Scalar implicatures - a view from processing Judith Degen - - PowerPoint PPT Presentation

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Scalar implicatures - a view from processing

Judith Degen

University of Rochester

September 18, 2009

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

1

Introduction

2

The psycholinguistic debate Recent past Present

3

Putting scalar implicatures in the grammar

4

A model of scalar implicature processing

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Scalar implicatures

(1) Peter: Did all of your guests stay until midnight? Mary: Some of them did. It’s not the case that all of them did. Scale: all, some (2) Peter: Who is in that room? Mary: John or Bill. It’s not the case that both John and Bill are. Scale: and, or

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

The (neo-)Gricean explanation

Grice’s conversational maxims: Quantity-1: Make your contribution as informative as is required (for the current purpose of the exchange). Truthfulness: Do not say what you believe to be false. Hearer’s reasoning about speaker S: S uttered the statement with some instead of all, which would have also been relevant the all statement entails the some statement if S knew that the all statement holds, she would have uttered it S is well-informed thus, it is not the case that the all statement holds

[Grice (1975), Horn (1984), Levinson (2000)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Cancelability

(3) Explicit

• a. Some of the guests stayed until midnight. In fact, they

all did. # Some, but not all of the guests stayed until midnight. (4) Implicit

• a. If some of the guests stayed until midnight, it must

have been a good party. # If they all stayed, it wasn’t.

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

The Relevance Theory explanation

“post-Gricean” no more maxims trade-off between cognitive effects and processing effort the implicature is computed only if the interpretation arrived at via the basic meaning of the scalar term does not satisfy the hearer’s expectations of relevance

[Sperber and Wilson (1995), Carston (1998)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Relevance Theory vs. Neo-Griceanism

Framing the empirical question Are scalar implicatures computed by default or as part of an effortful, context-driven process? Default model Pragmatic meaning (SI) is the default, cancellation is effortful Context-driven model Basic meaning is the default, SI derivation is effortful

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Relevance Theory vs. Neo-Griceanism

Framing the empirical question Are scalar implicatures computed by default or as part of an effortful, context-driven process? Default model Pragmatic meaning (SI) is the default, cancellation is effortful Context-driven model Basic meaning is the default, SI derivation is effortful Theory Empirical model Neo-Griceanism Default Relevance Theory Context-driven

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Scalar implicatures: default or not?

answer: it’s not that simple evidence supporting the Context-driven model: Noveck & Posada (2003) and Bott & Noveck (2004) - reaction times in a sentence verification task, Breheny et al. (2006) - reading times, Huang & Snedeker (2009) - eye movements evidence supporting the Default model: Grodner et al. (2007)

• eye movements

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Scalar implicatures: default or not?

answer: it’s not that simple evidence supporting the Context-driven model: Noveck & Posada (2003) and Bott & Noveck (2004) - reaction times in a sentence verification task, Breheny et al. (2006) - reading times, Huang & Snedeker (2009) - eye movements evidence supporting the Default model: Grodner et al. (2007)

• eye movements

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Noveck & Posada (2003)

sentence verification task on three kinds of sentences:

underinformative: Some elephants have trunks. patently true: Some houses have bricks. patently false: Some crows have radios.

for underinformative utterances, the ‘semantic’ interpretation leads to a TRUE response, the ‘pragmatic’ interpretation to a FALSE response predictions:

default: semantic responses slower than pragmatic responses context-driven: pragmatic responses slower than semantic responses

results:

63% pragmatic responses analysis of reaction times of pragmatic vs. semantic responses: pragmatic responses are slower

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Grodner et al.

eye-tracking study in the visual world paradigm

(A)

Click on the girl with some

• f the balls/all of the

balloons. default prediction: pragmatic interpretation of “some”should lead to early disambiguation results: early increase in fixations to the target in both conditions (200-300 ms after quantifier onset)

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Interpreting the Grodner results

not exclusively support for the Default view alternative: pragmatic constraints strongly affect earliest stages of interpretation seemingly slow interpretations may result from integration of resultant interpretation with relevant contextual information computation vs. verification processes requiring additional processing effort constraints on both complement sets for pragmatic interpretation, only on one for semantic interpretation

