d echaine & wiltschko’s (2002) decomposing pronouns€¦ · 3.2.2 pro-˚ person clitics...

17
echaine & Wiltschko’s (2002) Decomposing Pronouns 1 Introduction Exploring the fine-grained syntax of proforms! Q: How can we account for the differences in distribution of pronouns cross-linguistically? While recognizing the existence of different pronoun types, previous analyses have uniformly treated proforms as DPs. This makes it problematic to distinguish pronouns on the basis of external syntax Q: How can syntax can “see” the internal difference of pronouns? Q: What is the internal structure of a DP? Claims made: Pronouns are complex–“the morphosyntactic status of a given pronoun type determines its binding properties” Tripartite structure: 1. Pro-NP: syntactically like lexical nouns occur in predicate position semantic constants undefined with respect to binding theory 2. Pro-φP: any FP that intervenes between N and D encodes φ-features such as number, gender, person occur as predicates or arguments semantically vacuous (only spell-out φ-features) correspond to “Condition B pronouns” 3. Pro-DP: have a true DP shell function like R-expressions occur as arguments semantically definite 1

Upload: others

Post on 18-Oct-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

Dechaine & Wiltschko’s (2002) Decomposing Pronouns

1 Introduction

Exploring the fine-grained syntax of proforms!Q: How can we account for the differences in distribution of pronouns cross-linguistically?

• While recognizing the existence of different pronoun types, previous analyses have uniformlytreated proforms as DPs. This makes it problematic to distinguish pronouns on the basis ofexternal syntax

Q: How can syntax can “see” the internal difference of pronouns?Q: What is the internal structure of a DP?

Claims made:

• Pronouns are complex–“the morphosyntactic status of a given pronoun type determines itsbinding properties”

• Tripartite structure:

1. Pro-NP:

– syntactically like lexical nouns

– occur in predicate position

– semantic constants

– undefined with respect to binding theory

2. Pro-φP:

– any FP that intervenes between N and D

– encodes φ-features such as number, gender, person

– occur as predicates or arguments

– semantically vacuous (only spell-out φ-features)

– correspond to “Condition B pronouns”

3. Pro-DP:

– have a true DP shell

– function like R-expressions

– occur as arguments

– semantically definite

1

Page 2: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

– subject to Condition C

– include φPs and NPs as subconstituents

(1) Proform structure

(2)

(3) Binding Conditions

Condition A: A reflexive pronoun must be bound in its local domain.Condition B: A non-reflexive pronoun must be free in its local domain.Condition C: An R-expression must be free.

2 3 Categories of Pronouns

2.1 Pro-DPs: Halkomelem

Cetral Coast Salish language that is predicate initial and head-markingClaim: Halkomelem independent (emphatic) pronouns are Pro-DPs

2

Page 3: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(4)

Properties:

• have D syntax and are morphologically complex

(5)

– composed of the determiner tu and the rest of the pronoun tl’o, a pro-φP that spells outperson and number features (3 sg)

– have an empty NP position which can be overt or covert; if covert, we see the “pronom-inal” use of the pronoun:

(6)

• restricted to argument position

3

Page 4: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(7)

• pro-φP tl’o can appear in predicate position, but not argument position (although this ispredicted as a property of pro-φPs)

(8)

(9) Principle of Maximal SpecializationIn a grammatical dependency relation R, select the most specialized form. A form Ais more specialized than B if A can fulfill fewer functions than B. (Koster 1997)

• due to DP status, act like R-expressions with respect to binding

– cannot function as bound variables

(10)

(11)

Q: Would pro-φP tl’o be ok in examples (10) and (11)?

4

Page 5: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

2.2 Pro-φPs: Shuswap

Northern Interior Salish language (BC) that is predicate initial. Arguments marked on the verb asclitics or agreement affixes.

