“Where Was I?”A Psycholinguistic Investigation of Interruptions to Language Production : “Where Was I?” A Psycholinguistic Investigation of Interruptions to Language Production Benjamin Swets—Stony Brook University
Fernanda Ferreira—University of Edinburgh
Erik Altmann—Michigan State University
Slide2: File
papers Time
Slide3: File
papers Answer
phone Time Interruption
Slide4: File
papers Answer
phone Hang
up Time Interruption
Slide5: File
papers Answer
phone Hang
up Resume
filing Time Interruption ?
Slide6: File
papers Answer
phone Hang
up Resume
filing Time Interruption 25 minutes “Meet the Life Hackers”
New York Times
Clive Thompson
10/16/06
Conversational Interruptions: Conversational Interruptions
Intuitively disruptive
Bad manners
Example dialog:
Speaker 1 narrates scene from Lord of the Rings to Speaker 2
Conversational Interruptions: Conversational Interruptions
1: One of the hobbits’ names is Merry and he tells him to climb up the tree and they start climbing up the tree and they think that they’ve lost him again and he grabs the other one’s leg and pulls him down—
2: Do you hear that in the hallway?
1: No—
2: That buzzing noise?
1: No.
2: Sorry, that’s just annoying.
1: (Laughter)
2: Okay, go ahead.
1: Um…so…okay now I’ve…lost what I was—what was going on—
2: I’m sorry.
1: That’s o—no, that’s okay. Um…okay, so the monster is…pulls the one off of a tree, and…
Overview: Overview
When are conversational interruptions disruptive to speakers?
How can this knowledge be used?
When Do Interruptions Disrupt?: When Do Interruptions Disrupt?
Slide11: Begin
Primary
Task Begin
Secondary
Task End
Secondary
Task Resume
Primary
Task Interruption Time
Slide12: File
papers Answer
phone Hang
up Resume
filing Time Interruption
Slide13: Begin
Primary
Task Begin
Secondary
Task End
Secondary
Task Resume
Primary
Task Interruption Time Resumption
Lag
When Do Interruptions Disrupt?: When Do Interruptions Disrupt?
When there is no option to rehearse one’s location in the primary task before interruption begins (Altmann & Trafton, 2002)
Suggests that goal memory is vital for resumption
When interruptions come at fine break points rather than coarse break points (Adamczyk & Bailey, 2004).
Movie titling task
When Do Interruptions Disrupt?: When Do Interruptions Disrupt?
When there is no option to rehearse one’s location in the primary task before interruption begins (Altmann & Trafton, 2002)
Suggests that goal memory is vital for resumption
When interruptions come at fine break points rather than coarse break points (Adamczyk & Bailey, 2004)
Movie titling task
Psycholinguistic Issues: Psycholinguistic Issues
Levelt (1989): Bookkeeping
How do we keep track of where we are in a discourse?
Goal memory of language production
Recent theories of dialog (Pickering & Garrod, 2004)
Predictions can be generated about how interruptions will be handled in dialog
Interruptions with different goals should be disruptive
Definitions: Present Study: Definitions: Present Study
Interruption
A temporary event that causes a speaker to cease speech, and which allows speech to resume afterward.
Disruption
Operationalized by several variables measuring resumption difficulty:
Resumption lag
Resumption errors
Disfluencies
Overview of Studies: Overview of Studies
Study 1: Dialog
What disrupts bookkeeping in natural conversation?
Study 2: Sentence production
What disrupts bookkeeping when producing single sentences in controlled settings?
Study 1: Study 1
Research question
Does the goal of an interruption influence whether it is disruptive?
Method : Method Participant would…
Be introduced to confederate “participant”
Watch 2 film clips in a row (about 3 minutes each)
Describe film clips to confederate for a memory test (cover story)
Method (cont): Method (cont) Confederate would…
Interrupt 4 times per description (8 times total)
Interrupt at predictable points in each description…
with 4 different types of interruptions (2 aligning, 2 non-aligning)
Aligning Interruption Types: Aligning Interruption Types
Repetition interruptions
Summarize last few points to make sure she understood
Example: “Ok, wait, I’m sorry, let me make sure I got this right first: so a guy walks into a room…”
Clarification interruptions
Confederate asks speaker to clarify unmentioned detail
Example: “Wait, sorry—so what kind of basket was it? Was it like a flower basket or a, like, bigger basket?”
