Document 7346082

Download Report

Transcript Document 7346082

ESSLLI, Hamburg 2008
Helen de Hoop
Case marking patterns in the languages of the world
based on joint work with Andrej Malchukov
Research group Optimal Communication: www.ru.nl/optimalcommunication/
Two papers

de Hoop & Malchukov (2007) On fluid differential case
marking: A bidirectional OT approach, Lingua 117,
1636-1656.

de Hoop & Malchukov (2008) Case marking strategies,
to appear in Linguistic Inquiry.
Three violable case constraints
I. Identify: Encode internal argument properties
A specific instantiation of Identify that will be used during this
lecture:
•
Identify (A/erg): Ergative case identifies high-prominent
subjects
(A ↔ ERG).
Identify
•
sang.nyin nga-s las.ka ‘di
byed.kyi.yin
tomorrow I-ERG work this
do
‘I shall do this work tomorrow (of my own free will)’
•
sang.nyin nga
las.ka ‘di
byed.kyi.yin
tomorrow I
work this
do
‘I shall do this work tomorrow (whether I like it or not)’
Tibetan: volitional subject  ergative case
Three violable case constraints
II. Distinguishability: The two arguments of a transitive clause
should be distinguishable.
Cf. a.o., Peter de Swart (2007)
Distinguishability
• kappal
tiramaalakal-e bheediccu
ship
waves-ACC
split
“The ship broke through the waves.”
• tiramaalakal kappaline
bheediccu
waves
ship-ACC
split
“The waves split the ship.”
Malayalam: avoid ambiguity  accusative case
Three violable case constraints
III. PAIP: Avoid (case) marking of the unmarked argument.
Cf. Malchukov 2006.
Similar constraints are Tsunoda’s (1981) “Unmarked Case
Constraint” and Bobaljik’s (1993) “Obligatory Case
Parameter”.
PAIP
•
Onkrutil rulj.
he rotate wheel-ACC
‘He rotated the wheel (consciously).’
•
Onkrutil ruljom.
he rotate wheel-INSTR
‘He rotated the wheel unconsciously.’
Russian:the effect of PAIP
Case and voice
Legendre et al. (1993): prominence distinctions trigger
voice/case alternations
INPUT transitive construction: AP
INPUT passive: aP
INPUT antipassive: Ap (= Comrie’s “natural transitive”)
Antipassive
Jaaku-p
arnaq tuqut-p-aa
Jacob-ERG
woman
kill-IND-3sERG/3sNOM
‘Jacob killed the woman.’
Jaaku arna-mik
tuqut-si-v-uq
Jacob
kill-AP-IND-3sNOM
woman-INSTR
‘Jacob killed a woman.’
Passive formation
Is the passive a language universal?
No!
There are many languages in which passive formation is not
attested.
Languages without passives:
Dyirbal, Lezgian, Tongan, Samoan, Hungarian, etc.
Case and voice
• Legendre et al. do not account for the fact that passives are
found mostly in nominative-accusative languages (while
antipassives are found mostly in ergative languages).
Passive formation
Conflict between two constraints:
• Mark prominence distinction in A (faithfulness to the input)
• PAIP (do not mess around with the unmarked argument)
A voice alternation (passive formation) can be the optimal
outcome of this conflict.
Passive formation
Passive formation: the effect of PAIP
• A passive applies when the subject is indefinite, nonspecific or not important in the discourse (demoted agent).
• In some languages passive forms are used to indicate nonvolitionality of the subject (cf. Masica 1991 on Sinhala and
Dhivehi).
• By using the passive construction, the object is promoted to
the function of subject and hence becomes the unmarked
argument (in the nominative case), thus satisfying PAIP.
Differential subject marking
• In ergative languages, on the other hand, a non-volitional
subject may lead to a differential subject marking pattern. In
ergative languages marking a prominence distinction in A
can affect the form of the subject exclusively (DSM):
satisfaction of PAIP.
Differential subject marking
Lezgian (Haspelmath 1993)
Zamira-di
get’e xa-na
Zamira-ERG pot
break-AOR
‘Zamira broke the pot’
Zamira-diwaj get’e xa-na
Zamira-ADEL pot
break-AOR
‘Zamira broke the pot accidentally/involuntarily’
Passive & DSM
• Passive: low-prominent subject (a) (accusative languages)
• DSM: high/low-prom subject (A/a) (ergative languages)
Antipassive & DOM
• Antipassive: low-prominent object (p)
(ergative languages)
• DOM: high/low-prom object (P/p)
(accusative languages)
Case and voice alternations
INPUT
Nom-acc
Erg-abs
A/a
Active/passive
DSM
P/p
DOM
Active/antipassive
Two functions of case-marking
The identifying and disambiguating functions of case-marking
can go hand in hand.
• IDENTIFY: Mark the high-prom object
• DISTINGUISH: Mark the high-prom object (in order to
distinguish it from the subject)
DOM
• Raam
ek
bakraa
Raam
one
goat.NOM
‘Raam sells a goat’
bec-taa
sell
hae
is
• Raam
ek
bakre-ko
Raam
one
goat-ACC
‘Raam sells the goat’
bec-taa
sell
hae
is
Two functions of case-marking
DOM
DISTINGUISH
IDENTIFY
P-marking
p-marking
*
*
Two functions of case-marking
• IDENTIFY: Mark the high-prom subject
• DISTINGUISH: Mark the low-prom subject (in order to
distinguish it from the object)
DSM in Dyirbal
1st, 2nd person
3rd person
Transitive subject

