distinguishing contrastive, new and given information

84
1 Distinguishing contrastive, new and given information Angelika Kratzer & Elisabeth Selkirk Workshop on Prosody and Meaning, Barcelona September 17, 2009

Upload: vuongngoc

Post on 14-Feb-2017

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Distinguishing contrastive, new and given information

1

Distinguishing contrastive, new and given information

Angelika Kratzer & Elisabeth Selkirk Workshop on Prosody and Meaning, Barcelona

September 17, 2009

Page 2: Distinguishing contrastive, new and given information

Bolinger versus Chafe

An old debate that is not yet resolved

2

Page 3: Distinguishing contrastive, new and given information

Bolinger 1961, 87

No less troublesome than our inability to find a uniquely contrastive pitch, however, is our failure to define what is meant by ‘contrastive’. … In I found a bóok. Whóse book? Jóhn's book. Not Jóe's? there is no difference between whose, John's, and Joe's as far as accent or reasons for accent are concerned, but we are likely to limit the term ‘contrastive’ to the word Joe's in the last sentence of the series. In a broad sense every semantic peak is contrastive.

3

Page 4: Distinguishing contrastive, new and given information

Chafe 1976, 34

I favor, on the contrary, the view that contrastive sentences are qualitatively different from those which simply supply new information from an unlimited set of possibilities.

4

Page 5: Distinguishing contrastive, new and given information

No distinction between contrastiveness & newness

Bolinger 1961, Jackendoff 1972, Rooth 1985, 1992, Krifka 1992, 1993, Schwarzschild 1999, Ladd 1980, 1996, 2008, Gussenhoven 1983, 1992, 2004, Selkirk 1984, 1996, Kratzer 1991, Büring 2007, 2008, Féry & Krifka 2008, and many others.

5

Page 6: Distinguishing contrastive, new and given information

Distinction between contrastiveness & newness

Halliday 1967, Chafe 1976, Rochemont 1986, Kiss 1998, Vallduví & Vilkuna 1998, d’Imperio 1997, Pierrehumbert & Beckman 1998, Frota 2000, Face 2002, LeGac 2002, Selkirk 2002, 2003, 2007a, 2007b, Neeleman and Szendroi 2003, Féry and Samek-Lodovici 2006, Kratzer 2004, Kratzer and Selkirk 2007, Katz and Selkirk 2009, and many others.

6

Page 7: Distinguishing contrastive, new and given information

The semantics needs to distinguish newness and contrastiveness

Katz & Selkirk examples

7

Page 8: Distinguishing contrastive, new and given information

8

Contrastive versus New

Gary is a really bad art dealer. He gets attached to the paintings he buys. He acquired a few Picassos and fell in love with them. The same thing happened with a Cezanne painting.

So he would only offer that Modigliani to MoMA.

Page 9: Distinguishing contrastive, new and given information

9

Gary is a really bad art dealer. He gets attached to the paintings he buys. He acquired a few Picassos and fell in love with them. The same thing happened with a Cezanne painting.

•  So he would only offer that Modigliani to MoMA.

He wouldn’t offer that Picasso to MoMa.

He wouldn’t offer that Cezanne to MoMa.

Page 10: Distinguishing contrastive, new and given information

10

Contrastive versus New

•  Gary is an art dealer. Lately he’s been very picky about which museum he deals with; he doesn’t do business with the Metropolitan or the Guggenheim.

So he would only offer that Modigliani to MoMA.

Page 11: Distinguishing contrastive, new and given information

11

Gary is an art dealer. Lately he’s been very picky about which museum he deals with; he doesn’t do business with the Metropolitan or the Guggenheim.

•  So he would only offer that Modigliani to MoMA.

He wouldn’t offer that Modigliani to the Metropolitan

He wouldn’t offer that Modigliani to the Guggenheim

Page 12: Distinguishing contrastive, new and given information

The problem

He would only offer that [Modigliani]F to [MoMA]F.