Non-Aligning Interruption Types: Non-Aligning Interruption Types
Digressions
Use just-mentioned point to introduce to unhelpful topic
Example: “Is this a horror movie? I hate horror movies—they’re so cheesy.”
True interruptions
Unrelated and unhelpful topic
Example: “I’m sorry—do you hear that noise in the hallway? It’s really annoying.”
Non-Aligning Interruption Types: Non-Aligning Interruption Types
Digressions
Use just-mentioned point to introduce to unhelpful topic
Example: “Is this a horror movie? I hate horror movies—they’re so cheesy.”
True interruptions
Unrelated and unhelpful topic
Example: “I’m sorry—do you hear that noise in the hallway? It’s really annoying.”
Dependent Measures : Dependent Measures
Resumption lag (RL1 and RL2)
Slide26: Describe
scene “Do you
hear that?” “Sorry,
go on.” 1st
content
word RL1 Interruption Time
Slide27: Describe
scene “Do you
hear that?” “Sorry,
go on.” Interruption Time 1st
continuing
word RL2
Dependent Measures: Dependent Measures
1: One of the hobbits’ names is Merry and he tells him to climb up the tree and they start climbing up the tree and they think that they’ve lost him again and he grabs the other one’s leg and pulls him down—
2: Do you hear that in the hallway?
1: No—
2: That buzzing noise?
1: No.
2: Sorry, that’s just annoying.
1: (Laughter)
2: Okay, go ahead.
1: Um…so…okay now I’ve…lost what I was—what was going on—
2: I’m sorry.
1: That’s o—no, that’s okay. Um…okay, so the monster is…pulls the one off of a tree, and… Resumption Lag 1
Dependent Measures: Dependent Measures
1: One of the hobbits’ names is Merry and he tells him to climb up the tree and they start climbing up the tree and they think that they’ve lost him again and he grabs the other one’s leg and pulls him down—
2: Do you hear that in the hallway?
1: No—
2: That buzzing noise?
1: No.
2: Sorry, that’s just annoying.
1: (Laughter)
2: Okay, go ahead.
1: Um…so…okay now I’ve…lost what I was—what was going on—
2: I’m sorry.
1: That’s o—no, that’s okay. Um…okay, so the monster is…pulls the one off of a tree, and… Resumption Lag 2
Dependent Measures: Dependent Measures
1: One of the hobbits’ names is Merry and he tells him to climb up the tree and they start climbing up the tree and they think that they’ve lost him again and he grabs the other one’s leg and pulls him down—
2: Do you hear that in the hallway?
1: No—
2: That buzzing noise?
1: No.
2: Sorry, that’s just annoying.
1: (Laughter)
2: Okay, go ahead.
1: Um…so…okay now I’ve…lost what I was—what was going on—
2: I’m sorry.
1: That’s o—no, that’s okay. Um…okay, so the monster is…pulls the one off of a tree, and… Repeated Words
Dependent Measures: Dependent Measures
1: One of the hobbits’ names is Merry and he tells him to climb up the tree and they start climbing up the tree and they think that they’ve lost him again and he grabs the other one’s leg and pulls him down—
2: Do you hear that in the hallway?
1: No—
2: That buzzing noise?
1: No.
2: Sorry, that’s just annoying.
1: (Laughter)
2: Okay, go ahead.
1: Um…so…okay now I’ve…lost what I was—what was going on—
2: I’m sorry.
1: That’s o—no, that’s okay. Um…okay, so the monster is…pulls the one off of a tree, and… Disfluencies
Prediction: Prediction
Aligning types less disruptive than non-aligning types
Alignment in dialog (Pickering & Garrod, 2004)
Bookkeeping as goal memory (Altmann & Trafton, 2002; Grosz & Sidner, 1986; Levelt, 1989)
Slide33: Graph “something like this”
Summary of Results: Study 1 : Summary of Results: Study 1
More disruptive types:
clarifications (aligning)
true interruptions (non-aligning)
Less disruptive types:
repetition interruptions (aligning)
digressions (non-aligning)
Implications: Implications
Goal alignment is not the only factor driving disruption to bookkeeping
Other factors:
Interference: need to respond verbally (clarifications)
Shallow processing (digressions)
Study 2: Study 2
More traditional psycholinguistic approach
What disrupts sentence-level bookkeeping?