-ŋgu
Transitive object
-na

DSM in Dyirbal
SUBJECT
DISTINGUISHABILITY
PAIP
ERG, A
*
 ERG, a
*
 , A
, a
*
DSM in Manipuri
•
Əy-nƏ
tebƏl-dƏ
theŋŋi
I-ERG
table-LOC
touched
“I touched the table (volitionally)”
•
Əy
tebƏl-dƏ
I
table-LOC
“I touched the table”
Manipuri, Bhat & Ningomba 1997
theŋŋi
touched
DSM in Manipuri
SUBJECT
IDENTIFY (A/ERG)
 ERG, A
*
ERG, a
, A
 , a
PAIP
*
*
*
Two functions of case-marking
DSM
DISTINGUISH
A-marking
a-marking
IDENTIFY
*
*
Two functions of case-marking
• DOM: the effects of both types of constraints converge
• DSM: the two constraints conflict
We predict more cross-linguistic variation in DSM as compared
to DOM
This prediction is borne out
Differential Subject Marking
Edited by Helen de Hoop and Peter de Swart (2008)
Springer, Dordrecht
Differential Subject Marking
• Variation
–
–
–
–
Pronoun/ Noun
Agentivity, animacy, volitionality
Tense/aspect/mood
Main/dependent
• Prominence
Differential subject marking
•
Baw.a
nek
Mother.ERG
milk.NOM
‘Mother poured out the milk’
•
Baw.a-f-as
nek
atuzu-ne
Mother-AD-ELAT milk.NOM
pour out-PAST
‘Mother accidentally spilled the milk’
atuzu-ne
pour out-PAST
Agul: A  ergative case; a  oblique (locative) case
Ganenkov et al. (2008)
Exceptions to the general picture
• In Inuit, the verb agrees with both the subject and the object.
Therefore,
• not only the (absolutive) object but also the (ergative)
subject has properties of the ‘primary’ (unmarked) argument
referred to by PAIP.
• Inuit disfavors DSM and uses a passive construction
instead, although passive formation is otherwise typical of
nominative-accusative languages.
Exceptions to the general picture
• Similarly, when the verb agrees with the object in a
nominative-accusative language, the object has properties
associated with the ‘primary’ unmarked argument.
• Such a language may resist DOM and turn to the use of an
antipassive instead, a voice which is otherwise almost
exclusively found with ergative languages.
• Cf. Nichols’ (1992) observation that not only ergative
languages but also “those accusative languages in which
there is agreement with the O” have antipassives.
Exceptions to the general picture
• DSM in nominative/accusative languages
– Colloquial Korean (Lee 2008)
– Turkish (Kornfilt 2008)
– Uto-Aztecan languages (Arkadiev 2008)
DSM in colloquial Korean
Choykun-ey
wuli
younghwa-lul nemwu cacwu pwa
Nowadays
we
movie-ACC
too
often watch
‘Nowadays we watch too many movies’
The rate for case ellipsis for third person subjects (a) is
significantly lower than the rate for local person subjects
(A). Similar effects for animacy and definiteness.
Lee (2008)
Colloquial Korean: a  NOM marker; A  no case
PAIP not very active: Case-markers/case-ellipsis for subjects
and objects (DSM and DOM). Distinguishability satisfied.
DSM in Turkish
• Ali kitab-ı
Ali book-ACC
‘Ali read the book’
oku-du
read-PAST
• Ali-nin
kitab-ı
oku-dug-un-u
duy-du-m
Ali-GEN
book-ACC
read-FN-3SG-ACC heard-1SG
‘I heard that Ali read the book’
Turkish nominalizations  subject gets genitive case,
independent of its prominence (Kornfilt 2008); violation of
PAIP?
DSM in Chemehuevi
• Two cases only: DIR (S/A) and OBL (P). But in subordinate
clauses the marking of S/A switches to OBL.
• John Ann-i
karıtıa-j kıaw
taya-ai-n putucugaj
John Ann-OBL chair-OBL yesterday kick-PFV-NML knows
‘John knows that Ann kicked the chair yesterday’
Arkadiev (2008)
Violation of PAIP? Violation of distinguishability.
Two types of case alternations
•
•
•
•
Cross-linguistically, a merely distinguishing function of
case is rare.
Difference between split and fluid differential case
marking:
Split: two forms are in complementary distribution.
Fluid: two forms of the same noun phrase in the same
linguistic context + two closely related meanings.
Split DSM
• Raam
ek bakraa
Raam.NOM one goat
‘Raam sells a goat’
bec-taa
sell
• Raam-ne ek
bakraa
Raam-ERG one
goat
‘Raam sold a goat’
hae
is
bec-aa
sold
Fluid DOM
• Raam
ek
bakraa
Raam
one
goat.NOM
‘Raam sells a goat’
bec-taa
sell
hae
is
• Raam
ek
bakre-ko
Raam
one
goat-ACC
‘Raam sells the goat’
bec-taa
sell
hae
is
Fluid DSM
•
Raam-ne
chiikh-aa.
Raam-ERG
screamed
‘Raam screamed (purposefully).’
•
Raam
chiikh-aa.
Raam.NOM
screamed.
‘Raam screamed.’
Fluid differential case-marking
•
•
•
•
Unidirectional OT sufficient?
Difference between split and fluid differential case
marking:
Split: two forms are in complementary distribution.
Fluid: two forms of the same noun phrase in the same
linguistic context + two closely related meanings.
Fluid case marking
•
Volitional subjects of intransitive clauses in Hindi do not
bear ergative case usually. Only with a small class of
intransitive predicates fluid differential case marking occurs
and then it corresponds to a difference in volitionality.
•
Cf. Butt and King 1991; de Hoop and Narasimhan 2005,
2008
Fluid case marking
•
Mohan
ghar bhaag-aa
mohan.NOM
home ran
“Mohan ran home.”
•
Arunaa
zamiin-par
arunaa.NOM
ground-LOC
“Aruna fell on the ground.”
gir-ii
fell
An OT syntactic analysis
Intransitive volitional
subject in Hindi
ERGATIVE
 