•  If both Modigliani and MoMA were simply marked for focus (F-marked), we would get the wrong alternative set, assuming standard accounts like Rooth (1992), for example.

12

Page 13: Distinguishing contrastive, new and given information

Wrong alternative set

He would offer that ______ to ______ Modigliani MoMA

Cezanne Metropolitan Picasso Guggenheim etc. etc.

{p: ∃x ∃y [painter(x) & museum(y) & p = λw offer(x)(y)(Gary)(w)]}

13

×

Page 14: Distinguishing contrastive, new and given information

Conclusion

•  The semantics needs a distinction between focus of contrast and discourse-new

14

Page 15: Distinguishing contrastive, new and given information

The phonology makes that distinction, too

Katz & Selkirk (2009)

15

Page 16: Distinguishing contrastive, new and given information

Condition A: FoC - New

Gary is a really bad art dealer He gets attached to the paintings he buys. He acquired a few Picassos and fell in love with them. The same thing happened with a Cezanne painting. So he would only offer that [Modigliani]FOC to [MoMA]New. I bet the Picassos would have fetched a much higher price

16

Page 17: Distinguishing contrastive, new and given information

Condition B: New - FoC

Gary is an art dealer. Lately he’s been very picky about which museum he deals with; he doesn’t do business with the Metropolitan or the Guggenheim. He would only offer that [Modigliani]New to [MoMA]FOC. He says that’s the only place with a good enough space to hang it in.

17

Page 18: Distinguishing contrastive, new and given information

Gary was a successful art dealer, and could afford to be pretty demanding with his clients. He would never make a deal unless the price was right and he respected the buyer. He will probably offer that [Modigliani]New to [MoMA]New. But only for a six figure sum.

Condition C: New-New

18

Page 19: Distinguishing contrastive, new and given information

Tonal Representation

•  The contrastive focus and new constituents all bear H* or L+H* pitch accents.

•  They are separated into distinct major phonological phrases with a L- edge tone.

•  ( ….. Modigliáni )PhP( to MóMA)PhP )IP

H* L- H* L- L%

19

Page 20: Distinguishing contrastive, new and given information

Duration: Main Finding

•  The duration of a contrastive focus constituent is greater than the duration of a new constituent in the same sentential position.

20

Page 21: Distinguishing contrastive, new and given information

Duration results

Complement 1 Complement 2 21

Page 22: Distinguishing contrastive, new and given information

Pitch: Main Finding

•  The patterns of pitch prominence of Comp 1 relative to pitch prominence in Comp 2 differed significantly between the three conditions.

•  Two effects are responsible: Greater pitch protrusion on FoC-marked Greater pitch compression after FoC-marked

22

Page 23: Distinguishing contrastive, new and given information

Summary of main findings

•  FoC-marked constituents show greater pitch prominence than New-constituents within the same sentence.

•  In between-sentence comparisons, FoC-marked constituents show greater duration than New-constituents in corresponding syntactic positions

23

Page 24: Distinguishing contrastive, new and given information

Architecture question 1

Q: Can we capture the relation between phonetic prominence and information structure status in a model of grammar in which the phonology or phonetics does not speak to semantics or pragmatics?

A: Yes, if syntax mediates, via its interfaces with semantics and phonology.

24

Page 25: Distinguishing contrastive, new and given information

Architecture question 2 Q: How to capture this relation between

phonetic prominence and information structure marking in the syntax?

A: An interface which assigns FoC-marking a distinct phonological representation.

A phonetic implementation component which accesses phonological representation.

25

Page 26: Distinguishing contrastive, new and given information

An interface constraint referring to Focus of Contrast

•  Maximize FoCus Stress If a syntactic constituent is a focus of contrast

(FoC), it bears maximal stress within any prosodic domain that contains it.

(cf. Jackendoff 1972, Truckenbrodt 1995, 2006)

•  A stress is maximal in some prosodic domain if it is greater than any other stress within the same prosodic domain.