Claim: Shuswap independent (emphatic) pronouns are Pro-φPs

(12)

Properties:

• have neither D nor N syntax

(13)

– Shuswap N

(14)

(15)

5

Page 6: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

– Shuswap D

(16)

• can be predicates or arguments

(17)

(18)

• act like Condition B pronouns (e.g. they/them) with respect to binding

(19)

– can function as bound variables

(20)

6

Page 7: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

2.3 Pro-NPs: Japanese

Claim: pronoun kare is an Pro-NPProperties:

• has the syntax of N

(21)

– can follow an adjective, a possessive, or a demonstrative pronoun

(22)

• due to referential properties [male] and [marriageable age], kare does support coreferencewithout showing Condition C effects

(23)

– cannot function as a bound variable (semantic constant)

7

Page 8: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(24)

Q: D&W don’t mention whether kare be used in argument position. However, this is not reallyan issue. If its a pro-NP we expect kare to occur where NPs occur, including inside DPs in argumentpositions. It just so happens that Japanese D is null.

3 Extending the system to other languages

3.1 English

3.1.1 Pro-NP “one”

• can appear in a definite or indefinite phrase

• can follow a determiner, quantifier, or modifier

(25) a. the oneb. someonec. the real one

• can replace an elliptical noun

(26) The red [car]i is more expensive than the yellow [one]i

• cannot function as a bound variable

(27) *[Everybody]i thinks [one]i is a genius.

• has no referential content; cannot support coreference

(28) *[Mary]i loves [one]i’s mother

3.1.2 Personal Pronouns: Pro-DPs and Pro-φPs

• English personal pronouns can be split up into two groups: 1st/2nd person pronouns (Pro-DP) and 3rd person pronouns (Pro-φP)

8

Page 9: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(29)

• (no account for the impossibility of combining singular pronouns with lexical nouns: *I lin-guist, *you linguist, *him linguist)

• 1st/2nd person pronouns make an over NP subconstituent available (combine with overtnouns), while 3rd person pronouns do not

(30)

(31)

• D&W argue second dialect is still consistent, because them can be decomposed into a boundD-morpheme th- and a clitic φ-morpheme ’em

(32) [Them linguists], John likes ’em.

– they acts differently, because it does not reduce to a clitic φ-morpheme *’ey

9

Page 10: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

∗ Scottish does have clitic ’ey

– th- is paradigmatic (definite determiner, demonstratives)

(33)

3.1.3 3rd person pronouns

• function as a bound variable

(34) [Every candidate]i thinks that [he]i will win.

– Q: It is more common for people to say: [Every candidate]i thinks that [they ]i will win.Is this a problem for D&W’s system?

• support coreference

(35) [John]i thinks that [he]i will win.

3.1.4 1st/2nd person pronouns

• cannot function as bound variables

(36)

• subject to Condition C—uh oh!

(37) a. Ii said that John saw mei

b. Youi said that John saw youi

– Why do pro-DPs act so differently from names, if they are DPs?

(38) *Shei said that John saw Mary i

10

Page 11: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

Possible solution

• grammar only regulates bound variable anaphora (Reinhart 1986). English DPs are quantifi-cational and undergo QR (Condition C effects are SCO violations)

(39) a. *I know hei loves Oscari.b. LF: *[Oscar]i [I know hei loves ti]

• not all English DPs are quantificational—English focused DPs and deictics do not causeSCO violations. Since 1st and 2nd person pronouns are dependent on context, they are notquantificational

(40) a. Ii said that you saw mei.b. LF: Ii said that you saw mei.

• D&W: coreference is regulated in one of two ways

1. falls within domain of bound variable anaphora—disjoint reference is forced due to SCOviolations

2. if no QR, coreference possibilities determined by pragmatic coreference

3.1.5 Predicative Status of English Pro-φPs

• 3rd person pronouns can be arguments or predicates

(41) a. [He]arg saw [her ]argb. That’s [her ]predc. *She’s [that ]pred

• but...it is also possible for 1st/2nd person pronouns to appear in postcopular position; D&Wsuggest that (42)a. is okay as an equative structure (involving two DP expressions)

(42) a. That’s [me]b. *I’m [that ]pred

• 3rd person pronouns can participate in word formation, which shows that they are propertydenoting

11

Page 12: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(43)

(44) a. *[me]-maleb. *[you]-goat

Q: what about me-time and YouTube?

Q: OED entries:

1. thou verb meaning ‘to address a person with the pronoun thou’ in OED (as late as 2006from Tribune): “I am not the kind of guy comfortable running around in a leotard anda floppy hat while theeing and thouing everyone in sight”

2. 1967—G. R. Dickson Soldier, ask Not xvii. 122: But in spite of the fact he ‘youed’instead of ‘thoued’ me, he was not happy to see me.

