Grammatical relations in Malaya

8/12/2019 Grammatical relations in Malaya

1/55

GR MM TIC L

REL TIONS

AND N PHOR

I N M L Y L M

by

Karuvannul

Putllanveetti l Mohanan

SU MITTED

IN

P RTI L FULFILLMENT

OF THE

REQUIREMENTS

OF THE

DEGREE OF

~ S T R OF SCIENCE

IN

LINGUISTICS

t the

M SS CHUSETTS INSTITUTE OF TECHNOLOGY

June 9

o Karuvannur Puthanveettil Mohanan

9

The

author

hereby

grants

to

M I T

p erm is sio n to

repro-

duce and

to distr i ute copies

of

this thesis

document

in Whole

or

in part

Signature of

uthor

Cert i f ied

by ~ ~ ~ r ~ > ~ ~ < ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ y ~ ~ C ~ \ ; ~ ; ~ ~ 6 ~ o ~ ~ ~ k y

Thesfs ~ u p e r v i s o r

ccepte d

by

- - - - - ~ ~ ~ - - \ . N - - - - V \ , J - - - - > A ~ - x ~ H . . , . . . . . . . - . . ~ - { r a t i t y ~ e r

Chairman

~ ~ Q 1 ~ t a l

c ~ i t t e e

I fum .

M A S S A ~ H u s E T r s

INSTITUTE

0 TECHNOLOGY

Y lD

L\BRARtES


2/55

G R A ~ l l \ 1 A T C A L REL TIONS ND

N PHOR

IN IVI L Y L M

b y

K RUV NNUR PUTH NVEETTIL

MO N N

Submitted

to the Department

of L i n ~ u i s t i c s and Philosophy

on

8

y 1981 i n pa r t i a l fulfil:i6nent

of the

requiremel ts

of

the e ~ r e e of Master of Science

STR CT

This thesis r ~ s th at p ri nc ip le s governing the re la-

t ion between anaphors and antecedents are best stated a t

a

level

tha t encodes

grammatical

relat ions such as

subject

of

and object of.

This

level cannot be universal ly

ident i f ied

with

the

level of

configurat ional structure.

~ h e f i r s t section of the thesis presents

a

descript ion

of the behav iour of anaphors

and pronouns in

Malayalam, and

ident i f ies

those

proper t ies

of anaphora in

this

l a r ~ u a ~

tha t are of some theoret ical

in te res t

Section shows ti at

these properties recur in various other languages

such

as

Kannada,

Chinese,

Yoruba, and

Icelandic.

Sect ion

discuss-

es

the

problems that these

phenomena

pose fo r

the

Govern-

ment i n d i n ~ Theory as developed b y

Chomsky

in press) , and

s u ~ g e s t s

possible

r e v i s i o n s ~

Thesis

Supervisor. Noam

Chomsky

Title: ns t i tu te Professor


3/55

KNOWLEDGElfENTS

I am

extremely

grateful to my

supervisor

oam

Chomsky,

from whom learned to

do

syntax by o st i -

n tely

disagreeing w; th llim.

I

have benefi ted

a

great

deal from

the

llumerous discussions

I

have had wi

Joan Bresnan Ken Hale Alec l Y arantz, and Paul

Kipar

sky. I am

also

i ~ d e b t e d to a host of

people

who have

offered

me

thei r

suggetions

served

as in fo rman ts ur

col lected data for me,

spec ia l

mention must be made

of Douelas PUlleyblank James Huang Mark

L i b e r ~ a n

Mitch Marcus

Sreevas

Mandalam,

run and

Poornima. y

Wife

Tara is

responsible

for l l the

errors

since

she checked the

data c r i t i c i sed my arguments

and

pre-

sentation corrected

my

~ g l i s h and typed the

thes is .


4/55

4

TABLE OF CONTENT..

abstract

l

2

acknowle

dgements I

J

table

of

contents 4

O. INTR0DUCTION 5

1.

AN/LPHORA IN

MALA

YALAM

n I

6

9

11

14

17

21

25

)2

...........

Obligatoriness

of Antecedents

The C command Condi tion .

Disjoint Reference '

Long

Dis tance

Anaphora

General Properties

of Anaphora in

Malayalam

SUbjecthood

of

Antecedents

1.1. Introductory

Remarks

6

2

The

NoncoreferencE

Rule

.

1.3.

1 .4.

1.5.

1.6.

1.7.

1.8.

2. N PHOR IN KANNADA CHINESE

YORUBA

ND

ICELANDIC 1 35

2.1.

Anaphora

in

Kannada

2.2.

