Upload
barnard-jeffry-wells
View
229
Download
6
Tags:
Embed Size (px)
Citation preview
Brief roadmap
Language production research Today:
Models of Language production Levelt’s model Dell’s model Some of the characteristics that distinguish the
models Some of the research to investigate the models
Tuesday: Conversation – Production and Comprehension come together
Nitty-gritty details of the model
Message level
Morphemic level
Syntactic level
Phonemic level
Articulation
A central question: How do we do speak so quickly and fluently?
Are the stages discrete or cascading? Discrete (modular): must complete before
moving on Cascade: can get started as soon as some
information is available Is there feedback?
Top-down only (serial processing) Garrett, Levelt
Bottom up too (interactive processing) Dell, Stemberger, McKay
Two different models
Message
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
TACTIC FRAMES LEXICAL NETWORK
Dell (1986)Levelt (1989)
Levelt’s model Four broad stages:
Conceptualization Deciding on the message (= meaning
to express) Formulation
Turning the message into linguistic representations
Grammatical encoding (finding words and putting them together)
Phonological encoding (finding sounds and putting them together)
Articulation Speaking (or writing or signing)
Monitoring (via the comprehension system)
Message
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
Formalization on the Syntax side of the model
Works in parallel with the lexicon sideMessage
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
Levelt’s model
Functional processing: Assignment of roles
Direct object
Grammatical subject
Formalization on the Syntax side of the model
Works in parallel with the lexicon sideMessage
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
Levelt’s model
Positional processing: Build syntactic tree
NP VP
S
V NP
Tip of tongue state when lemma is retrieved without word-form being retrieved
Message
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
Levelt’s model
Involves lexical retrieval: Semantic/syntactic content
(lemmas) Phonological content
(lexemes or word-forms)
Formalization on the Lexicon side of the model
has stripes is dangerous
TIGER (X)
Fem.
Noun
countabletigre
/tigre/
/t/ /I/ /g/
Lexical concepts
Lemmas
Lexemes
Phonemes
Levelt’s model (see chpt 5, pg 115-117)
has stripes is dangerous
TIGER (X)
Levelt’s model: conceptual level
Conceptual level is not decomposed one lexical concept node for
“tiger” instead, conceptual links from
“tiger” to “stripes”, etc.
Fem.
Noun
tigre
/tigre/
/t/ /I/ /g/
countable
TIGER (X)
Fem.
Noun
tigre
Levelt’s model: meaning & syntax
First, lemma activation occurs This involves activating a lemma or
lemmas corresponding to the concept thus, concept TIGER activates lemma
“tiger”
has stripes is dangerous
/tigre/
/t/ /I/ /g/
countable
TIGER (X)
Fem.
Noun
tigre
Levelt’s model: meaning & syntax
First, lemma activation occurs This involves activating a lemma or
lemmas corresponding to the concept thus, concept TIGER activates lemma
“tiger”
But also involves activating other lemmas
TIGER also activates LION (etc.) to some extent
and LION activates lemma “lion”
LION (X)
lion
/tigre/
/t/ /I/ /g/
has stripes is dangerous
TIGER (X)
Fem.
Noun
tigre
Levelt’s model: meaning & syntax
First, lemma activation occurs Second, lemma selection occurs
LION (X)
lion
Selection is different from activation
Only one lemma is selected Probability of selecting the target
lemma (“tiger”) ratio of that lemma’s activation to
the total activation of all lemmas (“tiger”, “lion”, etc.)
Hence competition between semantically related lemmas
/tigre/
/t/ /I/ /g/
has stripes is dangerous
Morpho-phonological encoding (and beyond)
The lemma is now converted into a phonological representation
called “word-form” (or “lexeme”) If “tiger” lemma plus plural (and
noun) are activated Leads to activation of morphemes
tigre and s Other processes too
Stress, phonological segments, phonetics, and finally articulation/tigre/
/t/ /I/ /g/
has stripes is dangerous
Fem.