3.2 French

3.2.1 Pro-N Clitic “en”

• can replace a lexical noun in the context of adjectival (weak) quantifiers:

(45) a. J’ai vu trois/un livre(s)b. J’en ai vu trois/un

• can appear only with an indefinite phrase; with a definite determiner pro-N is null

(46) a. *J’en ai vu tousb. J’ai vu tous

• cannot function as a bound variable

(47) *[Chacun]isomeone

pensethinks

quethat

JeanJean

[en]ien

ahas

vuseen

‘Someonei thinks that Jean has seen themi’

• cannot support coreference

12

Page 13: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(48) *[Marie]iMarie

pensethinks

quethat

JeanJean

[en]ien

ahas

vuseen

‘Mariei thinks that Jean has seen heri.’

3.2.2 Pro-φ Person Clitics

• earlier theories have treated clitics like articles, namely D

• l-clitics

– can be pro-arguments (inflects for gender) or pro-predicates (does not inflect for gender)

(49) Jeanne la/le voit

(50) MarieMarie

estis

unea

avocate,lawyer(FEM)

etand

JeanneJeanne

le/*lait

serawill.be

aussi.too

‘Marie is a lawyer and Jeanne will be one too.’

(51) JeanJean

estis

unea

avocat,lawyer(MASC)

etand

FrancoisFrancois

le/*lait

serawill.be

aussi.too

‘Jean is a lawyer and Francois will be one too.’

– can function as bound variables

(52)

– D&W: since l-clitics have the same form as French articles, these, too, will be φ-Ps andwill behave differently than English DPs

– this is borne out by the fact that French l-article does not have a fixed interpretation(sometimes definite, sometimes generic)

(53) Jean aime le vin= i. ‘Jean likes wine.’=ii. ‘Jean likes the wine.’

• 1st/2nd person clitics

– can be locally bound and used for coreference, or used for disjoint reference

13

Page 14: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(54)

– can function as bound variables (compare to English Pro-DPs)

(55)

3.3 Reference-Tracking Systems

• obviation: obviative-marked argument is obligatorily disjoint from proximate-marked argu-ment

• switch-reference: when subject of a dependent clause is distinct from subject of a mainclause, it is obligatorily coded by different-subject marking

(56)

3.3.1 Obviation

• a way to distinguish two 3rd person participants (discourse-sensitive)

14

Page 15: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

• one is marked proximate, one obviative

• e.g. Plains Cree marks overt DP arguments with the suffix -(w)a

• obligatorily marked on possessed nouns

(57)

• D&W propose that obviation markers are of category D

– given that they respect Condition C, they must be disjoint

• leads us to predict that when there is pronominal obv. agreement, it will be an augment of“normal” pronominal agreement

(58)

• proximate φ-agreement is discourse dependent (salient discourse referent); obviate D-agreementsignals a distinct discourse referent

(59)

Q: If obviation markers are D and prevent conference because DPs dont corefer, would thispredict that ‘HePROX showed herOBV herselfOBV ’ would be ungrammatical in such a system?

15

Page 16: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

3.3.2 Switch-Reference

• a switch in subject or agent is obligatorily marked

• obligatory coreference is marked by same-subject agreement; noncoreference is marked bydifferent-subject agreement (e.g. Mojave)

(60)

• D&W propose that same-subject agreement is φ-agreement, while different-subject agreementis categorically D-agreement; respects Condition C

• similar prediction as obviation: different-subject agreement is an augment of “normal” φ-agreement (e.g. Amele markers)

(61)

• switch-reference is restricted to subjects (unlike obviation), due to tense dependency

• Tense of subclause is dependent on matrix Tense

• so, since φ is dependent on T, and the lower φ is dependent on the higher φ, we can only getsame-subject or different-subject marking (not same-object or different-object)

16

Page 17: D echaine & Wiltschko’s (2002) Decomposing Pronouns€¦ · 3.2.2 Pro-˚ Person Clitics earlier theories have treated clitics like articles, namely D l-clitics { can be pro-arguments

(62)

4 Conclusion

• Pronouns are not primitive

• Q: What consequences does this have?

→ provide a solution to the differences in distribution of proforms

→ formalize similarities and differences between obviation and switch-reference

→ should extend to nominals as well (reflexives, pro, agreement, full XPs)

17