contrast in cleft sentences emilie destruel (project in collaboration with dan velleman, ut austin)...
TRANSCRIPT
Contrast in cleft sentences
Emilie Destruel(project in collaboration with Dan Velleman, UT Austin)
[email protected] University of Iowa
Semantics ColloquiumFrankfurt
01/21/2014
Goals for today
1. OFFER background on the notions of focus and contrast,
2. MOTIVATE the present study,
3. PROVIDE empirical evidence for the felicity of the English it-cleft,
4. REFINE the notion of contrast in discourse-semantics terms,
5. EXPLORE cross-linguistics implications.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
2
THEORETICAL BACKGROUND
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
3
What is focus?• Universal category of information structure (Krifka 2008)• Focus evokes a set of alternative propositions which the
speaker takes to be salient (Rooth 1992)
(1) Q: Who likes Sue? A: [Mary]F likes Sue (alternatives: [Jane]F likes Sue, [Peter]F likes Sue, etc.)
A: * Mary likes [Sue]F
• Focus does not have to be “new”, but must be unpredictable (not yet in the common ground). (Beaver & Velleman 2011)
(2) Gary, Larry, Harry, Barry and Mary all showed up at the party. And you won't believe who got the drunkest. It was_____ !
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
4
Two types of focus• Informational focus(3) Q: Who is chewing tobacco?
A: [John]F is chewing tobacco.
• Identificational focusAssociated with a semantically stronger interpretation:
ExhaustiveContrastiveCorrectiveVerum
(4) S1: Last Halloween, John went as Batman.S2: No, John went as [Spiderman]F.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
5
How is focus realized?
• Different across languages (Büring 2008)(5) Q: Who is chewing tobacco?
A: JOHN is chewing tobacco EnglishA: C’est Jean qui mâche du tabac FrenchA: Está masticando tabaco Juan Spanish
• Identificational focus must be realized in a special position, via more marked strategy (Kiss 1998)
(6) A: It’s John who is chewing tobacco it-cleft Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
6
The English it-cleft
(7) It’s John who is chewing tobacco.
• Associated with an exhaustive inference (a) • and/or contrast (b)
(a) Nobody other than John is chewing tobacco.(b) John as opposed to, for example Mary, is chewing tobacco.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
7
Contrast• Previously understood on the basis of semantic terms:
• Exclusion of alternatives (Kiss 1998)• Presence of an explicit antecedent focal alternative
(Rooth 1992; Schwarschild 1999)
• Account for why clefts are good as corrections but sound odd as direct answers to explicit wh-questions.
• … but some facts remain unexplained. Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
8
Motivation for the current study
• In contexts where the speaker provides a complete, exhaustive answer, the it-cleft is not always produced.
• In contexts in which an antecedent is available, speakers may nevertheless choose not to use an it-cleft.
(8) S1:Darren sounded really excited about his vacation. I think he’s going to Canada.
S2a: Actually, he’s going to Mexico.S2b: ? Actually, it’s Mexico that he’s going to.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
9
So …
• Exhaustivity and presence of a focal antecedent may be necessary conditions, but we think they are not sufficient to explain the natural use of the cleft in English …
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
10
What more is needed ?
Specific research questions
• Under which conditions are it-clefts actually produced in English? What factors are relevant for their felicity?
• How can the notion of contrast be refined to reflect these factors?
• What might this mean for the bigger cross-linguistic picture? D
estr
uel,
Sem
antic
s Co
lloqu
ium
Fr
ankf
urt,
01/2
1/14
11
Our proposal
Broaden the notion of contrast to include two features:• LINGUISTIC: Rooth-style contrast with a prejacent
proposition. Makes clefts more felicitous. Directly promotes cleft production.