At what point in sentence production is it most costly to be interrupted?
Study 2: Paradigm: Study 2: Paradigm Participant described arrays of objects
Two-clause target sentences
“The ear moves | above the book | and the shoe moves | below the book.”
Slide42: Study 2: Paradigm Subject began describing array
Slide43: Study 2: Paradigm Experimenter pressed a button to initiate interruption
Slide44: Study 2: Paradigm Speaker: stop speaking, perform task
Interruption Tasks: Interruption Tasks
Grammaticality judgment
Is the sentence grammatical?
Example: “The whole team lost the coach’s money himself.”
Arithmetic
Is the problem solved correctly?
Example: 23 + 35 = 58
Time limit: 3.25 seconds
Slide46: Study 2: Paradigm After 3.25 seconds, subject would hear another BEEP.
Slide47: Study 2: Paradigm Speakers instructed to resume from exactly where they left off
Manipulations: Manipulations
Interruption type
Location of interruption
1st clause, between clauses, 2nd clause
The ear moves | above the book | and the shoe moves | below the book
Slide49: “The ear
moves…” Interrupt
BEEP Resumption
BEEP “…above
the book…” Resumption
Lag Interruption
Task Time Study 2: Measures RL: Latency to first vocal response after interruption
Slide50: “The ear
moves…” Interrupt
BEEP Resumption
BEEP “…above
the book…” Resumption
Lag Interruption
Task Time Study 2: Measures Resumption errors
Slide51: “The ear
moves…” Interrupt
BEEP Resumption
BEEP “…moves
above…” Resumption
Lag Interruption
Task Time Study 2: Measures Repeated words
Slide52: “The ear
moves…” Interrupt
BEEP Resumption
BEEP “…and the
shoe moves …” Resumption
Lag Interruption
Task Time Study 2: Measures Skipped words
Hypotheses and Predictions: Hypotheses and Predictions
Locality predictions:
Temporal bookkeeping: Early vs. late interruptions
Size of planning load: Early more disruptive than late
Structural bookkeeping: Between- vs. within-clause
Coarse vs. fine break points
Study 2: Summary of Results: Study 2: Summary of Results
Influence of locus on recovery:
Early interruptions resulted in longer resumption lags.
Between-clause interruptions resulted in fewer resumption errors.
Implications: Implications
Two kinds of disruption
Resumption lag increases due to reloading larger plans
Speakers are planning a clause in advance
Resumption errors measure disrupted bookkeeping
Unit of bookkeeping is the clause
How to Interrupt: How to Interrupt
In socially responsible ways:
Make shallow and pointless interruptions
Interrupt late in a sentence, or between clauses
How to Interrupt: How to Interrupt
In socially irresponsible ways:
Interrupt at cross purposes
Make interlocutor address interruption in detail
Interrupt early, before clause boundary
Thank you: Thank you
Future Directions: Future Directions
Reduced interference from non-verbal interruptions?
Gestures as interruption warnings—helpful to speakers at resumption?
Prosody of resumptions (Very robust!)
Focus on global measures—are interruptions sometimes helpful?
Untangling goal structure
Priming investigations
What is bookkeeping?: What is bookkeeping?
What is the nature of the bookkeeping mechanism?
Is it a domain-general memory mechanism?
Or a domain-specific language production mechanism?
Similarity hypothesis (Gillie & Broadbent, 1989)
Verbal interruptions should be more disruptive than non-verbal interruptions
Domain-specific verbal interference
Preliminary analyses: Grammaticality judgments more disruptive than spatial judgments
Interruption lengths Study 1: Interruption lengths Study 1
Awareness: Awareness
29% fully unaware of purpose/unusual behavior of confederate
44% aware that other participant was behaving strangely/interrupting a lot
27% figured out the purpose of the experiment
Awareness did not interact with interruption type
Practice effects: Practice effects
Some evidence that practice reduced effects in Study 1
Missing data
Better preparation?
Practice reduced RL and errors in Study 2, but did not interact with IVs.