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Things get tricky

2 problems with getting at the question of Relevance Theory

• vs. Neo-Griceanism via the question of Default:

1 mapping of theories to empirical predictions 2 the empirical data Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Problem 1 - from theory to processing claim

the claim that Neo-Griceanism as a whole should endorse the Default model is not justified ambiguity of the term ” default”

applies to output (e.g. Grice) - the mechanism may be quite complex/require lots of processing effort applies to processing mechanism (Levinson)

conclusion: the question of Default is not a fruitful way of resolving the debate between Relevance Theory and Neo-Griceanism

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Problem 2 - divergent data. . . why?

shift from the question of Default to the question of what factors influence implicature processing

to whether or not the implicature arises to what influences processing, and how

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Brief pause

3 questions

1 Does integration of pragmatic information occur at the

earliest stages of language processing?

2 What are the factors that influence whether or not the

implicature arises?

3 How do these factors influence the processing mechanism

itself?

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Brief pause

3 questions

1 Does integration of pragmatic information occur at the

earliest stages of language processing?

2 What are the factors that influence whether or not the

implicature arises?

3 How do these factors influence the processing mechanism

itself?

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Question 1 - evidence from the gumball paradigm

participants: 28 paid undergraduates from the University of Rochester procedure:

1

display 1 (2s)

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Question 1 - evidence from the gumball paradigm

participants: 28 paid undergraduates from the University of Rochester procedure:

1

display 1 (2s)

2 “KA-CHING” Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Question 1 - evidence from the gumball paradigm

participants: 28 paid undergraduates from the University of Rochester procedure:

1

display 1 (2s)

2 “KA-CHING” 3

display 2

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Question 1 - evidence from the gumball paradigm

participants: 28 paid undergraduates from the University of Rochester procedure:

1

display 1 (2s)

2 “KA-CHING” 3

display 2

4 “You got some of the gumballs.” Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Question 1 - evidence from the gumball paradigm

participants: 28 paid undergraduates from the University of Rochester procedure:

1

display 1 (2s)

2 “KA-CHING” 3

display 2

4 “You got some of the gumballs.” 5

respond YES (agree)/NO (disagree) (timeout: 4s)

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Materials

Subsets Quantifier Condition constructions some some some of the summa all of the alla none of the nunna

• ne of the
• nea

two of the twoa three of the threea seven of the sevena eleven of the elevena 112 randomized trials

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Proportions of judgments at 13 gumballs (full set)

alla summa Proportion of YES responses 0.0 0.2 0.4 0.6 0.8 1.0

“You got all of the / some

• f the gumballs.”

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Mean reaction times for“some of the”at full set

pragmatic semantic

Mean reaction time

500 1000 1500 2000

• “You got some of the

gumballs.”

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Reaction times for YES responses to“some/some of the”

• 1

2 3 4 5 6 7 8 9 10 11 12 13 1000 1250 1500 1750 2000 2250 2500

Gumballs Reaction time in ms

summa some

slowdown effect at full set of gumballs for YES responses interpretation: intrusion of the pragmatic discourse context even at the earliest stages of interpretation

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Brief pause

3 questions

1 Does integration of pragmatic information occur at the

earliest stages of language processing?

preliminarily - yes.

2 What are the factors that influence whether or not the

implicature is computed?

3 How do these factors influence the processing mechanism

itself?

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Brief pause

3 questions

1 Does integration of pragmatic information occur at the

earliest stages of language processing?

preliminarily - yes.

2 What are the factors that influence whether or not the

implicature is computed?

3 How do these factors influence the processing mechanism

itself?

