psyc 363 cognitive psychology al cl karlo clark-foosacfoos/courses/363/11-12... · psyc 363...
TRANSCRIPT
Psyc 363Psyc 363Cognitive Psychology
A l Cl k FArlo Clark-Foos
Part XI-XIIPart XI XIIPsycholinguistics
PsycholinguisticsPsycholinguistics
• What is the nature of language?What is the nature of language?– Humans often defined by use of language– What makes it so special?
• Productivity and Regularityy g y– Grammars attempt account for both using rules
• Syntax (Violation: “The girls hits the boys”• Semantics (Violation: “Colorless green ideas sleep furiously”• Phonology (Mispronunciations)
PsycholinguisticsPsycholinguistics• Culture makes language
h d d !hard to study!– Slobin: Daddy Example
• English• EnglishGermanTurkish
Slobin (1982) on Canonical Structure [schema][ ]
English: DADDY THREW THE BALLDADDY THREW THE BALL agent action object [focus] [past] [definite]
German: VATER WARF DEN BALL agent action object [focus] [past] [definite][focus] [past] [definite]
[3rd person] [singular][masculine]
[object]
Turkish: TOP-- U BABA– M AT-- TI object agent action
[d fi it ] [ d [ t][definite] [possessed [past][object] by speaker] [3rd person]
[witnessed by speaker]Very odd to us
Competence vs. PerformanceCompetence vs. Performance
• Two components of languageTwo components of language (another dichotomy…)– Competence vs Performance– Competence vs. Performance
• Linguists (structure)Linguists (structure)• Psycholinguists (process and
function)
Structure of LanguageStructure of LanguageDISCOURSE (constituents)
SENTENCES (constituents)
PHRASES (constituents)
WORDS (RE-WORK-ED)
MEANING (morphemes)
SOUND (phonemes)
Understanding LanguageUnderstanding Language
• Three regularities of language
– Grammar
W it th ’ th– Meaning
– Sound
Wait, there’s another one of these!
Sound
• No grand theory to explain it all• No grand theory to explain it all– But there are some small ones to explain
parts…
Understanding LanguageUnderstanding Language
• Pragmatics
– Combination of regularities
– What is the aim?
– Example: Indirect speech acts
Understanding LanguageUnderstanding Language
• Phrase Structure Grammar (PSG)Phrase Structure Grammar (PSG)
R lh t’ t– Rummelhart’s story– Productivity– Knowledge of rules
– Boomer (1965): Length of pauses in speech
Phase Structure GrammarPhase Structure Grammar
They are washing machines. They are washing machines.
h h hThey are washing machines.
I forgot how good beer tastes. I forgot how good beer tastes.I forgot how good beer tastes.
A woman without her man is nothing. A woman without her man is nothing. A woman without her man is nothing.
Phase Structure GrammarPhase Structure Grammar
They are washing machines. They are washing machines. They are washing machinesThey are washing machines.
Garden Path Sentences(MacKay & Osgood, 1959)
• The cotton shirts are made from comes from Arizona.
• The old man the boats.
• The horse raced past the barn felle o se aced past t e ba e
Evidence for PSGEvidence for PSG
• Where do speech errors happen?Where do speech errors happen?
• Phrase boundaries
• Reaction time to sentences
Caplan (1972)Caplan (1972)
Now that artists are working fewer hours oil prints are rare.
Now that artists are working in oil prints are rare.
(…fewer hours) (oil prints…) VS.
(…in oil) (prints are…)
Indicates clause boundary at which chunk and purge operation takes place
McDermott & Chan (2006)McDermott & Chan (2006)• Study: Read sentences
“The children’s snowman vanished when the temperatures reached 80.”
• Test: Fill-in sentences exactly as seen in study y y“The children’s snowman ______ when the temperatures
reached 80.”
• Results• Results– Errors on 1/3 of sentences
• Error example: “vanished” became “melted”
• Conclusion– “Pragmatic inference”
• Expectations change memory; info implied by sentencep g y; p y• Dump verbatim, retain gist
Intent in LanguageIntent in Language
• Structure and functionStructure and function– Goals
Grice’s Implicatures– Grice s Implicatures• Sentence meaning vs. Speaker’s meaning
Intent in LanguageIntent in Language
• Austin’s (1962) ideasAustin s (1962) ideas– Two uses of language: Descriptive & Performative
(Purposeful)(Purposeful)
John SearleJohn Searle
Locutionary Illocutionary PerlocutionaryyLiteral Meaning
yImplied Meaning Illocutionary Effect
The sky is gray(it must be a normal Michigan Winter day)( g y)
Intent (Function) in LanguageIntent (Function) in Language
• Grice’s CooperativeGrice s Cooperative Principle
– Definition
– Purpose
– 4 maxims
Grice’s MaximsGrice s Maxims
1. Quantity: Say as much as is Q y yrequired: No more, no less.
2. Quality: Say only what you have reason to believe is true.
Geniuses
Grice’s MaximsGrice s Maxims3. Relation: Be relevant.
