “the typist tried to read the letter without her glasses” - nearly half (46%) of the time the speaker thought the listeners had correctly understood the sentence when they hadn’t - limits to how much speakers tailor their performance to their listeners, they also do not always do so correctly even when they try