Noun countable
tigre
TIGER (X)
Modularity Later processes cannot affect
earlier processes No feedback between the word-form
(lexemes) layer and the grammatical (lemmas) layer
Also, only one lemma activates a word form
If “tiger” and “lion” lemmas are activated, they compete to produce a winner at the lemma stratum
Only the “winner” activates a word form (selection)
The word-forms for the “losers” aren’t accessed
Model’s assumptions
Message
Lexicon
Grammatical
Form
Articulation
FunctionalProcessing
PositionalProcessing
Dell’s interactive account Dell (1986) presented the one of the best-known
interactive accounts other similar accounts exist (e.g., Stemberger, McKay)
Network organization 3 levels of representation
Semantics (decomposed into features) Words and morphemes phonemes (sounds)
These get selected and inserted into frames
Dell (1986)
A moment in the production of:
“Some swimmers sink”
TACTIC FRAMES LEXICAL NETWORK
Dell’s interactive account
as well as “downwards”
information
inform
ation
Interactive because information flows “upwards”
Dell (1986)
Cascading because processing at lower levels can start early
TACTIC FRAMES LEXICAL NETWORK
Dell’s interactive account
these send activation back to the word level, activating words containing these sounds (e.g., “log”, “dot”) to some extent
Dell (1986)
this activation is upwards (phonology to syntax) and wouldn’t occur in Levelt’s account
FURRY BARKS
dog log
/a//g//d/ /l/
MAMMAL e.g., the semantic features mammal, barks, four-legs activate the word “dog”
this activates the sounds /d/, /o/, /g/dot
/t/
Dell’s interactive account
Model comparisons
Levelt’s Dell’s
Similar representations
Frames and slots
Insertion of representations into the frames
Serial
Modular
External monitor(comprehension)
Interactive
Cascaded
Similarities
Differences
Testing Models of language production
Experimental investigations of some of these issues
Time course - cascading vs serial Picture word interference
Separation of syntax and semantics Subject verb agreement
Abstract syntax vs surface form Syntactic priming
tiger
Picture-word interference task Task:
Participants name basic objects as quickly as possible
Distractor words are embedded in the object (or presented aloud)
Participants are instructed to ignore these words
Experimental tests
Semantic interference Meaning related words can
slow down naming the picture
e.g., the word TIGER in a picture of a LION
Experimental tests
tiger
Picture-word interference task
Form-related words can speed up processing
e.g., the word liar in a picture of a LION
liar
Experimental tests Picture-word interference task
Semantic interference
Experiments manipulate timing: picture and word can be presented
simultaneously
liar
time
liar
or one can slightly precede the other We draw inferences about time-course of processing
liar
Experimental tests
SOA (Stimulus onset asynchrony) manipulation -150 ms (word …150 ms … picture) 0 ms (i.e., synchronous presentation) +150 ms (picture …150ms …word)
Schriefers, Meyer, and Levelt (1990) DOT phonologically related CAT semantically related SHIP unrelated word
Evidence against interactivity
Schriefers, Meyer, and Levelt (1990) DOT phonologically related CAT semantically related SHIP unrelated word
EarlyOnly Semantic effects
LateOnly Phonological effects
Evidence against interactivity
Schriefers, Meyer, and Levelt (1990) Also looked for any evidence of a mediated
priming effect
hat dog
DOG (X) CAT (X)
cat
/cat/ /hat/
/t//a//k/ /h/
Found no evidence for it
Evidence against interactivity
Early semantic inhibition Late phonological facilitation Fits with the assumption that semantic processing
precedes phonological processing No overlap
suggests two discrete stages in production an interactive account might find semantic and phonological
effects at the same time
Interpretation
Mixed errors Both semantic and phonological relationship to target word Target = “cat”
semantic error = “dog” phonological error = “hat” mixed error = “rat”
Occur more often than predicted by modular models if you can go wrong at either stage, it would only be by chance
that an error would be mixed
Evidence for interactivity
Dell’s explanation The process of making an error
The semantic features of dog activate “cat” Some features (e.g., animate, mammalian) activate “rat” as well “cat” then activates the sounds /k/, /ae/, /t/ /ae/ and /t/ activate “rat” by feedback This confluence of activation leads to increased tendency for
“rat” to be uttered Also explains the tendency for phonological errors to be real
words (lexical bias effect) Sounds can only feed back to words (non-words not
represented) so only words can feedback to sound level
Evidence for interactivity
A number of recent experimental findings appear to support interaction under some circumstances (or at least cascading models) Damian & Martin (1999) Cutting & Ferreira (1999) Peterson & Savoy (1998)
Evidence for interactivity
Damian and Martin (1999) Picture-Word interference The critical difference:
the addition of a “semantic and phonological” condition
Picture of Apple peach (semantically related) apathy (phonologically related) apricot (sem & phono related) couch (unrelated)
peach
Evidence for interactivity
Damian & Martin (1999)
early semantic inhibition
couch (unrelated)
peach (semantically related)
apathy (phonologically related)
apricot (sem & phono related)
Evidence for interactivity
Damian & Martin (1999)
late phonological facilitation (0 and + 150 ms)
early semantic inhibition
couch (unrelated)
peach (semantically related)
apathy (phonologically related)
apricot (sem & phono related)
Evidence for interactivity
Damian & Martin (1999)
late phonological facilitation (0 and + 150 ms)
Shows overlap, unlike Schriefers et al.
