deriving (uni)directionality - stanford universitykiparsky/papers/berlin-anaph-new.pdf ·...

ABabcdfghiejkl

Grammaticalization Formal grammaticalization Non-conve rgence Anaphora Aspect to tense 1 / 54

Deriving (uni)directionality

Paul Kiparsky

ABabcdfghiejkl

Grammaticalization Formal grammaticalization Non-convergence Anaphora Aspe ct to tense 2 / 54

Outline

1 Grammaticalization

2 Formal grammaticalization

3 Non-convergence

4 Anaphora

5 Aspect to tense

ABabcdfghiejkl


What is grammaticalization?

1 DEF: a change “by which the parts of aconstructional schema come to have strongerinternal dependencies” (Haspelmath 2002).

ABabcdfghiejkl




2 DEF: “The change whereby lexical items andconstructions come in certain contexts to servegrammatical functions, and, once grammaticalized,continue to develop new grammatical functions”

ABabcdfghiejkl





Hopper & Traugott 2003: 18, cf. Kuryłowicz 1958,Traugott 1991, Heine, Claudi, and Hünnemeyer1991).

ABabcdfghiejkl





Hopper & Traugott 2003: 18, cf. Kuryłowicz 1958,Traugott 1991, Heine, Claudi, and Hünnemeyer1991).

3 DEF: a change which gives rise to a new grammaticalcategory [a category previously unexpressed in thelanguage]. (Meillet 1912)

ABabcdfghiejkl


Jerzy Kuryłowicz Antoine Meillet

ABabcdfghiejkl


Grammaticalization puzzles

1 What do “grammaticalization” processes have incommon?

ABabcdfghiejkl




2 What causes grammaticalization, and why is itunidirectional?

ABabcdfghiejkl





3 What explains the order in which grammaticalizeditems “continue to develop new grammaticalfunctions”?

ABabcdfghiejkl






4 If grammaticalization is unidirectional, why haven’t alllanguages converged by now?

ABabcdfghiejkl






4 If grammaticalization is unidirectional, why haven’t alllanguages converged by now?

5 What accounts for the exceptions to unidirectionality?

ABabcdfghiejkl


The (apparent) heterogeneity ofgrammaticalization

Formal grammaticalization:e.g. postpositions > clitics > case suffixes,leading to “stronger internal dependencies”, but notnecessarily to any change in function or meaning.

ABabcdfghiejkl



Formal grammaticalization:e.g. postpositions > clitics > case suffixes,leading to “stronger internal dependencies”, but notnecessarily to any change in function or meaning.Functional grammaticalization:e.g. deontic > epistemic modals, changes in bindingproperties of anaphors,leading to more/new grammatical functions, but notnecessarily to stronger internal dependencies.

ABabcdfghiejkl



Formal grammaticalization:e.g. postpositions > clitics > case suffixes,leading to “stronger internal dependencies”, but notnecessarily to any change in function or meaning.Functional grammaticalization:e.g. deontic > epistemic modals, changes in bindingproperties of anaphors,leading to more/new grammatical functions, but notnecessarily to stronger internal dependencies.Grammaticalizations that do not seem to fit eitherdefinition: aspect > tense (e.g. perfect > past).

ABabcdfghiejkl



Formal grammaticalization:e.g. postpositions > clitics > case suffixes,leading to “stronger internal dependencies”, but notnecessarily to any change in function or meaning.Functional grammaticalization:e.g. deontic > epistemic modals, changes in bindingproperties of anaphors,leading to more/new grammatical functions, but notnecessarily to stronger internal dependencies.Grammaticalizations that do not seem to fit eitherdefinition: aspect > tense (e.g. perfect > past).

We’ll look at a representative case of each type andpropose a way to unify them theoretically.

ABabcdfghiejkl

Grammaticalization Formal grammaticalization Non-convergence Anaphora Aspect to tense 7 / 54

Outline



3 Non-convergence

4 Anaphora

5 Aspect to tense

ABabcdfghiejkl


Origin of a case ending

postpositions > case clitics > case suffixes

1 *kätehand

pälV-kinside-Lative

(Finno-Ugric)

‘to the inside of the hand’

ABabcdfghiejkl




1 *kätehand


(Finno-Ugric)


2 kéz-behand-Illative

(Hungarian)

‘into the hand’

ABabcdfghiejkl




1 *kätehand


(Finno-Ugric)


2 kéz-behand-Illative

(Hungarian)

‘into the hand’

3 *pälV-k > *belV-j > *-belé > *-bele > *-be

ABabcdfghiejkl


Formal grammaticalization

Other things being equal, the learner prefers“stronger internal dependencies”.