Anaphora

in

Chinese

C

,.

2.3.

Anaphora

in

Yoruba

2.4.

naphora

in

Icelandic

35

38

40

3.

N PHOR ND

G

46

References

54


5/55

GR MM TI L REL TIONS ND N PHOR IN M L Y L M

O

INTRODUCTION

Thia

t h e s i s

i s concerned

with

on e o f the

binding

con

i t i ~ in the G overnment

B inding) theory as developed

in

ChomSky

1979;

in p r e s s ) .

The p r i n c i p l e

is

s t a t e d thus.

1 Anaphors are bound in

t h e i r

governing

category.

In Chomsky i n

p r e s s ) ,

is

assumed

t h a t the

condi

t ion a p p lie s

a t

the l e ve l

o f

e -st ruc t ure

to

syntactic

con-

f igur ations, I s h a l l argue t h a t must,

inGtead,

be assum

ed to apply

to

a nonconfigurational le v e l o f representa

tio n

containing What Chomsky

calls

lexica l V P s ,

The

l e v e l containing l e x i c a l

VP s,

which

may be

c a l l e d

the

l ex i ca l s tr uc tu ra ,

is

the

l e v e l

t h a t

u n iv e rsa lly encodes

grammatical

r e l a t i o n s l i k e subject

and o b je c t.

I s h a l l

also

argue

t h a t

pr inciple

1)

should

be

revised

to

in

clude c e r t a i n parametric

options

so

t h a t

ca n

account

fo r anaphora in languages l i k e Malayalam,

Chinese,

Yoruba,

and

Ic e la n d ic .

The

f i r s t se c tio n

o f the

th e sis presents

a

descrip

tio n o f the

behaviour

o f

anaphors

and

pronouns in

Malaya

lam,

and

i d e n t i f i e s

the

p ro p e rtie s o f

anaphora

in th i s

language t h a t are o f some

t h e o r e t i c a l

in terest

S ectio n

2 shows t h a t these p ro p e rtie s

recu r in various

ot he r la n -

5


6/55

guages,

such as

K a n n a d a ~

Chinese, Yo ruba

and c ~ l a n d i c

In secti.on 3, I sha l l di.scuss the problems tha t these

pllenomena pose

for

the

bi

nding condi t ion

in 1 ) .

and

suggest

possible

solutions.

1 . ANAPHORA

IN

MALAYAI,AM

1.1.

Introductory

R e ~ a r k s

Malayalam

i s a

f ree

word order language with

the

following ~ t = VP less) clause structure. 1

2)

S

X X

V

The structure

of

Ja) is given in

3b as

an

example:

J)a.

u ~ ~ inna1e aanaye

nu i.

child-n

yesterday

elephant-a pinched

2

The

child

p inched the

elephant

yesterday.)

b. S

Adv.

V

u ~ ~

innale aanaye

nu i

1. F,)r detai led arguments

to

show

tha t Malayalam

does

not

have

a VP node, see Mohanan in press) ,

2. n

=

nominative, a = accusative, d =

dative,

and so on.

The

unmarked

subject in Malayalam is

in

the nominative

case,

and

in the dative case for a few special verbs

and

modals. Animate direct object

is

in the accusative

case,

and inanimate

in the

nominative.

The indi rec t

object

is

in the

dative case.


7/55

The

constituents

directly

dominated by S are order

free in Malayalam. ~ h u s S Y

SV

SV VS

t

e tc .

are a l l

possible

word

orders This property reveals

in teres t ing

character i s t ics

of

anapnor-antecedent

relat ions in the

language.

There are

three

types of overt elements that

par t ic i

pate

in

syntact ical ly governed coreference

relat ionships ,

i . e . elements that can take antecedents.

I

sh all re fer to

them

as pronouns, pronominal anaphors, and

nonpronominal

anaphorsl

(4)

Pronouns awarl he; awal she awar they

b. Pronominal anaphorsl

se l f

c. Nonpronominal anaphorsl 8wa- se l f-

Pronouns

and

pronominal

anaphors can take the entire

range of cases.

The

nonpronominal anaphor swa-, on the

other

hand, can

take only the

accusative

swayam

and

the

genit ive

(swantam).

The

relevant

propert ies that dist inguish these

ele-

ments from

one

another, which

is

what this

sect ion

i s con-

cerned

with,

may be summarised as follows

(5)a. a c k w a r ~ d anaphoral

Pronouns do not

allow

the i r

antecedents

to

follow them.

Pronominal and nonpronominal

anaphors do not have

this

res t r ic t ion .