4. Manner: Avoid obscurity and ambiguity. Be brief. Be orderly.
Clark (1979)Clark (1979)
• 2 kinds of requestsq– Similar to Grice’s
implicaturesp
“Do you know what time it is?”Do you know what time it is?• Direct• Indirect
Clark (1979)Clark (1979)“There are at least six sources of information listeners use in judging whether
the literal meaning was intended seriously or pro forma, as well as whetherthe literal meaning was intended seriously or pro forma, as well as whether there was intended to be any indirect meaning.” (p. 1, Clark, 1979)
Clark (1979)Clark (1979)
C ti l R tConventional Request:
Could you tell me the price for a fifth of Jim Beam?Yes, $5.79 15% ← answers direct + indirect$5.79 77% ← answers indirect only
Novel Request:
Do you have a price for a fifth of Jim Beam?Yes 2% direct + Yes, $5.79 37% indirect$5.79 49% ← indirect only
d li ll dBased on N = 200 liquor stores called
Clark (1979)Clark (1979)
Caller might care (might be formal):Caller might care (might be formal):
I want to buy some bourbon. Does a fifth of Jim Beam cost more than $5?
Yes 28%Yes 28%Yes, $5.79 20%$5.79 48%
C ll h ld ( b bl f l)Caller should care (probably formal):
I’ve got $5 to spend. Does a 5th of Jim Beam cost more than $5?Yes 32%Y $5 79 34%Yes, $5.79 34%$5.79 32%
Based on N = 100 liquor stores called
Searle’s Literal-First ViewSearle s Literal First View
Can you open the window?
STEPS: 1. Compute the literal meaning. 2. Decide if literal meaning is appropriate. 3. If not, compute indirect meaning using Gricean Maxims, etc.
Predictions? What about RTs?
Searle’s Literal-First ViewSearle s Literal First View
• Back to indirect requests (Clark & LucyBack to indirect requests (Clark & Lucy, 1975)– Double negatives are more complicated.
RTs to “Does the sentence satisfy the request?”– RTs to Does the sentence satisfy the request?
“I’ll be very happy if you open the door.”“I’ll be very sad unless you open the door.”
– Can we ever process indirect meaning faster than direct meaning?
Gibbs (1983)Gibbs (1983)Group 1: Controlp
1. “Can’t you be friendly?”2. Verify paraphrases of
• LiteralLiteral• Indirect requests
Group 2: Experimental1. Context / Story2 “C ’t b f i dl ?”2. “Can’t you be friendly?”3. Verify paraphrases of
• Literal• Indirect requests
Gibbs (1983)Gibbs (1983)• Indirect Story: Mrs. Norman was watching her kids play in the back yard. One
of the neighbor’s children came over to play. But Mrs. Norman’s kids refused to share their toys. She angrily walked outside and said to one of the children, “Can’t you be friendly?”
• Literal Story: Rod was talking to his psychiatrist. He was having lots of problems establishing relationships. “Everyone I meet I seem to alienate,” Rod said. “I just turn hostile for no reason,” he continued. The shrink said, “Can’t j , ,you be friendly?”
Indirect Target: “Please be friendly to other people.”Literal Target: “Are you able to act friendly.”Unrelated Target: “Running is excellent for the heart.”
Syntactic ParsingSyntactic Parsing
• Sorry SearleSorry Searle…– Back to PSG’s
Verbatim and gist– Verbatim and gist– 3 syntactic universals
• Grouping Word Order & Agreement• Grouping, Word Order, & Agreement
Syntactic Universals1. Grouping:
Syntactic Universalsp g
[Mary] [drove] [her 1959 Mustang] [to New York City]
2 W d O d2. Word Order:
Three dogs ADJ. Noun[The old man] [who was my uncle][The old man] [who was my uncle]
3. Agreementg[Three dogs] [on my street] [are vicious]
Syntactic ParsingSyntactic Parsing
• Basic syntactic unitBasic syntactic unit
• Caplan “oil” experiment (review)Caplan oil experiment (review)
• What about physical/visual clues?• What about physical/visual clues?– Line breaks in language?
Syntactic ParsingSyntactic Parsing
“During World War II,
even fantastic schemes
“During World War
II, even fantastic
received consideration
if they gave promise
f h t i th fli t ”
schemes received
consideration if they gave
i f h t i th fli t ”of shortening the conflict.” promise of shortening the conflict.”