ABabcdfghiejkl


Formal grammaticalization

Other things being equal, the learner prefers“stronger internal dependencies”.

This preference drives formal grammaticalization.

ABabcdfghiejkl


Minimalist morphology

Two components:

1 A generative component specifies the potentialexpressions of the language.

ABabcdfghiejkl



Two components:


Affixation operates freely, provided the feature contentof the affix unifies with the feature content of the stem.

ABabcdfghiejkl



Two components:



2 The competition between the potential expressionswhose meaning is compatible with a given inputmeaning (the ‘intended’ meaning) is resolved by theinteraction of (FAITHFULNESS) and (MARKEDNESS)(BLOCKING, Wunderlich 1996, Kiparsky 2004).

ABabcdfghiejkl



Two components:




FAITHFULNESS: Express the meaning of the input.

ABabcdfghiejkl



Two components:




FAITHFULNESS: Express the meaning of the input.MARKEDNESS: Avoid complexity.

ABabcdfghiejkl


A toy example: why best is best

Assume that the input (or other constraints) specifythat -est is a suffix which denotes the maximaldegree of a property and that most is a word with thesame meaning.

Input: Max(good) FAITHFULNESS MARKEDNESS

1. good *2. ☞ best3. good-est *4. most good **

ABabcdfghiejkl


Analogy from reduced input

If best is not a candidate, goodest wins:

Input: Max(good) FAITHFULNESS MARKEDNESS

1. good *2. best3. ☞ good-est *4. most good **

Logically, it would be equally possible for the complex(synthetic) form to block the simple (analytic) form.But this never seems to happen: the distributionalgeneralizations are always most perspicuously statedon the simple form.

ABabcdfghiejkl


Grammaticalization from reduced input

Suppose a learner detects no evidence for the categoryand morphological composition of bele-i. She willconsider two structures of kéz belei : as a noun pluspostposition, or as a noun plus a case affix.

Input: ‘into the hand’ FAITHFULNESS MARKEDNESS

1. [kéz]ω [bele-i]ω *2. ☞ [kéz-belei]ω

MARKEDNESS, under any ranking, guarantees apreference for “stronger internal dependencies”,which drives grammaticalization.

ABabcdfghiejkl


Grammaticalization respectslanguage-specific constraints

Grammaticalization of most as a prefix in English is notlikely to happen because English inflects only withsuffixes.

Input: Max(good) FAITH RT-HEAD MARKEDNESS

1. ☞ [most]ω[good]ω *2. [most-good]ω *

ABabcdfghiejkl


Unidirectionality

Under this regime it is strictly impossible to get“upgrading” (e.g. of affixes to clitics or clitics towords).

ABabcdfghiejkl


Unidirectionality


Thus we derive the origin of the innovations from thesame principles that determine the direction of theirspread.

ABabcdfghiejkl


Unidirectionality



Moreover, these principles also organize synchronicmorphological systems.

ABabcdfghiejkl


Unidirectionality



Moreover, these principles also organize synchronicmorphological systems.

Contrast evolutionary theories, which are only aboutselection between existing variants.

ABabcdfghiejkl


But . . . what about real cases of upgrading?!

In the late 14th century, the genitive suffix -s becamea clitic (the man I met yesterday’s wife).

ABabcdfghiejkl




Around the same time, the prefix to became anonfinite modal (to quickly say, I want to, to danceand sing).

ABabcdfghiejkl





☞ Specific constraints trump general constraints.

ABabcdfghiejkl





☞ Specific constraints trump general constraints.

Apparent degrammaticalizations always turn out toeliminate language-specific complications (Plank1995: response to “Systemstörung”). They areanalogical changes.

ABabcdfghiejkl


Upgrading as system-internal regularization

The loss of case inflections left English with thegenitive as sole case suffix. Such systems are highlymarked (no instances in Arkadiev’s 2006 survey ofminimal case systems). So genitive case inflectionwas eliminated.