• METALINGUISTIC: Conflict with hearer’s expectations. Takes the discourse in an unexpected direction. Makes canonical sentences less felicitous, which
indirectly promotes cleft production.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
12
Roadmap
1. Evidence from a production study
2. Evidence from a rating study
3. Discussion
4. Implications for cross-linguistics data & Direction for future work D
estr
uel,
Sem
antic
s Co
lloqu
ium
Fr
ankf
urt,
01/2
1/14
13
PRODUCTION STUDY:THE CONTRA-PRESUPPOSITIONAL USE OF CLEFTS
Experiment 1
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
14
Goal of the study
• A practical goal: how to elicit felicitous it-clefts?
• An observation: natural English it-clefts often seem to be counter-presuppositional … not informational.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
15
Counter-presuppositional context:
• Your friend says: We were planning Amy's surprise party for weeks. I can't believe someone ruined the surprise. Do you have any idea why Alice told her about it?
• Correction: Ken.
• You say: __________________________________
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
16
Informational context:
• Your friend says: We were planning Amy's surprise party for weeks. I can't believe someone ruined the surprise. Who told her about it?
• Answer: Ken.
• You say: ___________________________
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
17
Prediction
Counter-presuppositional contexts will elicit clefts where informational contexts did not.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
18
Methodology2 x 2 design: • context (informational/ counter-presuppositional) • grammatical function of element in the answer
(subject/ object).
• 15 native English speakers (UT undergraduates).• 5 lexicalizations created per context.• 2 versions of the written questionnaire.• Each participant saw a total of 10 experimental stimuli +
5 fillers.• Collected 150 sentences.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
19
Results (raw numbers)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
20
• Difference in distribution of sentence form across both contexts is highly significant. (χ2 (1) = 36.24, p<0.01)
• The odds of using a cleft is 13.1 times higher given the counter-presuppositional context than the informational context.
• Difference in distribution of sentence form in informational context is highly significant. (z = 7.27, p<0.01)
• Difference in distribution of sentence form in counter-presuppositional context is not highly significant. (z = -0.57, p=0.5)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
21
Interim discussion
• It-clefts are actually produced, mostly to offer a correction to a presupposition. They do make really bad direct answers.
• While clefts are clearly contrastive, clefting is not the only strategy available to speakers. Canonicals (with prosodic prominence) are also produced.
So, what makes a cleft a better alternative?
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
22
RATING STUDY:DEGREE OF BELIEF OR DEGREE OF AT-ISSUENESS?
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
23
Experiment 2
The intuition behind this study…
• The production study can only tell us so much about the cleft’s use.
• Question: what difference actually matters between the informational & counter-presuppositional contexts?
• Intuition: cleft is doing more than just linguistic contrast... D
estr
uel,
Sem
antic
s Co
lloqu
ium
Fr
ankf
urt,
01/2
1/14
24
• The cleft seems increasingly better when the speaker’s expectations are expressed more strongly, and the conflict with hearer’s expectations intensifies.
(related to Zimmerman 2008, 2011)
Two types of expectations seem important:• Expectations about the world (Strength of belief)• Expectations about the discourse (At-issueness)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
25
Strength of belief
• Expectations involving speaker’s beliefs about the world (common ground)
… expressed as assertions or presuppositions.
• Gradient notion: “no overt belief” “weak belief” “strong belief”
• More strongly expressed beliefs lead to stronger conflict between interlocutors.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
26
At-issueness
• Expectations the speaker has about the direction in which discourse is progressing…
… expressed by marking (part of) the proposition as:
• At-issue: asserted content. “The main point of the utterance”
• Non-at-issue: presupposed content. “propositions which the sentences are not primarily about”
(Simons et al. 2010, Tonhauser 2012, Tonhauser et al. 2013)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
27
Hypothesis
Both factors Strength of belief and At-issueness play a role in the felicity of the it-cleft:
• clefts will become increasingly felicitous as strength of belief increases, and vice-versa for canonicals.