early semantic inhibition
couch (unrelated)
peach (semantically related)
apathy (phonologically related)
apricot (sem & phono related)
Evidence for interactivity
Cutting and Ferreira (1999) Picture-Word interference The critical difference:
Used homophone pictures Related distractors could be to
the depicted meaning or alternative meaning
“game”
“dance”
“hammer” (unrelated)
Only tested -150 SOA
dance
Evidence for interactivity
ball
BALL (X) BALL (X)
ball
/ball/
DANCE (X)
dance
GAME (X)
game
Cascading Prediction: dance ball /ball/
Cutting and Ferreira (1999)
Evidence for interactivity
Early Facilitation from a phonologically mediated distractor
Early semantic inhibition
Cutting and Ferreira (1999)
Evidence of cascading information flow (both semantic and phonological information at early SOA)
Evidence for interactivity
Peterson & Savoy (1998) Slightly different task
Prepare to name the picture
If “?” comes up name it?
Evidence for interactivity
Peterson & Savoy (1998) Slightly different task
Prepare to name the picture
If “?” comes up name it If a word comes up
instead, name the word
liar
Manipulate Word/picture relationship SOA
Evidence for interactivity
Peterson & Savoy (1998) Used pictures with two
synonymous names
Used words that were phonologically related to the non dominant name of the picture
sofa couch
DominantSubordinate
soda
Evidence for interactivity
Peterson & Savoy Found evidence for phonological activation of near
synonyms: Participants slower to say distractor soda than unrelated
distractor when naming couch Soda is related to non-selected sofa
Remember that Levelt et al. assume that only one lemma can be selected and hence activate a phonological form
Levelt et al’s explanation: Could be erroneous selection of two lemmas?
Evidence for interactivity
Can the two-stage account be saved?
Evidence for interaction is hard to reconcile with the Levelt account However, most attempts are likely to revolve
around the monitor Basically, people sometimes notice a problem and
screen it out Levelt argues that evidence for interaction
really involves “special cases”, not directly related to normal processing
Levelt et al.’s theory of word production: Strictly modular lexical access Syntactic processing precedes phonological
processing Dell’s interactive account:
Interaction between syntactic and phonological processing
Experimental evidence is equivocal, but increasing evidence that more than one lemma may activate associated word-form
Overall summary
Conversational interaction“the horse raced past
the barn”
Conversation is more than just two side-by-side monologues.
“the kids swam across the river”
Conversational interaction“The horse raced past
the barn”
Conversation is a specialized form of social interaction, with rules and organization.
“Really? Why would it do that?”