26

Page 27: Distinguishing contrastive, new and given information

Proposed difference between contrastive focus and new: Degree of phrasal stress at Comp 2

Condition B: presence of head stress prominence of IP

( x )IP

( x ) ( x )PhP Subject only Verb Complement1New Complement2FoC

Condition C: absence of head stress prominence of IP

( Ø )IP

( x ) ( x )PhP Subject Ø Verb Complement1New Complement2New

27

Page 28: Distinguishing contrastive, new and given information

Condition A: presence of head stress prominence of IP

( x )IP

( x ) ( x )PhP Subject only Verb Complement1FoC Complement2New

Condition C: absence of head stress prominence of IP

( Ø )IP

( x ) ( x )PhP Subject Ø Verb Complement1New Complement2New

Proposed difference between contrastive focus and new: Degree of phrasal stress at Comp 1

28

Page 29: Distinguishing contrastive, new and given information

A phonetic implementation that refers to (degrees/levels of) stress

in phonological representation

•  The greater the stress, the greater the duration.

•  The greater the stress on a H*, the greater the pitch protrusion.

•  Etc.

29

Page 30: Distinguishing contrastive, new and given information

Interim Summary

Both the semantics and the phonology/phonetics require that a distinction be made between the syntactic representation of focus of contrast and discourse-new.

30

Page 31: Distinguishing contrastive, new and given information

Representing the three-way distinction

31

Page 32: Distinguishing contrastive, new and given information

The syntactic representation of a 3-way distinction

Chafe 1976

What we propose

focus of contrast new given

FoC-marking N-marking G-marking

32

Page 33: Distinguishing contrastive, new and given information

The attraction of this possibility

• N-marking is in fact superfluous: The prosodic properties that would have to be associated with N-marking already follow from independent principles.

•  An account of the distribution of G-marking & FoC-marking and their effects on prosody and semantic interpretation is straightforward.

33

Page 34: Distinguishing contrastive, new and given information

Typological support for this approach

• Whether FoC-marking or G-marking are expressed in sentence prosody varies from one language to the next. Some languages do not mark givenness prosodically. Others mark neither givenness nor FoCus.

•  All languages display a default prosody associated with unmarked, ‘neutral’, all-new sentences.

34

Page 35: Distinguishing contrastive, new and given information

To be shown here

The ‘default’ status of the phonology of New A simple phonology of G-marking A straightforward semantics of G-marking √ A simple phonology of FoC-marking A straightforward semantics of FoC-marking

√ = done

35

Page 36: Distinguishing contrastive, new and given information

‘New’ has default prosody

36

Page 37: Distinguishing contrastive, new and given information

Basic fact

•  In Standard British and American English every new phrase must contain a pitch accented word.

Page 38: Distinguishing contrastive, new and given information

Pitch accents in all-new sentences

(3) Why do [chíldren] avòid [végetables]?

(4) [Wíttgenstein] hànded [Ánscombe] [a glàss of [ shérry ]]

38

Page 39: Distinguishing contrastive, new and given information

Default sentence prosody: constituency interface & phonological constraints

•  Match Phrase A phrase of syntactic structure matches up with a phonological

phrase in the prosodic structure representation of the sentence. (Kratzer and Selkirk 2007, Selkirk 2009)

•  Prosodic Prominence (ϕ) Every phonological phrase (ϕ) contains a (head) stress prominence. [Assigns “default” phrase stress.]

•  No-Toneless-Phrase-Stress Every phrase stress must be associated with a tone (pitch accent). [Pitch accents may be epenthesized onto syllables with phrase stress.]

39

Page 40: Distinguishing contrastive, new and given information

( ) intonational phrase ( x ) ( ( x )) phonological phrase ( x ) ( x ) ( x ) prosodic word Why do [chíldren] [avoid [végetables]]? H* H*

H* pitch accents associate to syllables with phrase stress

Page 41: Distinguishing contrastive, new and given information

The Prosody of Givenness

•  In Standard British & American English a given phrase fails to show necessary pitch accent; in some contexts within the sentence a given constituent is necessarily “deaccented”, while in others it is only optionally accented.