ABabcdfghiejkl




The modals lacked non-finite forms. This gap wasfilled by recategorizing to as a non-finite modal. Thisis part of a much larger morphosyntacticreorganization in Late Middle English.

ABabcdfghiejkl




The modals lacked non-finite forms. This gap wasfilled by recategorizing to as a non-finite modal. Thisis part of a much larger morphosyntacticreorganization in Late Middle English.

☞The “counterexamples” to unidirectionality are opti-mizations. They follow from the same principles asunidirectionality itself.

ABabcdfghiejkl


Grammaticalization as optimization

analogical change(optimization)

exemplar-based non-exemplar-based(grammaticalization)

proportional non-proportionalanalogy analogy

ABabcdfghiejkl


Outline



3 Non-convergence

4 Anaphora

5 Aspect to tense

ABabcdfghiejkl


Why grammaticalization causes noconvergence

The original structure is eventually renewed from otherresources. Language is a STABLE DYNAMIC SYSTEM: i.e.,linguistic change is not linguistic evolution.

Short cycles: e.g. Jespersen’s negation cycle

ABabcdfghiejkl




Short cycles: e.g. Jespersen’s negation cycleLong cycles

ABabcdfghiejkl





Latin future *ama=bhw-o > ama-b-o ‘I will love’,renewed in Romance: amare habeo > aimerai,and again in French: je vais aimer.

ABabcdfghiejkl





Latin future *ama=bhw-o > ama-b-o ‘I will love’,renewed in Romance: amare habeo > aimerai,and again in French: je vais aimer.

Superlong cycles: e.g. agglutination > fusion >

isolation > agglutination . . .

ABabcdfghiejkl


Otto Jespersen

ABabcdfghiejkl


Otto Jespersen Jespersen’s cycle

. . . “the original negative ad-verb is first weakened, thenfound insufficient and there-fore strengthened, gener-ally through some additionalword, and this in its turn maybe felt as the negative properand may then in the course oftime be subject to the samedevelopment as the originalword.” (Jespersen 1917:4)

ABabcdfghiejkl


Jespersen’s cycle in English

PLAIN STRENGTHENING WEAKENING

I. ne −→ ne a ‘not ever’ −→ na ‘not’

II. ne −→ ne . . . na ‘not ever’ −→ ne . . . na ‘not’

III. ne . . . na −→ ne . . . na wiht −→ (ne) . . . naught ‘not’‘not a creature’

IV. not −→ not a bit

ABabcdfghiejkl


Strengthening and weakening

Morphological/syntactic strengthening: A plainnegation is emphasized with a focused indefinite .

ABabcdfghiejkl




Semantic weakening (“bleaching”): The emphaticnegation becomes noncompositional, and turns into aplain negation.

ABabcdfghiejkl





Strengthening adds an expressive resource;weakening eliminates it.

ABabcdfghiejkl






Strengthening is MORPHOSYNTACTIC change,weakening is SEMANTIC change.

ABabcdfghiejkl






Strengthening is MORPHOSYNTACTIC change,weakening is SEMANTIC change.

Semantic weakening can be followed by phonologicalreduction or loss of the original head.

ABabcdfghiejkl


What drives the cycle?

Functions of emphatic negation

1 Denial of a (possibly implicit) assertion.

ABabcdfghiejkl





2 Denial of a presumption or an expectation.

ABabcdfghiejkl






3 Strengthening of a negative assertion.

ABabcdfghiejkl







4 Aspectual disambiguation.

ABabcdfghiejkl







4 Aspectual disambiguation.

Assumption: All languages distinguish emphatic negationfrom plain negation.(Eckardt (2002, 2006), Condoravdi and Kiparsky 2004)

ABabcdfghiejkl


Outline



3 Non-convergence

4 Anaphora

5 Aspect to tense

ABabcdfghiejkl


Pronominal parameters

Pronouns differ systematically in how their referencemay be determined from the context.

ABabcdfghiejkl




These differences can be characterized in terms ofhierarchically nested sets of constraints.

ABabcdfghiejkl





The constraints are invariant; pronouns vary in howmuch of the constraint hierarchy they are subject to.