Conversation Herb Clark (1996)
Joint action People acting in coordination with one another
doing the tango driving a car with a pedestrian crossing the street
The participants don’t always do similar things Autonomous actions
Things that you do by yourself Participatory actions
Individual acts only done as parts of joint actions
Conversation Herb Clark (1996)
Speaking and listening Traditionally treated as autonomous actions
Contributing to the tradition of studying language comprehension and production separately
Clark proposed that they should be treated as participatory actions
Conversation Herb Clark (1996)
Speaking and listening Component actions in production and
comprehension come in pairs
Speaking Listening A vocalizes sounds for B
A formalizes utterances for B
A means something for B
B attends to A’s vocalizations
B identifies A’s utterances
B understands A’s meaning
The actions of one participant depend on the actions of the other
Conversation Herb Clark (1996)
Face-to-face conversation - the basic setting Features
Co-presence Visibility Audibility
Instantaneity
Evanescence Recordlessness Simultaneity
Extemporaneity Self-determination Self-expression
Immediacy Medium Control
Other settings may lack some of these features e.g., telephone conversations take away co-presence and
visibility, which may change language use
Meaning and understanding ABBOTT: Super Duper computer store. Can I help you? COSTELLO: Thanks. I'm setting up an office in my den, and I'm thinking about buying a
computer. ABBOTT: Mac? COSTELLO: No, the name is Lou. ABBOTT: Your computer? COSTELLO: I don't own a computer. I want to buy one. ABBOTT: Mac? COSTELLO: I told you, my name is Lou. ABBOTT: What about Windows? COSTELLO: Why? Will it get stuffy in here? ABBOTT: Do you want a computer with windows? COSTELLO: I don't know. What will I see when I look in the windows? ABBOTT: Wallpaper. COSTELLO: Never mind the windows. I need a computer and software. ABBOTT: Software for windows? COSTELLO: No. On the computer! I need something I can use to write proposals, track
expenses and run my business. What have you got? ABBOTT: Office.
Meaning and understanding COSTELLO: Yeah, for my office. Can you recommend anything? ABBOTT: I just did. COSTELLO: You just did what? ABBOTT: Recommend something. COSTELLO: You recommended something? ABBOTT: Yes. COSTELLO: For my office? ABBOTT: Yes. COSTELLO: OK, what did you recommend for my office? ABBOTT: Office. COSTELLO: Yes, for my office! ABBOTT: I recommend office with windows. COSTELLO: I already have an office and it has windows!OK, lets just say, I'm sitting at
my computer and I want to type a proposal. What do I need? ABBOTT: Word. COSTELLO: What word? ABBOTT: Word in Office. COSTELLO: The only word in office is office. ABBOTT: The Word in Office for Windows.
Meaning and understanding COSTELLO: Which word in office for windows? ABBOTT: The Word you get when you click the blue "W.” COSTELLO: I'm going to click your blue "w" if you don't start with some straight
answers. OK, forget that. Can I watch movies on the Internet? ABBOTT: Yes, you want Real One. COSTELLO: Maybe a real one, maybe a cartoon. What I watch is none of your
business. Just tell me what I need! ABBOTT: Real One. COSTELLO: If it ユ s a long movie I also want to see reel 2, 3 and 4. Can I watch them? ABBOTT: Of course. COSTELLO: Great, with what? ABBOTT: Real One. COSTELLO; OK, I'm at my computer and I want to watch a movie.What do I do? ABBOTT: You click the blue "1.” COSTELLO: I click the blue one what? ABBOTT: The blue "1.” COSTELLO: Is that different from the blue "W"? ABBOTT: The blue 1 is Real One and the blue W is Word. COSTELLO: What word?
Meaning and understanding ABBOTT: The Word in Office for Windows. COSTELLO: But there are three words in "office for windows"! ABBOTT: No, just one. But it ユ s the most popular Word in the world. COSTELLO: It is? ABBOTT: Yes, but to be fair, there aren't many other Words left. It pretty much wiped out all the other
Words. COSTELLO: And that word is real one? ABBOTT: Real One has nothing to do with Word. Real One isn't even Part of Office. COSTELLO: Stop! Don't start that again. What about financial bookkeeping you have anything I can
track my money with? ABBOTT: Money. COSTELLO: That's right. What do you have? ABBOTT: Money. COSTELLO: I need money to track my money? ABBOTT: It comes bundled with your computer. COSTELLO: What's bundled to my computer? ABBOTT: Money.
Meaning and understanding COSTELLO: Money comes with my computer? ABBOTT: Yes. No extra charge. COSTELLO: I get a bundle of money with my computer? How much? ABBOTT: One copy. COSTELLO: Isn't it illegal to copy money? ABBOTT: Microsoft gave us a license to copy money. COSTELLO: They can give you a license to copy money? ABBOTT: Why not? THEY OWN IT!
(LATER) COSTELLO: How do I turn my computer off?? ABBOTT: Click on "START".