Page 42: Distinguishing contrastive, new and given information

“Deaccenting” of given phrases

•  [Ànscombe] has been fèuding with [her cólleagues]. But [Wíttgenstein] bròught [some shérry] over to [Anscombe]Given. (They must have made up.)

42

Page 43: Distinguishing contrastive, new and given information

An interface constraint referring to Givenness

•  Avoid Given Stress If a constituent is G-marked, it does not contain

major phrase stress.

•  A major phonological phrase is immediately dominated by intonational phrase (Ito and Mester 2007)

cf. Destress Given (Féry and Samek-Lodovici 2006) and Ladd 1980 on avoidance of phrase stress on given

43

Page 44: Distinguishing contrastive, new and given information

( ) intonational phrase … ( x )maj( *x )maj phonological phrase … ( x ) ( x ) prosodic word … [some shérry] [to [Anscombe]G] H* H* not possible: violates Avoid Given Stress

H* pitch accent associates to syllables with phrase stress.

Page 45: Distinguishing contrastive, new and given information

( ) intonational phrase … ( ( x ) )maj phonological phrase … ( x ) ( x ) prosodic word … [some shérry] [to [Anscombe]G] H*

H* pitch accent associates to syllable with phrase stress. G-marked phrase doesn’t bear any degree of phrase stress here.

Page 46: Distinguishing contrastive, new and given information

46

Conclusion

• Marking of new information (N-marking) can be dispensed with. The prosodic properties that would have to be associated with N-marking follow from independently motivated interface and phonological principles.

•  The G-marking interface with phonology is trivial.

Page 47: Distinguishing contrastive, new and given information

What if a sentence is all-given?

•  Guess what, Amanda read the newspaper!

If [Amànda read the néwspaper]G, there must be cartoon in it.

47

Page 48: Distinguishing contrastive, new and given information

Main phrase stress in all-given sentences

A. Amánda rèad the néwspaper.

B. If AmàndaGiven rèadGiven the néwspaperGiven, there must be cartoons in it.

⇑ violation of Avoid Given Stress

48

Page 49: Distinguishing contrastive, new and given information

Why necessary final pitch accent in all-given sentences?

•  Constituency interface and phonological constraints dominate the Givenness interface constraint:

Match (Clause, IP), ProsHierchIP/PPh, ProsProm(ϕ)

> > Avoid Given Stress

49

Page 50: Distinguishing contrastive, new and given information

Constituency interface

•  Match Clause A clause must be parsed as an

intonational phrase.

50

Page 51: Distinguishing contrastive, new and given information

Phonological Constraints

Prosodic Hierarchy: IP/PPH An intonational phrase dominates at least one phonological phrase.

In addition, constraints from above requiring that –  A phonological phrase carries main phrase stress. –  That main phrase stress is optimally rightmost. –  Phrase stressed syllables carry pitch accent.

51

Page 52: Distinguishing contrastive, new and given information

( ) intonational phrase ( x ) phonological phrase ( x ) ( x ) ( x ) prosodic word If AmàndaGiven readGiven the néwspaperGiven

H*

⇑violation of Avoid Given Stress

52

Page 53: Distinguishing contrastive, new and given information

Interim Checklist

√ The ‘default’ status of the phonology of New √ A simple phonology of G-marking A straightforward semantics of G-marking √ A simple phonology of FoC-marking A straightforward semantics of FoC-marking

53

Page 54: Distinguishing contrastive, new and given information

The interface with the semantics/pragmatics

The interpretation and distribution of G-marking

54

Page 55: Distinguishing contrastive, new and given information

G-marking

55

Syntactic feature G-marking

Syntactic constraints for the distribution of the feature

Functional vocabulary is not G-marked.