• clefts will become increasingly felicitous when the proposition corrected is not-at-issue, and vice-versa for canonicals.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
28
StimuliContexts:“This bean dip is fantastic. I really want the recipe... • …Who made it ?” Informational• …I guess maybe Shannon made it.” At-issue weak belief• …Shannon made it.” At-issue strong belief• …I can’t believe Shannon made it.” Counter-presuppositional
Responses to rate:• Actually, Tim made it. Canonical• Actually, it was Tim who made it. It-Cleft D
estr
uel,
Sem
antic
s Co
lloqu
ium
Fr
ankf
urt,
01/2
1/14
29
Methodology
• Felicity judgment task on 1-5 point likert scale: 1- not natural at all, I would never say this.3- sounds natural, I would sometimes say this.5- extremely natural, exactly how I would say this.
• 12 native English speakers (undergraduates at St Edwards University).
• 5 lexicalizations created per context.• Each participant saw a total of 20 experimental stimuli
(5 in each context).• Collected a total of 240 ratings.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
30
Results (raw numbers, collapsed for GF)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
31
Ratings of canonicals by context(based on 240 observations)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
32
Ratings of it-clefts by context (based on 240 observations)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
33
• For clefts: ANOVA shows a significant effect of context on rating (p<0.001). Tukey’s HSD test shows significant differences between:• Informational and low contrast (p=0.015)• Informational and strong contrast (p<0.001)• Informational and counter-presuppositional (p<0.001)
• For canonicals: ANOVA shows a significant effect of context on rating (p<0.001). Tukey’s HSD test shows significant differences between:• Counter-presuppositional and informational (p<0.001)• Counter-presuppositional and low contrast (p<0.001)• Counter-presuppositional and strong contrast (p<0.001)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
34
DISCUSSION
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
35
Canonicals• Canonicals are significantly worse in the counter-
presuppositional context than in the other three.
• Strength of belief is not significant, but at-issueness plays a role.
• Our interpretation: The use-conditions of the canonical must make reference to
the distinction at-issue/ non-at-issue. Canonical sentences are preferred when used to address or
contradict at-issue (asserted) content. Canonical sentences signal “things are proceeding as
normal” and are infelicitous in contexts where the component contradicted is non-at-issue.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
36
It-clefts• It-clefts are significantly worse in the informational
context than in the other three.
• Strength of belief has some effect, but it only slightly significant.
• Our interpretation: All it takes for the clefts to be felicitous is linguistic
contrast … but for clefts to be preferred, the canonical must not be
available, and that due to metalinguistic contrast: conflict about the direction of the discourse.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
37
Competition is key!
• Clefts and canonicals are in competition.• Clefts are felicitous when the canonical is a “bad” option.• Specifically, clefts are better than canonicals in the
counter-presuppositional context due to a combination of two effects:• Clefts improve because there’s an antecedent.• Canonicals degrade because the antecedent is not-at-
issue. Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
38
Summing up
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
39
• It-clefts are used in English when the corresponding canonical is less felicitous or unavailable.
• The notion of contrast must include a metalinguistic component: conflict with hearer’s expectations, specifically about the advancement discourse.
• So far, the only statistically significant feature is at-issueness: canonicals are worse when addressing non-at-issue content.
• In-line with results on Chinese (Greif 2012) where increased phonetic marking and clefts are produced with corrections of presupposed information.
CROSS-LINGUISTIC IMPLICATIONS
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
40
What’s next: cross-linguistically
• Felicity contexts for clefts (and other strong focus constructions) are supposed to be the same crosslinguistically…
• … So why don’t all languages use clefts in the same contexts?
A speculative hypothesis: the felicity conditions for canonical sentences vary.
• A canonical sentence signals “things are proceeding as normal” – but what’s “normal” could vary.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
41
What’s normal?