ABabcdfghiejkl






The binding domain of a pronoun is determined by aranking of markedness and faithfulness constraints.

ABabcdfghiejkl






The binding domain of a pronoun is determined by aranking of markedness and faithfulness constraints.

UG hypothesis (generative grammar, OT): learner’ssearch space = the typological space.

(Details: Kiparsky 2002, Gast 2006)

ABabcdfghiejkl


Referential dependence

REFERENTIALLY INDEPENDENT pronouns can (butneed not) introduce something new into thediscourse.

ABabcdfghiejkl




Deixis (pointing use): It’s him. It’s her.

ABabcdfghiejkl




Deixis (pointing use): It’s him. It’s her.Restrictive relative clauses: He who hesitates is lost.

ABabcdfghiejkl





REFERENTIALLY DEPENDENT pronouns require adiscourse antecedent.

ABabcdfghiejkl






*It’s it!

ABabcdfghiejkl






*It’s it!*It which does not kill you makes you stronger.

ABabcdfghiejkl


How far can referentially dependentpronouns go for their antecedent?

The discourse topic (non-reflexive referentiallydependent pronouns, e.g. it, German er, sie (forinanimates), Greek o idhios, Turkish kendisi, Marathiaapan

˙

ABabcdfghiejkl




˙Within the sentence (very long-distance reflexives,e.g. Icelandic sig)

ABabcdfghiejkl





Within the finite domain (long-distance reflexives, e.g.Swedish sig, Russian sebja)

ABabcdfghiejkl





Within the finite domain (long-distance reflexives, e.g.Swedish sig, Russian sebja)

First accessible subject (local reflexives, e.g. himself,German sich)

ABabcdfghiejkl


The antecedent domain hierarchy

locally bound bound reflex. ref.dep. ref.indep.

ABabcdfghiejkl


Obviation

Some pronouns can’t be coreferential with a coargument(except for certain predicates like “shave”, “wash”).Swedish sig in an obviative long-distance reflexive.

Generalen igeneral-the

tvingadeforced

översten jcolonel-the

attto

beask

löjtnanten klieutenant-the

attto

hjälpahelp

sig i ,?j ,∗kself

‘The general forced the colonel to ask the lieutenant tohelp him.’

For the local domain, obviation meanssubject-orientation.

ABabcdfghiejkl


The antecedent domain hierarchy forobviative pronouns

German sich Sw. sig Icel. sig Gk. o idhios him

ABabcdfghiejkl


Unidirectionality of change

Referentially independent (demonstrative) pronounsbecome referentially dependent (“weak” pronouns,requiring an antecedent).

ABabcdfghiejkl



Referentially independent (demonstrative) pronounsbecome referentially dependent (“weak” pronouns,requiring an antecedent).The domain in which referentially dependentpronouns must find their antecedent becomesnarrower over time.

ABabcdfghiejkl




discourse > finite domain > non-finite domain > localdomain

ABabcdfghiejkl





Non-obviative pronouns become obviative.

ABabcdfghiejkl






Reflexives become subject-oriented.

ABabcdfghiejkl






Reflexives become subject-oriented.

☞ The binding properties of pronouns become MORE

RESTRICTIVE in time.

ABabcdfghiejkl


Finno-Ugric

1 Reflexive: itse & cognates

ABabcdfghiejkl


Finno-Ugric


2 Pronoun (> Reflexive): hän & cognates

ABabcdfghiejkl


Finno-Ugric



3 Demonstratives (> Pronouns): se, tämä & cognates

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite

finite disc. None (ref.indep.)

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite


Votyak Refl Pron Pron Pron Pron

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite


Votyak Refl Pron Pron Pron PronStd. Finnish Refl Pron Pron Pron Pron ∼ Dem

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite


Votyak Refl Pron Pron Pron PronStd. Finnish Refl Pron Pron Pron Pron ∼ DemSW Finnish Refl Pron Pron Pron Dem

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite


Votyak Refl Pron Pron Pron PronStd. Finnish Refl Pron Pron Pron Pron ∼ DemSW Finnish Refl Pron Pron Pron DemTornio Finnish Refl Pron Pron Dem Dem

