An experimental comparisonbetween presuppositions and
indirect scalar implicaturesJacopo Romoli and Florian Schwarz
(in collaboration with Cory Bill and Stephen Crain)
The project
Comparing the processing of
Indirect scalar implicatures
Presuppositions
Direct scalar implicatures
••••
Inferences
Indirect scalar implicatures
(1) John didn’t always go to the movies ⤳"John went sometimes
••
Inferences
Direct scalar implicatures
(2) John sometimes went to the movies ⤳"John didn’t always go
••
Inferences
Presuppositions
(3) John didn’t stop showing up late forclass ⤳"John used toshow up late for class
••
The processing of directscalar implicatures
Direct SIs associated with a delay (Bott and Noveck 2004, Huang andSnedeker 2009, Chemla and Bott under review, a.o; but see Grondner et al2010, Degen and Tanenhaus 2011, 2012 a.o.)
•
The processing ofpresuppositions
(Global) presuppositions not associatedwith a delay (Schwarz and Tiemann 2012, 2013, Chemla and Bott underreview)
•
The processing ofindirect SI?
What about indirect SIs?•
Traditional view
Same derivation of indirect & direct scalar implicatures
Predicts equivalence in processing(with costs for both)
Contrast with presuppositions expected
•
•
•
Our experiment
Testing this prediction
Comparing the processing of indirect SIsand that of presuppositions
••
Our results
The processing of indirect SIs is closer tothat of presuppositions than that of direct SIs
A challenge for the traditional view
•
•
In other words
Traditional view: Direct SIs = Indirect SIs ≠Presuppositions
Results: Direct SIs ≠ Indirect SIs ≈Presuppositions
•
•
TodayThe phenomena in brief
The traditional approach
Predictions
Experiment
Results, implications, and further directions
•••••
The phenomena
Direct and indirect scalarimplicatures
(1) John sometimes went to the movies ⤳"John didn’t always go
(2) John didn’t always go to the movies ⤳"John sometimes went
•
•
Presuppositions
(3) John didn’t stop showing up late forclass ⤳"John used toshow up late for class
•
Presence and absence ofinferences
All these three inferences can be absent •
Direct scalarimplicatures
(1) John sometimes went to the movies... In fact, he always did!
•
Indirect scalarimplicatures
(2) John didn’t always go to the movies... In fact, he never went!
•
Presuppositions
(3) John didn’t stop showing up late... because he never did!
•
Compare
(4) John sometimes went to the movieslast week... #In fact he never went!
•
In sumDirect and indirect SIs
Presuppositions
Theoretical Goals:
explain how they arise
predict where these inferences occur
•••••
The traditional view
Traditional grouping
Direct = indirect SIs ≠ Presuppositions•
Deriving SIs: the Gricean algorithm
Hear an utterance
Comparison with an alternative utterance
Competitor is false if stronger than the assertion
•••
Deriving SIs: the Gricean algorithm
The speaker said A
The speaker might have said B
If B is stronger than A, B is false
•••
How do we obtaincompetitors?
replace certain words in the assertion
<some, all>
<sometimes, always >
...
••••
Deriving direct SIs
(1) John sometimes went to the movies (2) John always went to the movies
⤳"John didn’t always go to the movies
••
Deriving indirect SIs
(1) John didn’t always go to the movies (2) John didn’t sometimes go to the movies
⤳"It’s false that John didn’t sometimes go to themovies="John went sometimes
••
A unified approach
A unified account of direct and indirect SIsA scalar implicature algorithm A theory of competitors
•••
Deriving presuppositions
... ...
(1) John stopped showing up late for class
only defined/can only update a context that entailsthat John used to show up late Stalnaker 1974, Karttunen 1973 1974, Heim 1983 and much
subsequent work
•
•
•
•
Presuppositionprojection
Predicting the presuppositions of complexsentences
(2) John didn’t stop showing up late for class
(3) If John stopped showing up late for class, ...
(4) Did John stop showing up late for class?
⤳"John used to show up late for class
•
•
•
•
Deriving projection
Presuppositions and indirect implicatures mightlook similar on surfaceBut the derivations of these inferences are quite
different on standard assumptions
•
•
Parenthesis
More recent approaches brings SIs andpresuppositions closer (Abusch 2009, Chemla 2009, Romoli 2012 a.o.)
We’ll come back to this
•
•
Deriving the absence:scalar implicatures
(2) John sometimes went to the movies... In fact, he always went!
•
Deriving the absence:scalar implicatures
The scalar implicature is not computed
The speaker uttered the competitor
She cannot think that it’s false
•••
Deriving the absence:presuppositions
(2) John didn’t stop showing up late for class... because he never did!