Semantics/pragmatics interface constraint

An eligible constituent is G-marked iff it matches a constituent in the active preceding discourse.

Phonology interface constraint If a constituent is G-marked, it does not contain major phrase stress.

Page 56: Distinguishing contrastive, new and given information

G-marking interface constraint

•  An eligible constituent is G-marked iff it matches a constituent in the active preceding discourse.

•  Recall the ‘deaccenting’ examples ….

56

Page 57: Distinguishing contrastive, new and given information

Matching

•  A constituent β matches a constituent α in a context c iff [[α]]c,O entails [[β]]c,O.

•  Allow for a cross-categorial notion of entailment.

•  Allow for a context-dependent notion of entailment.

57

Page 58: Distinguishing contrastive, new and given information

Cross-categorial entailment

For any a, b of any type, ⇒ is the smallest relation satisfying the following conditions: (a) If a = b, then a ⇒ b. (b)  If a, b ∈ Dt, then a ⇒ b if a = 0 or b=1. (c)  If a, b ∈ D<τ, t> for some type τ, then a ⇒ b if for all c, c ∈ Dτ, a(c) ⇒ b(c).

58

Page 59: Distinguishing contrastive, new and given information

Examples

•  Jason’s being a cat entails Jason’s being an animal.

•  Being a cat entails being an animal.

•  In some contexts: being Schubert entails being the composer.

59

Page 60: Distinguishing contrastive, new and given information

The interface with the semantics/pragmatics

The interpretation and distribution of FoC-marking

60

Page 61: Distinguishing contrastive, new and given information

61

Alternative Semantics (Rooth 1992)

•  FoC-marking of a constituent α makes available

sets of alternatives for α that can be used by the semantics/pragmatics in various ways.

•  [[α]]O Ordinary denotations

•  [[α]]FoC FoCus denotations

Page 62: Distinguishing contrastive, new and given information

62

Alternative sets

[[AnnFoC praised SueFoC]]FoC =

{Ann praised Sue, Sue praised Ann, Ann praised Ann, Sue praised Sue, John praised Sue, Sue praised John, John praised Ann, Ann praised John, John praised John, etc….}

Page 63: Distinguishing contrastive, new and given information

Licensing of FoC-marking

63

Page 64: Distinguishing contrastive, new and given information

Optional Focus of Contrast

• Guess what, Amanda read the newspaper!

If [ [Amánda]FoC read the newspaper]G, there must be cartoons in it.

If [Amànda read the néwspaper]G, there must be cartoons in it.

64

Page 65: Distinguishing contrastive, new and given information

But ….

•  There are also cases of obligatory FoC-marking.

•  ‘Recycling’ cases of Schwarzschild 1999.

65

Page 66: Distinguishing contrastive, new and given information

Obligatory Focus of Contrast

•  Mary did John’s homework.

No, [ [Jóhn]G]FoC [did John’s homework]G

# No, [John]G [did John’s hómework]G

66

Page 67: Distinguishing contrastive, new and given information

Obligatory Focus of Contrast

•  John’s mother gave Mary a book.

No, [John’s mother]G [gave]G [ [Jóhn]G]FoC [a book]G.

# No, [John’s mother]G [gave]G [John]G [a bóok]G.

67

Page 68: Distinguishing contrastive, new and given information

And ….

• Sometimes, you just can’t FoC-mark! • Analogues of Schwarzschild’s Avoid F.

68

Page 69: Distinguishing contrastive, new and given information

No overFoCusing!

•  Sue invited Mary. No, [John]FoC invited Mary. # No, [John]FoC invited [Mary]FoC.

•  Büring 2005, Krifka 2007, building on Schwarzschild’s 1999 Avoid F.

69

Page 70: Distinguishing contrastive, new and given information

Accounting for the licensing of FoC-marking

70

Page 71: Distinguishing contrastive, new and given information

Matching FoC-meanings with preceding discourse

(a) (Mary) invited Sue. (b) (She also) invited [Fred]FoC.