ABabcdfghiejkl


Finno-Ugric




Antecedentdomain

local non-finite


Votyak Refl Pron Pron Pron PronStd. Finnish Refl Pron Pron Pron Pron ∼ DemSW Finnish Refl Pron Pron Pron DemTornio Finnish Refl Pron Pron Dem DemEstonian Pron Pron Dem Dem Dem

(Viinikka-Kallinen & Trosterud 1999)

ABabcdfghiejkl


Referentially independent > referentiallyindependent

Indo-European k′e- (demonstrative) > OE he(referentially dependent)

ABabcdfghiejkl




Indo-European *ey-, -i (demonstrative) > Latin is, ea,id (referentially dependent), also Avestan a-

ABabcdfghiejkl




Indo-European *ey-, -i (demonstrative) > Latin is, ea,id (referentially dependent), also Avestan a-

Indo-European *swe- (pronominal adjective meaning“own”) has been recruited as an reflexive in manybranches. The predicted intermediate stage, areferentially dependent pronoun, is attested inRigvedic (with logophoric function).

ABabcdfghiejkl


Nonreflexive > reflexive

Classical Greek ho- was a referentially dependentpronoun in Homeric, only a (long-distance) reflexivein later Greek.

ABabcdfghiejkl


Nonreflexive > reflexive

Classical Greek ho- was a referentially dependentpronoun in Homeric, only a (long-distance) reflexivein later Greek.

Old Chinese ji, zìjia (apparently a referentiallydependent pronoun) has developed into the modernChinese reflexive zìji.

ABabcdfghiejkl


Long-distance reflexive > local reflexive

Middle High German to Modern German

. . .

. . .batasked

er ihe

sih iself

ketrencanlet-drink

dazthe

uuip jwoman

‘. . . he asked the woman to give him something todrink’. . . bat eri das Weibj ihni zu “tränken” (ModernGerman)

ABabcdfghiejkl


Long-distance reflexive > local reflexive

Middle High German to Modern German

. . .

. . .batasked

er ihe

sih iself

ketrencanlet-drink

dazthe

uuip jwoman

‘. . . he asked the woman to give him something todrink’. . . bat eri das Weibj ihni zu “tränken” (ModernGerman)Latin to Romance

Ariovistus iAriovistus

responditanswered

omnesall-A

GalliaeGaul’s

civitatesstates-A

adto

se iself-A

oppugnandumattack-Grnd

venissecome-Prf-Inf

‘Ariovistus answered that all the states of Gaul hadcome to attack him’

ABabcdfghiejkl


Non-obviative > obviative

Swedish sig

ABabcdfghiejkl



Swedish sig

Marathi aapan˙

(from Sanskrit atman, non-obviative)

ABabcdfghiejkl



Swedish sig

Marathi aapan˙

(from Sanskrit atman, non-obviative)

Rise of subject-orientation (Dogon, data fron C. Culy)

ABabcdfghiejkl


Explaining the unidirectionality

The binding constraint system includes

ABabcdfghiejkl



The binding constraint system includesConstraints on binding domains, which form astringency hierarchy, and an OBVIATION constraint.

ABabcdfghiejkl



The binding constraint system includesConstraints on binding domains, which form astringency hierarchy, and an OBVIATION constraint.A FAITHFULNESS constraint that dictates retention ofarbitrary input binding relations.

ABabcdfghiejkl



The binding constraint system includesConstraints on binding domains, which form astringency hierarchy, and an OBVIATION constraint.A FAITHFULNESS constraint that dictates retention ofarbitrary input binding relations.

Particular anaphors are characterized by a specificranking of these constraints. FAITHFULNESS

represents a cutoff-point such that constraints aboveit are strictly obeyed and constraints below it areviolable.

ABabcdfghiejkl


Swedish vs. Icelandic

Input

[ . . . Ai . . . [ . . . Bi . . . ]CP ]CP

DIS

CO

UR

SE

FIN

ITE

FA

ITH

FU

LN

ES

S

NO

N-F

INIT

E

LO

CA

L

[ . . . Ai . . . [ . . . Bi . . . ]CP ]CP * * *[ . . . Ai . . . [ . . . Bj . . . ]CP ]CP * * *

ABabcdfghiejkl



In the learner’s initial state, markedness constraintsoutrank faithfulness constraints (as in phonology).