•
Deriving the absence:presuppositions
Presuppositions are always present
But not necessarily at the global level
••
PS(p) = p and its presuppositions•
Global presuppositions
PS[not[John stopped showing up late]] =John didn’t stop showing up late and he usedto do it
•
Local presuppositions
not[PS[John stopped showing up late]] =John didn’t stop showing up late
•
In sum
Traditional grouping
Direct SIs and indirect SIs ≠ Presuppositionsunified treatment of presence/absence of
direct and indirect SIs different from how the presence/absence of
presuppositions is derived
•
••
•
Processing
Predictions forprocessing
Direct SIs and indirect SIs ≠ PresuppositionsThe processing of direct = indirect SIsTheir processing ≠ that of presuppositions
•••
What we know
The presence of direct scalar implicatures associatedwith a delay (Bott and Noveck 2004, Breheny 2006, Huang and Snedeker2009)
The presence of (global) presuppositions notassociated with a delay (Schwarz and Tiemann 2011, 12, Chemla and Bott underreview)
Global presuppositions are faster than local ones (Chemla and Bott under review)
•
•
•
What we know
Response Times:
absence of direct SI < presence of direct SI
‘absence’ of presupposition (= local PS) > presence of (global) presupposition
•••
Expectations
the processing of indirect SIs = theprocessing of direct SIs
Response Times:
absence of indirect SI < presence of indirect SI
•
••
Experiment
Prediction testedResponse Times:
absence of indirect SI < presence of indirectSI
absence of presupposition > presence ofpresupposition
Prediction: Cross-over interaction
••
•
•
Participants
25 native speaker of English
UPenn undergraduates
Received course credits for participation
•••
Material and Procedure
Sentence picture matching task
Pictures representing a character and herschedule
••
Material and Procedure
Participants chose among three pictures
One target, one distractor, and one covered (Huang et al 2013, Romoli et al 2011)
•
•
Material and Procedure
Instruction
one and only one picture matches thesentence
••
Test and control trials
24 test trials
(12 always; 12 stop)
40 controls (no negation)
20 target; 20 covered
••
••
Design2 x 2
Type of trigger
stop vs. always
Inference?
inference or no-inference
•••
••
DesignType Sentence Inference?
Stop (PSP)John didn’t stop eating
strawberries onWednesday
Inference: He ate strawberries before
No Inference: He didn’t eat strawberries
before
Always(ISI)
John didn’t always eatstrawberries this week
Inference: He sometimes ate
strawberries
No Inference: He never ate strawberries
Always-inference
Distractor Target
Always-no inference
DistractorTarget
Stop-inference
Distractor Target
Stop-no inference
Distractor Target
Dependent variables
Choice of target vs. covered picture
Reaction times of target choices
••
Remember predictions
RTs of target choices
always-no-inference < always-inference
stop-no-inference > stop inference
•••
Results and discussion
Main effect of inference
Simple effects of inference for both ‘stop’and ‘always’
No interaction
••
•
0.00
0.25
0.50
0.75
1.00
always:Inf always:NoInf stop:Inf stop:NoInfInference
%Loc
Inf
NoInf
% of Target Choices
Note Split in subjects
Only close to 50% of subj. had any localresponses at all
Among those, no-inference targets werechosen much more frequently
••
•
0.00
0.25
0.50
0.75
1.00
always:Inf always:NoInf stop:Inf stop:NoInfInference
%
Loc
Inf
NoInf
% of Target Choices ('Local' Subjects)
Response Data
Overall proportion of No-Inferenceinterpretations relatively low
Huang et al. (2013) found 87% No-Inference choices for direct SIs with ‘some’
(We have found similar ‘logical’ responserates in TVJ studies)
•
•
•
(marginal) main effect of Inference Simple effects of Inference for both ‘always’
(marginal) and ‘stop’ No interaction
••
•
RT for Target Choices
Inference
ms 5000
10000
Inf NoInf
expression
always
stop
Remember theexpectation
RTs of target choices
always-no-inference < always-inference
stop-no-inference > stop inference
•••
What we found
RTs of target choices
always-no-inference > always-inference
stop-no-inference > stop inference
•••
Results
The processing profiles ofpresupposition and indirect SIs are similar
Neither one involves additional cost forInference
•
•
Results
The presupposition part is consistentwith previous findings (Chemla and Bott under review)
The indirect SIs is a novel finding
•
•
Implications
The traditional view
The processing of indirect SIs = that of direct SIs
⤳"associated with a cost
••
Implications
Our results contrary to this prediction
Indirect SIs are not costly
Their processing is similar to that ofPresuppositions
A challenge for the traditional grouping
•••
•
Implications
Some recent accounts (Chemla 2010, Romoli 2012a.o.)
Indirect SIs ≈ PresuppositionsBut no difference predicted between direct and
indirect Direct SIs = Indirect SIs
••
•
•
Possible explanations
Two directions
Indirect SIs are presuppositions
Indirect SIs are a different type of SIs
••
Indirect SIs arepresuppositions
What if (1) presupposes (2)?