8/55

b. Obligatoriness o f antecedentsl

Antecedents

in

the same

sentence

a re o blig ato ry fo r

p ro no mi na l a nd nonpronominal

anaphors.

Pronouns

do

n o t

have

t h i s

res t r ic t ion

c .

Di sj o i n t

~ f r n l

Pronouns

and

pronominal ana-

p h o rs cannot have

the ir

antec Jnts in

the

same minimal NP S

t h a t contains

the m. N onpr onor nina l anaphors do

n o t

have th is c o n d i t i o n .

d. The c command condition. The antecedents

o f

pro-

nominal

and

nonpronominal anaphors must

c command them

This does

n o t

apply to

pronouns.

e. Subjecthood o f antecedentsl The antecedents o f

pronominal

and nonpronominal anaphors

must

be s u b j e c t s .

f Subjecthood o f anaphors: Both

pronominal

and

nonpronominal a na phor s allow long dis -

tance

anaphora, i e

they

can

f i n d

antecedents

in

higher

up cl auses. ow-

ever, 8wa is allow ed to

have

long d i s -

tan ce a n a p h o r a o n ly when

i s contain-

ed in

the

s u b j e c t .

In

th e s e c

t ions t l t

f ) l l o w ,

I

sha l l demons

t ra te each

o f the prope rtie s

l is ted in 5).


9/55

9

1.2. The Noncoreference Rule

One

of the

conditions

governing

the relation bet

ween pronouns and

the ir

antecedents in

Malayalam

is

s ta

ted

as

follows

I

(6) Pronominal

Noncoreference (Malayalam)

Pronouns

cannot

precede

the i r

antecedents . )

This property sharply

dis t inguishes

pronouns f rODl .

pronominal and nonpronominal anaphors.

Whatever

be

the

c-command

relat ion

between

pronouns

and antecedents, a l l

and only those versions

in

which the pronoun follows the

antecedent are grammatical, as shown

~

folloWing

ex-

ampleSt

7

a moohan Cawan

ce

b h a a F y a y ~ DU:n.i

4

Mohan-n

his wife-a pinched

Mohan pinched

his wife.)

b.

*

Un00hante bhaaryaye] Dul-l-i

(8)a. hmoohante Ehaar;yaye]

awan

DU1 l i

b. *

[awante

b h a a r y a ~ ] moohan D U ~ l i

9 a

Cmoohan

te

bhaarya]

awane

D U ~ J i

(Mohan s wife p inc hed him.)

b.

*

[awante bhaarya]

moohane nu:J 1 i

; . The

in tui t ive

meaning of the term

antecedent

is ob

vious. In Oscar thirlks that he i s

br i l l i an t

Oscar

is the antecedent of he.

For a

formal characterisation,

see

Mohanan

(1981). - -

4. Here, as in wha

t

follows,

underl ined

NP s ir ldi ca te the

coreferent

reading.


10/55

lO a. moohane awante bhaarya

null i

10

b.

awane moohante bhaarya nu i

l l a .sl tutti aanaye

nU:J :J i

enna]s awan pararLi u

child

elephant pinched

t r l ~ . t

he

said

He said

that

tria

child

pinched

~

elephant.

b. * ~ w a n aanaye nU:J :J i enn-]s kut:k

i

paraYffiu

12)a.

k u ~ ~ i paranfiu

S

[aYlan aanaye

D l l:J ;J i

e p ~ S

The

ch ild said that he

pinched the

elephant.

b.

awan parannu S [ k u t ~ i aanaye nu:J

i

e D D ~

Note

that

pronouns

can

c-command

their

n t e ~ e e n t s

in

Malayalam,

as

shown by

8a ,

9a , and

11a .

This

property

distinguishes

Malayalam

pronouns

from Englisr

pronouns.

In

contrast to the behav iour of

pronouns, pronominal

and

n o r ~ r o n o m i n l anaphors can

precede

their antecedentsl

lJ a .

[ ~ n t e swaotam

bhaaryaye]

Maohan nu:J J i

se l f s

b aanaye

nUffi

enna] ku paranfiu

se l f

\ cf .

b

cf . l lb)

n

the

basis of these

data,

w conclude that pronouns,

and

not pronominal

anaphors,

obey ~ condition against

fo llowing antecedents,

as stated

in

6 .


11/55

1 Obliga.to r iness of r ~ t e c e d e n t s

I

shal l

assume in

this

th esis tha t

anaphors

are

u -

versal ly

characterised

as

t h ~ s e

elements

that

require

n ~

tecedents.