Meaning and understanding Common ground
Knowledge, beliefs and suppositions that the participants believe that they share
Members of cultural communities Shared experiences What has taken place already in the conversation
Common ground is necessary to coordinate speaker’s meaning with listener’s understanding
Conversations are purposive and unplanned Typically you can’t plan exactly what you’re going
to say because it depends on another participant Conversations look planned only in retrospect
Conversations have a fairly stable structure
Structure of a conversation
Joe: (places a phone call) Kevin: Miss Pink’s office - hello Joe: hello, is Miss Pink in Kevin: well, she’s in, but she’s engaged at
the moment, who is it? Joe: Oh it’s Professors Worth’s secretary,
from Pan-American college Kevin: m, Joe: Could you give her a message “for
me” Kevin: “certainly” Joe: u’m Professor Worth said that, if Miss
Pink runs into difficulties, .. On Monday afternoon, .. With the standing subcommittee, .. Over the item on Miss Panoff, …
Structure of a conversation Kevin: Miss Panoff? Joe: Yes, that Professor Worth would be
with Mr Miles all afternoon, .. So she only had to go round and collect him if she needed him, …
Kevin: ah, … thank you very much indeed, Joe: right Kevin: Panoff, right “you” are Joe: right Kevin: I’ll tell her, Joe: thank you Kevin: bye bye Joe: bye
Structure of a conversation Action sequences: smaller joint projects to fulfill a
goal Adjacency pairs
Opening the conversation Kevin: Miss Pink’s office - hello Joe: hello, ..
Exchanging information about Pink Joe:.., is Miss Pink in Kevin: well, she’s in, but she’s engaged at the moment…
Structure of a conversation Action sequences: smaller joint projects to fulfill a
goal Adjacency pairs
Exchanging the message from Worth Joe: u’m Professor Worth said that, if Miss Pink runs into
difficulties, .. On Monday afternoon, .. With the standing subcommittee, .. Over the item on Miss Panoff, …
Closing the conversation Kevin: I’ll tell her, Joe: thank you Kevin: bye bye Joe: bye
Opening conversations Need to pick who starts
Turn taking is typically not decided upon in advance Potentially a lot of ways to open, but we typically restrict
our openings to a few ways Address another Request information Offer information Use a stereotyped expression or topic
Opening conversations Has to resolve:
The entry time Is now the time to converse?
The participants Who is talking to whom?
Their roles What is level of participation in the conversation?
The official business What is the conversation about?
EavesdropperAll listeners
Identifying participants Conversation often takes place in situations
that involve various types of participants and non-participants
BystanderSide
participantsAll participants
Speaker Addressee
Taking turns Typically conversations don’t involve two (or
more) people talking at the same time
Individual styles of turn-taking vary widely Length of a turn is a fairly stable
characteristic within a given individual’s conversational interactions
Standard signals indicate a change in turn: a head nod, a glance, a questioning tone
Taking turns Typically conversations don’t involve two (or
more) people talking at the same time Three implicit rules (Sacks et al, 1974)
Rule 1: Current speakers selects next speaker Rule 2: Self-selection: if rule 1 isn’t used, then next
speaker can select themselves Rule 3: current speaker may continue (or not)
These principles are ordered in terms of priority The first is the most important, and the last is the least
important Just try violating them in an actual conversation (but
debrief later!)
Taking turns Typically conversations don’t involve two (or more)
people talking at the same time
Use of non-verbal cues Drop of pitch Drawl on final syllable Termination of hand signals Drop in loudness Completion of a grammatical clause Use of stereotyped phrase
“you know”
Negotiating topics Keep the discourse relevant to the topic
(remember Grice’s maxims) Coherence again
Earlier we looked at coherence within a speaker, now we consider it across multiple speakers
Must use statements to signal topic shifts
Closing conversations Closing statements
Must exit from the last topic, mutually agree to close the conversation, and coordinate the disengagement
signal the end of conversation (or topic) “okay”
Justifying why conversation should end “I gotta go”
Reference to potential future conversation “later dude”
Summary “People use language for doing things with
each other, and their use of language is itself a joint action.” Clark (1996, pg387) Conversation is structured
But, that structure depends on more than one individual Models of language use (production and
comprehension) need to be developed within this perspective