(1) John always went to the movies
(2) John sometimes went to the movies
•••
Indirect SIs arepresuppositions
By projection mechanism (3) presupposes (2)
(3) John didn’t always go to the movies
(2) John sometimes went to the movies
•••
Indirect SIs arepresuppositions
Indirect SIs as presuppositions
It can explain why their processing profiles aresimilar
••
Open issues
Other differences between indirect SIs andpresuppositions
•
Projection
Differences in projection between presuppositionsand indirect SIs (Chemla 2009, Romoli 2012)
Projection from antecedent of conditionals,questions, modals ...
Universal projection in negative quantifiers
•
•
•
Differences inprojection?
(1) John stopped smoking
(2) If John stopped smoking, Mary must be happy.
(3) Did John stop smoking?
(4) It’s possible that John stopped smoking.
⤳"John used to smoke
•••••
Differences inprojection?
(1) If John didn’t always come, Mary must be happy.
(2) Did John not always come?
(3) It’s possible that John didn’t always come.
?⤳##John came sometimes
••••
Differences in projection
Universal projection inferences in negativequantifiers
Accepted more often for presuppositions thanIndirect SIs (Chemla 2009)
•
•
Differences in projection
(1) None of these students stopped smoking
⤳"All of these students used to smoke
••
Differences in projection
(3) None of these students did all of the readings
?⤳"All of these students did some of the readings.
••
Other questions
What about the Scalar Implicaturealgorithm?
Both operative? two ways of derivingindirect scalar implicatures?
•
•
In sum
The indirect SIs-as-presuppositions line
can account for the processingsimilarities
open issues about differences withpresuppositions
••
•
Indirect SIs areobligatory SIs
Indirect SIs are a different type of scalarimplicatures
Obligatory SIs (Spector 2007, Chierchia 2004, 2013, Magri 2011)
•
•
Indirect SIs areobligatory SIs
Obligatory SIs cannot be suspended
How do we account for theirabsence?
local computation
••
•
Notation
Imagine for any sentence p
SI(p) = p and its scalar implicatures
••
Global computation
John didn’t always come
SI([not[John always come]]) = John didn’t always come and hecame sometimes
••
Local computation(vacuous)
John didn’t always come
[not[SI(John always come)] = John didn’t always come
••
Indirect SIs areobligatory SIs
Indirect SIs computed locally in (1)
(1) John didn’t always come... In fact he never did!
••
Embedded SI andnegation
Scalar implicatures are marked undernegation (Chierchia et al 2012 a.o.)
(1) Jack didn’t meet Paul or Sue... he met both!
neg(SI(Jack meet Paul or Sue) = he metneither or he met both
•
•
•
Indirect SIs areobligatory SIs
Always-no inference involves a localscalar implicature under negation
Local SIs under negation are marked
marked readings are reflected inprocessing
•
••
Indirect SIs areobligatory SIs
explaining the delay associated withsuspending indirect SIs
•
Differences withpresuppositions
What about the differences betweenindirect SIs and presuppositions?
•
No projection fromantecedent ofconditionals
(1) If John always came, Mary must be happy
(2) If John sometimes came, Mary must be happy
⤳"It’s false that if John sometimes came, Mary must behappy.
•••
No projection throughpossibility modals
(1) It’s possible that John always came.
(2) It’s possible that John sometimes came.
(1) entails (2) so no SI predicted here
•••
Universal projectionfrom negative quantifiers
Universal projection can be predicted for IndirectSIs (Chemla 2010, Romoli 2012, to appear)
•
Predicting universalinferences
(1) None of these students did all of the readings
(2) Not all of these students did all of thereadings
⤳"All of these students did some of the readings
••
•
Predicting universalinferences
Universal inference for obligatory scalarimplicatures
But not a presuppositions so compatible withdifferences found in Chemla 2009
•
•
In sum
The indirect SIs as obligatory scalar implicatures
can account for the processing similarities
might explain differences with presuppositionsbetter than the presupposition line
For now both directions are open!
•••
•
Conclusions
Conclusions
Experimental results on the processingof indirect SIs and presuppositions
Neither is associated with a cost
Contrary to previous work on directSIs
•
••
Conclusions
A challenge for the traditionalgrouping of these inferences Direct SIs = indirect SIs ≠ Presuppositions
••
Conclusions
Hypotheses explored Indirect SIs are presuppositions
Indirect SIs are obligatory SIs
••
•
Conclusions
Remaining questions Other differences between presuppositions
and Indirect SIs
How many ways of deriving SIs?
••
•
Further directions
Establishing the result: comparing directly notalways and sometimes
Acquisition of presuppositions versus that ofindirect scalar implicatures
•
•
Thanks!Collaborators
Cory Bill Florian Schwarz Stephen Crain
Others
Rosalind Thornton, Kelly Rombough, Dorothy Ann, EmmanuelChemla, Danny Fox, Clemens Mayr. Yasutada Sudo, Lyn Tieu
•
•
•
•