PAIP
IDENTIFY
volitional/ERG
*!
*
An OT syntactic analysis
Intransitive nonvolitional subject in
Hindi
ERGATIVE
 
PAIP
IDENTIFY
volitional/ERG
*!
*
Problem for OT-syntactic analysis
•
Raam-ne
chiikh-aa.
Raam-ERG
screamed
‘Raam screamed (purposefully).’
•
Raam
chiikh-aa.
Raam.NOM
screamed.
‘Raam screamed.’
Reranking? Not possible in OT
volitional subject of
‘scream’ in Hindi
IDENTIFY
volitional/ERG
 ERGATIVE

PAIP
*
*!
Fluid differential case-marking
When we have two forms and two meanings:
• The markedness principle: Marked forms are used for
marked meanings (and unmarked forms for unmarked
meanings) (a.o., Horn 1984)
• The markedness principle can be proven to result from
(weak) bidirectional OT (Blutner 2000), see yesterday’s
lecture
Solution: Bidirectional OT
Subject of ‘scream’ verbs in Hindi

PAIP
IDENTIFY
volitional/ERG
ERG, volitional
*
ERG, non-volitional
*
, volitional
 , non-volitional
*
*
Conclusions of today’s lecture
• Prominence distinctions in the input may trigger voice or
case alternations.
• Sometimes the high-prominent subject (A) is case-marked
and sometimes the low-prominent subject (a) (Identify and
Distinguishability are conflicting constraints)
• By the economy constraint PAIP, we can account for the
fact that a subject case alternation is mostly found in
ergative languages, while passive formation (a voice
alternation) is mostly found in accusative languages.
Conclusions - continued
• In an OT framework, we can account for the fact that DSM
and antipassive formation are mostly found in ergative
languages, while DOM and passive formation are mostly
found in accusative languages.
• Asymmetries between different case-systems can be
derived from the different functions of case-marking,
IDENTIFY and DISTINGUISH
• Unidirectional OT syntax cannot adequately account for
fluid differential case marking
• Bidirectional OT can!