ABabcdfghiejkl




Consequently, learners expect a strict binding system(subject to all locality constraints and to obviation).

ABabcdfghiejkl





Positive data reveals which markedness constraintsare violated in the language.

ABabcdfghiejkl






Learners promote FAITHFULNESS to defeat these.

ABabcdfghiejkl







☞ This learning bias explains the unidirectionality ofchange in binding systems.

ABabcdfghiejkl







☞ This learning bias explains the unidirectionality ofchange in binding systems.

Because the changes produce no overt change in theoutput, speaker-based accounts are problematic.

ABabcdfghiejkl


Expressiveness

The maximally unrestricted values of each parametermust be instantiated: every language must have at least areferentially independent pronoun, and a non-obviativepronoun. This ensures the possibility of markingcoreference and non-coreference in any domain.

ABabcdfghiejkl


Expressiveness forces upgrading

Old English, the personal pronouns were referentiallydependent. They are not used deictically and cannothead restrictive relative clauses. They were recruitedfor reflexive uses when the Germanic reflexivepronoun was lost.

ABabcdfghiejkl




When the Old English masc. and fem.demonstratives se, seo were lost, he, she becamereferentially independent again (upgrading).

ABabcdfghiejkl




When the Old English masc. and fem.demonstratives se, seo were lost, he, she becamereferentially independent again (upgrading).

The reflexive function was taken over by thepronoun+self.

ABabcdfghiejkl


1 þonnethen

woldewould

heoshe

ealraof all

nyhstlatest

hyher

baþianbathe

&and

þweanwash

‘then she would last of all bathe and wash herself’[having first washed the others] (Bede 4 19.318.20)

ABabcdfghiejkl


1 þonnethen

woldewould

heoshe

ealraof all

nyhstlatest

hyher

baþianbathe

&and

þweanwash


2 acand

midwith

inneweardreinmost

heortanheart

monicoften

midwith

hinehim

sprecendespeaking

smeadereflected-3Sg

‘in his innermost heart he often argued with himself’(Bede 2 8.124.22)

ABabcdfghiejkl


1 þonnethen

woldewould

heoshe

ealraof all

nyhstlatest

hyher

baþianbathe

&and

þweanwash


2 acand

midwith

inneweardreinmost

heortanheart

monicoften

midwith

hinehim

sprecendespeaking

smeadereflected-3Sg

‘in his innermost heart he often argued with himself’(Bede 2 8.124.22)

3 þættethat

nænigno

biscopabishop

hinehim

oðrumothers-DAT

forbæreadvance-SUBJ3P

‘that no bishop shall put himself above others’ (Bede4 5.278.27)

ABabcdfghiejkl


Two consequences

Because the neuter demonstrative þæt was retained,it remains referentially dependent.

ABabcdfghiejkl


Two consequences

Because the neuter demonstrative þæt was retained,it remains referentially dependent.

Because him, her became obviative, him+self,her+self became non-compositional, and thecomplex reflexives were reanalyzed as morphologicalunits.

ABabcdfghiejkl


Outline



3 Non-convergence

4 Anaphora

5 Aspect to tense

ABabcdfghiejkl


Unidirectional grammaticalization paths

RESULTATIVE > PERFECT > PERFECTIVE/PAST

ABabcdfghiejkl




Kru, Chinese, Ewe, French, Italian, German (Dahl1985, 2000; Bybee et al 1994)

ABabcdfghiejkl





(LOCATIVE) > FOCALIZED PROGRESSIVE >

PROGRESSIVE > IMPERFECTIVE/PRESENT

ABabcdfghiejkl





(LOCATIVE) > FOCALIZED PROGRESSIVE >

PROGRESSIVE > IMPERFECTIVE/PRESENT

Yoruba, Scots Gaelic, Turkish, Maa, Margi, Kui(Comrie 1976; Bybee et al. 1994)

On focalized progressives, see Bertinetto 2000.

ABabcdfghiejkl


Focalized progressive

The focalized progressive yields the set of points inthe run-time of the event.