•  The ordinary denotations of the VP in (a) is a member of the FoC-denotation of the VP of (b): the VP of (b) FoC-represents the VP of (a).

71

Page 72: Distinguishing contrastive, new and given information

FoC-representation

• A phrase β FoC-represents a phrase α in iff [[α]]O ≠ [[β]]O and [[α]]O ∈ [[β]]FoC.

72

Page 73: Distinguishing contrastive, new and given information

FoC-interface constraint: first try

• FoC-represent constituents from the active preceding context whenever you can.

• A phrase β FoC-represents a phrase α in iff [[α]]O ≠ [[β]]O and [[α]]O ∈ [[β]]FoC.

73

Page 74: Distinguishing contrastive, new and given information

Obligatory Focus of Contrast

•  Mary did John’s homework.

No, [ [Jóhn]G]FoC [did John’s homework]G

# No, [John]G [did John’s hómework]G

74

Page 75: Distinguishing contrastive, new and given information

Obligatory Focus of Contrast

•  John’s mother gave Mary a book.

No, [John’s mother]G [gave]G [ [Jóhn]G]FoC [a book]G.

# No, [John’s mother]G [gave]G [John]G [a bóok]G.

75

Page 76: Distinguishing contrastive, new and given information

Optional Focus of Contrast

• Guess what, Amanda read the newspaper!

If [ [Amánda]FoC read the newspaper]G, there must be cartoons.

If [Amànda read the néwspaper]G, there must be cartoons.

76

Page 77: Distinguishing contrastive, new and given information

Trivial FoC-representations

(a)  (Sue) [cooked dinner] (b) (She) [invited Fred]FoC

•  The ordinary denotations of the VP of (a) is in the FoC-denotation of the VP of (b). This FoC-representation is trivial, though, since the FoC-denotation of (b) is identical to the set of all VP-denotations.

77

Page 78: Distinguishing contrastive, new and given information

Trivial FoC-representation

•  If α and β are of semantic type τ, then β is a trivial FoC-representation of α iff [[β]]FoC = Dτ.

78

Page 79: Distinguishing contrastive, new and given information

No OverFoCusing!

a. Sue invited Mary. b.  No, [John]FoC invited Mary. b’. No, [John]FoC invited [Mary]FoC.

•  Both (b) and and (b’) FoC-represent (a).

79

Page 80: Distinguishing contrastive, new and given information

Smaller alternative sets are better

•  If phrases β and β’ both FoC-represent a phrase α, then β is a better FoC-representation of α than β’ iff

[[β]], FoC ⊂ [[β’]]FoC.

80

Page 81: Distinguishing contrastive, new and given information

FoC-Interface constraint

•  You must FoC-represent constituents from the active preceding discourse, you must do so optimally, but you don’t have to bother with trivial FoC-representations.

81

Page 82: Distinguishing contrastive, new and given information

82

Predicting disjoint reference

(a) Bill blamed Amanda’s father, and then [ [Amánda]G ]FoC [blamed]G [ [hím]G]FoC.

•  If him referred to Amanda’s father, (a) would be overFoCused, hence the second conjunct would not represent an optimal contrast with what the first conjunct says.

Page 83: Distinguishing contrastive, new and given information

FoC-marking

83

Syntactic feature FoC-marking

Syntactic constraints for the distribution of the feature

None

Semantics/pragmatics interface constraint

FoC-marking introduces alternative sets. FoC-marking must produce an optimal FoC-representation of preceding discourse.

Phonology interface constraint If a constituent is FoC-marked, it bears maximal stress within any prosodic domain that contains it.

Page 84: Distinguishing contrastive, new and given information

Representing preceding discourse

•  Standard British or American English employ two distinct devices for representing matches with preceding discourse: G-marking and FoC-marking.

•  For Rooth 1992, Schwarzschild 1999, and many others, there is just one such device: F-marking.

84