Graf & Torrey (1966)
Kimball’s 3 Syntactic RulesKimball s 3 Syntactic Rules
1. Functors1. FunctorsWhat do they help us do?
2. Right-Most RuleVerbatim and gist
3. AdjacencyWhat makes it difficult to understand?
FunctorsFunctors
A, AN, THE A, AN, THE NPNP
Some, all, many, etc.Some, all, many, etc. NPNP
Auxiliaries (is, are, can, were)Auxiliaries (is, are, can, were) VPVP
OfOfOf, at, in, etc.Of, at, in, etc. PPPP
AdjacencyAdjacency
h f d d k h• John figured Mary wanted to take the train to NYC out.
John figured out Mary wanted….
• Look the address in the NYC telephone book up.
Look up the address in the NYCLook up the address in the NYC….
Semantic ParsingSemantic Parsing
• Another way to identify constituent structure?Another way to identify constituent structure?
– Hints in languageHints in language– Disambiguation
– Relationship to syntactic parsing?
Semantic Parsing: Semantic Parsing: MarotsosMarotsosgg
Age Competence Reason
3 BAD3 BAD
3 - 3 ½ GOOD Rote Understanding
3 ½ - 4 BAD Too broad a generalization
4 ½ GOOD Big conceptual change/advance
(Children act out passive sentences with toy animals)
Syntactic and Semantic ParsingSyntactic and Semantic ParsingHow do these two operate in relation to each other?
Autonomous: At constituent boundary info. from the two systemsAutonomous: At constituent boundary info. from the two systems converges on Central Processor (no red line – no connection b/w two systems)
I t ti R d li ti O l i thi i d th t tInteractive: Red line connection. Only in this view do the two systems interact and share info (i.e., they are interdependent)
Autonomous vs. InteractiveAutonomous vs. Interactive
• Is semantic information used before reaching the clause f f gboundary?
If you walk near the runway, landing planes….(N.P.)
If you’ve been trained as a pilot, landing planes….(V P )(V.P.)
Next word? IS or ARENext word? IS or ARE
Marslen-Wilson & Tyler (1987)
Next word?Next word?
Marslen Wilson & Tyler (1987)
Next word?Next word?
Autonomous vs. InteractiveAutonomous vs. Interactive
• Swinney’s (1979) worky ( )
– The role of context (i.e., bugs)
– Sentence and then surprise LDT:ANT (C ll R l d)• ANT (Contextually Related)
• SPY (Contextually Inappropriate)• SEW (Unrelated)
– Conclusion?
Marslen-Wilson & TylerMarslen Wilson & Tyler
1 NORMAL Th h h b k i t l t i ht S thi t l t f th l d1. NORMAL: The church was broken into last night. Some thieves stole most of the leadoff the roof.
2. SYNTAX ONLY: The power was located into great water. No buns puzzle some on the l d ff th t tlead off the text.
3. RANDOM: Into was power water the great located. Some the no puzzle buns in leadtext the off.
Recognition Time: NORMAL < SYNTAX < RANDOM
PragmaticsPragmatics1. Fillenbaum’s Perverse Sentences
D ’t i t th t I ’t !Don’t print that or I won’t sue you! People Paraphrase:
If you print that, I’ll sue you.∴ Not syntax not semantics but pragmatics∴ Not syntax, not semantics, but pragmatics
2. Wason’s Uninterpretable Sentences.No head injury is too trivial to be ignored.No head injury is too trivial to be ignored. People paraphrase:
One should always treat head injuries, no matter how small.
∴ Not just semantics and syntax, pragmatics matters
Text ComprehensionText Comprehension
• Why might this be more important?Why might this be more important?
i• Two inputs– Reader & Text
– Knowledge, goals, and expectations
Kintsch’s Model of Text ComprehensionKintsch s Model of Text Comprehension
• comprehension easiest if we can relate new information to ppropositions in working memory
• comprehension slower if must retrieve related concepts from long-term memory (reinstatement search)
• comprehension slower still if reinstatement search fails; must make inference as to how new information relates to previous information
Kintsch’s Model of Text ComprehensionKintsch s Model of Text Comprehension
• Reinstatement Searches– LTM and STM
– What effect do they have?
• Inferences• Inferences– When do they occur?
– What effect do they have?
Text ComprehensionText Comprehension
• Britton et al. (1990) study
– Historical passages with or without “Harmony”
– What hurt performance?p
SummarySummary
• Competence vs. • ProcesspPerformance
• Structuree.g., Parsing Strategies,
Autonomy vs. Interaction
• Pragmaticse.g., Garden Path, Constituents
• Functione g Implicatures Maxims
Pragmatics• Text comprehension
e.g., Reinstatement searches, e.g., Implicatures, Maxims, Indirect Requests inference, harmony