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Face-threatening contexts

influence of politeness on scalar implicature generation context +“Some people loved/hated your poem.” (face-boost/face-threat) “Do you think it’s possible that everyone loved/hated your poem?” fewer pragmatic interpretations of“some”in contexts where the stronger statement would lead to a loss of face for the listener (83% vs. 58%)

[Bonnefon et al. (2009)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Cognitive load

sentence verification task with underinformative sentences executive cognitive resources burdened by memorization of complex dot patterns fewer pragmatic interpretations of“some”under cognitive load

[De Neys and Schaeken (2007)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Focus

paraphrase selection task with auditory stimuli (pitch accent either on the disjunction or the auxiliary) (5) Mary will/WILL invite Fred OR/or Sam to the barbecue. Paraphrases:

• 1. She will invite Fred or Sam or possibly both;
• 2. She will invite Fred or Sam but not both.

increased exclusive (pragmatic) interpretations with pitch accent on“or”(16.5% exclusive vs. 28.6% exclusive)

[Schwarz et al. (2009)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing Recent past Present

Partitive (evidence from the gumball paradigm)

alla summa some Proportion of YES responses 0.0 0.2 0.4 0.6 0.8 1.0

“You got all of the/ some of the/ some gumballs.” increase in pragmatic interpretations

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Grammatical accounts of scalar implicatures

• bservation 1: scalar implicatures arise in embedded positions,

e.g. under attitude verbs (not predicted by a Gricean analysis) (6) Joe believes that some of the students will show up.

• a. ¬[Joe believes that all of the students will show up.]
• b. Joe believes that ¬[all of the students will show up.]

[Chierchia (2004); Chierchia, Fox & Spector (2008)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Grammatical accounts of scalar implicatures

• bservation 1: scalar implicatures arise in embedded positions,

e.g. under attitude verbs (not predicted by a Gricean analysis) (8) Joe believes that some of the students will show up.

• a. ¬[Joe believes that all of the students will show up.]
• b. Joe believes that ¬[all of the students will show up.]
• bservation 2: scalar implicatures are often suspended in

downward-entailing (DE) contexts (9) Joe doubts that Sue or Mary will show up. # Joe doubts that ¬[Sue and Mary will show up.] (i.e. he believes they will both show up)

[Chierchia (2004); Chierchia, Fox & Spector (2008)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Experimental evidence on SIs in DE contexts

paraphrase selection task; embedding of“or”under negation; (10)

• a. Maria asked Bob to invite Fred or Sam to the

barbecue.

• b. Maria asked Bob not to invite Fred or Sam to the

barbecue. What did Maria ask Bob to do? reduced exclusive interpretations under negation (6.8%

• vs. 64.7%)

[Schwarz et al. (2009)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Experimental evidence on SIs in DE contexts

paraphrase selection task with number terms in antecedent of conditional, restriction of“every” , or UE control sentence results: 78% exact interpretations in UE conditional controls, 49% in DE context results: 55% exact interpretations in UE quantified type controls, 27% in DE context effect of DE context in both cases (but very different numbers)

[Panizza et al. (2009)]

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Grammatical accounts of scalar implicatures

implicatures are computed in the grammar via a silent only

• perator O

O a function that takes as arguments a proposition φ and a set of alternative propositions ALT, and returns the conjunction of φ and the negation of all members of ALT that are stronger than φ under downward entailing operators such as negation, antecedent of conditionals, restriction of every, entailment pattern reverses (unless there is focus on the scalar item. . .)

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Against a (purely) grammatical account

critics, e.g. Russell (2006):

departure from rational principles of conversation neither desirable nor warranted grammatical mechanism is arbitrary scalar implicature as inherently pragmatic phenomenon

alternative:

fixation on labeling SIs as pragmatic or semantic does not provide much insight from a processing perspective: view scalar implicature computation as a cue integration problem

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Modeling scalar implicatures - potential cues

prosodic focus - judgment data politeness considerations - judgment data cognitive load - judgment and response latency data DE contexts - judgment and reading time data partitive vs. bare quantifier - judgment and reaction time data relevance of stronger statement? - in preparation contrast set (visual or conceptual)? - in preparation contrast set (linguistic)? information focus? . . .

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Conclusion

from the question of Default to boundary conditions scalar implicature processing as cue integration problem (where prosodic, syntactic, semantic, and pragmatic cues will play a role)

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Thanks!

Katie Carbary Christine Gunlogson Florian Jaeger Patricia Reeder Dana Subik Mike Tanenhaus

Judith Degen Scalar Implicatures

Introduction The psycholinguistic debate Putting scalar implicatures in the grammar A model of scalar implicature processing

Thanks!

Judith Degen Scalar Implicatures