ENGLISH: “Don’t use canonicals to address not-at-issue content.”(normally, we discuss what’s at issue)
FRENCH: “Don’t use canonicals for subject focus.”(normally, subjects are old information/ topics)
K’ICHEE’: “Don’t use canonicals for transitive subject focus.”(normally, transitive subjects are old information)
HUNGARIAN: “Don’t use canonicals for partial answers.”(normally, complete answers are preferred)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
42
What’s next: English
• A more thorough ratings experiment – having two additional contexts: Informational At-issue weak belief At-issue strong belief Not at-issue weak belief Not at-issue strong belief Counter-presuppositional
• A production experiment covering all contexts.• Run the experiment in the oral modality.• Supplement with corpus data.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
43
Thank you very much!
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
44
And thanks to:
• David I. Beaver• Aixin Tan• Malte Zimmermann• Mira Grubic• And our participants (the undergraduates at UT Austin &
St Edwards University)
References • Declerck, R. (1984). The pragmatics of it-cleft and wh-clefts. Lingua, 64, p.251-289.• Greif, M. (2012). Corrective focus in Mandarin Chinese: a question of belief? Muenchen:
Lincom Europa.• Katz, J. and E. Selkirk. (2011). Contrastive focus vs. discourse-new: Evidence from phonetic
prominence in English. Language, 87, p. 771-816. • Kiss, K. E. (1998). Identicational focus versus information focus. Language, 74, p. 245-273.• Rooth, M. (1992). A theory of focus interpretation. Natural Language Semantics, 1, p. 75-
116.• Schwarschild, R. (1999). Givenness, avoidf and other constraints on the placement of
accent. Natural Language Semantics, 7, p. 141-177.• Tonhauser, J. (2012). “Diagnosing (non)- at-issue content”. In Proceedings of under-
represented languages of the Americas (SULA) 6, p. 239-254. UMass: Amherst, GLSA.• Velleman, D., Beaver, D., Destruel, E., Bumford, D., Onea, E., and Coppock, E. (2012). It-clefts
are IT- (inquiry terminating) construction. In Proceedings of SALT 22 , p. 441-460.• Zimmermann, M. (2008). Contrastive focus and emphasis. In Acta Linguistica Hungarica , p.
347-360.• Zimmermann, M. (2011). The grammatical expression of focus in West Chadic: Variation
and uniformity in and across languages. Linguistics , 49, p. 1161-1211.
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
45
APPENDIX
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
46
Stimuli sample from newly conducted rating experiment (1)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
47
• Informational a. Speaker A: This bean dip is fantastic. I really want to get the recipe. Who
brought it?b. Speaker B: Lyle brought it. vs. Speaker B: It’s Lyle who brought it.
• At-issue weak beliefa. Speaker A: This bean dip is fantastic. I really want to get the recipe. I think
Sharon brought it.b. Speaker B: No, Lyle brought it. vs. Speaker B: No, it’s Lyle who brought it.
• Not at-issue weak beliefa. Speaker A: This bean dip is fantastic. I really want to get the recipe. But
Sharon -- who I'm guessing brought it -- already left.b. Speaker B: No, Lyle brought it. vs. Speaker B: No, it’s Lyle who brought it.
Stimuli sample from newly conducted rating experiment (2)
Des
true
l, Se
man
tics
Collo
quiu
m
Fran
kfur
t, 01
/21/
14
48
• At-issue Strong beliefa. Speaker A: This bean dip is fantastic. I really want to get the recipe. It
turns out that Sharon brought it.b. Speaker B: No, Lyle brought it. vs. Speaker B: No, it’s Lyle who brought it.
• Not At-issue Strong beliefa. Speaker A: This bean dip is fantastic. I really want to get the recipe. But
Sharon, who brought it, already left.b. Speaker B: No, Lyle brought it. vs. Speaker B: No, it’s Lyle who brought it.
• Counter-Presuppositionala. Speaker A: This bean dip is fantastic. I really want to get the recipe. I can’t
believe that Sharon brought it -- she’s normally not a very good cook.b. Speaker B: No, Lyle brought it. vs. Speaker B: No, it’s Lyle who brought it.