τ (e)•

ABabcdfghiejkl




τ (e)•

Ol-i-nbe-1Sg

luke-ma-ssaread-Ptc-Iness

kirja-abook-Part

(*2(2-Acc

tunti-a)hour-Part)

‘I was reading the/a book (for 2 hours)’ (Finnish)

ABabcdfghiejkl




τ (e)•

Ol-i-nbe-1Sg


kirja-abook-Part

(*2(2-Acc

tunti-a)hour-Part)


A focalized progressive denotes a point of time,therefore does not allow phrases denoting extent oftime.

ABabcdfghiejkl




τ (e)•

Ol-i-nbe-1Sg


kirja-abook-Part

(*2(2-Acc

tunti-a)hour-Part)


A focalized progressive denotes a point of time,therefore does not allow phrases denoting extent oftime.

The Focalized Progressive in Finnish is formed withthe Inessive (internal locative) case of the SecondInfinitive -ma (roughly ‘in -ing’).

ABabcdfghiejkl


Incompatibility with stative predicates

*Pyykkilaundry

o-nbe-3Sg

loju-ma-ssalie-Ptc-Iness

lattia-llafloor-Adess

‘The laundry is lying on the floor’ (Finnish)

ABabcdfghiejkl


Incompatibility with stative predicates

*Pyykkilaundry

o-nbe-3Sg

loju-ma-ssalie-Ptc-Iness

lattia-llafloor-Adess

‘The laundry is lying on the floor’ (Finnish)

Stative predicates (whether episodic or non-episodic)do not denote points of time, therefore do not allowfocalized progressives.

ABabcdfghiejkl


Stativity and episodicity

predicates

stative non-stative episodic

non-episodic episodic

Stative predicates (whether episodic or non-episodic)do not denote points of time.

ABabcdfghiejkl


Stativity and episodicity

predicates

stative non-stative episodic

non-episodic episodic

Stative predicates (whether episodic or non-episodic)do not denote points of time.

Non-episodic stative predicates are not located intime.

ABabcdfghiejkl


Durative progressive

The durative progressive yields the set of intervals inthe run-time of the event.

τ (e)t′i

ti

ABabcdfghiejkl




τ (e)t′i

ti

I was reading the/a book for 2 hours.

ABabcdfghiejkl




τ (e)t′i

ti

I was reading the/a book for 2 hours.*The earth is being round.

ABabcdfghiejkl




τ (e)t′i

ti

I was reading the/a book for 2 hours.*The earth is being round.

Non-episodic stative predicates are not located intime, therefore do not allow durative progressives.

ABabcdfghiejkl


Imperfective: Deo’s analysis

The imperfective yields the closure of the set ofintervals in the run-time of the event under thesuperinterval relation.

tiτ (e)

ABabcdfghiejkl




tiτ (e)

t′jt′i

ABabcdfghiejkl




tiτ (e)

t′jt′i t′k

ABabcdfghiejkl




tiτ (e)

t′jt′i t′k t′l

ABabcdfghiejkl




tiτ (e)

t′jt′i t′k t′ltj

ABabcdfghiejkl




tiτ (e)

t′jt′i t′k t′ltj tk

ABabcdfghiejkl




tiτ (e)

t′jt′i t′k t′ltj tk

☞

The denotation of the imperfective is a superset ofthe denotation of the progressive. Thus the imper-fective arises by a further step in the aspect-to-tensetrajectory.

Ashwini Deo, Tense and Aspect in Indo-Aryan languages: variation and diachrony(Stanford Diss. 2006)

ABabcdfghiejkl


Direction of change

Aspect/Tense morphemes lose their idiosyncraticproperties.

ABabcdfghiejkl


Direction of change


locative focalized progr. durative progr. imperfect

ABabcdfghiejkl


Direction of change



Anaphors lose their idiosyncratic properties.

ABabcdfghiejkl


Direction of change



Anaphors lose their idiosyncratic properties.

locally bound bound reflex. ref.dep. ref. indep.

ABabcdfghiejkl


Conclusion

Grammaticalization eliminates structurally arbitrary(albeit historically motivated) grammatical restrictions.

ABabcdfghiejkl


Conclusion

Grammaticalization eliminates structurally arbitrary(albeit historically motivated) grammatical restrictions.

It is non-exemplar-based analogical change.

ABabcdfghiejkl


deriving (uni)directionality - stanford universitykiparsky/papers/berlin-anaph-new.pdf ·...

Documents