5

Eoth pronouns and reflexives

take

antecedents,

unlike narne

l ike

John.

The

difference

between tham is

that antecedents are

optional

for pronouns while they are

o b l i ~ t o r y

for ref lexives. A pronoun tha t does not have

an

antecedent in the

sencence

i s a deict ic pronoun,

and

a

pronoun

that dOdS

not,

is

a

deict ic

one.

6

Seen

in

th is

l i ~ h t anaphors are

a subclass of nominals

which have

no

lex ica l reference,

namely,

5 cf:

An anaphor

is

something

lexical ly

specified

as

needinu

an

antecedent . tlomsky 1979116

6.

Pronouns and anaphors are not

the

only nominals

that

take

antecedents.

Definite

noun phrases such as the

boy (as opposed to a

boy) can also

have antecedents

in the discourse, and one may

argue

tha t

they

can have

antecedents

even within sentences, as in (i):

i A

boy

and a came in , and the boy took

off

his

shoes

iremediately.

I do not quite know how the antecedentship

of

defini te NP s f i t s

in with the

~ e n e r l theory of

ante-

cedents.

I t

must

also

be

pointed out

~ h t

no t

a l l

pronouns

take antecedents,

only

defini te pronouns do. Thus, in -

def ini te pronouns l ike one, someone,

anyone,

etc . do

not take antecedents.


12/55

12

those

t t must fin d t h e i r a n t e c e d e n t s i n the ~ D E r sen-

tence

and

n o t

i n t ~ discourse, l i k e

pronouns).

We

may

l e x i c a l l y encode thi.s p ro p erty

wi

th the

fea ture ~ a n a p h o r i c

and give the

following

d e f i n i t i o n .

14 i s

[.,.

anaPhoric] i f f

t

i s l e x i c a l l y r e

quired to have an

antecedent i n th e same sen

ten ce. ?

I f not ,

i s

[-a 1aph 1ric]).

We found

i n 1 . 2 . t h a t

the

noncoreference

ru le groups

pronom inal a na ph ors and

nonprollominal

anaphors

to g eth er,

d i s t i n g u i s h i n g them from

pronouns.

The

property o f

o b l i

g a to ry a n te c ed e nt sh ip

expressed

by 14)

o ffe rs y e t

another

c r i t e r i o n

fo r

making

e xa c t l y the

same

groupi.ngl pronominal

7. Compare t h i s

d e f i n i t i o n with the one given

in ,

s a y ,

Chomsky

i n

p r e s s ) ,

which

gives a l e s s

i n t u i t i v e

notion

o f

anaphor

.

For

Chomsky,

anaphors

a re

t ho se ele men ts

which

de n o t

have

i n h e r e n t referen ce, and

pr onominals

are

those

which have

th e featu re

o f number, gender,

and

p e r s o n .

This

r a i s e s the

i s s ue

why r e f l e xi ve s l i k e i m s l ~

a re

n o t considered

to

be pronominal, sin ce

they

too, l i k e

he

and she

have

the

f e a t ur e s o f number, gender

and

person.

Perhaps,

pronominals a re those

which

are

NOT araphors

and

have the f e a t ur e s o f

number,

gender,

and

person.

But

then,

nominals l i k e

man

a l s o have these f e a t u r e s , and to exclude

them

from being

pronominal,

one has

to

sa y

t h a t

pronominals

are

nonanaphors

which

have O Y the featu res

o f

number ,

gender, and p e r s o n . This, i n turn, l e a d s

to f u r t h e r ques

t i ons

about

pronouns

Which

have

ot he r

featu res

such

a s

nearness e . g . Malayalam

awan

t h a t

h e vs iwan

t h i s

h e ) .

These pI oblems,

no

doubt

t a re

n o t insunnountable, but

none o f

these

problems a r i s e with regard

to

the c ha r a c t e r

i s a t i o n o f anaphors

~ v n

i n 14).

Observe t h a t 14)

would lead us to

conclude t h a t

PRO

i s E-anaphoric],

a s

i s n o t n e c e s s a r y t h a t PRO

shoul d

have an antecedent

in

the

same

sentence.

I

see

nc

serio u s

problems

a r i s i n g

o ut o f

t h i s

conclusion, except those

r a i s e d

by

some o f

the assumptions

t h a t

a re theory i n t e r -

n a l

to G


13/55

13

anaphors and nonpronominal anaphors, and

not

pronouns,

are

required

to

take

antecedentsl

15)a. awan aanaye

he-n

elephant-a

pinched

He pinched the

elephant.)

b.

*taan aanaye DU li

self-n

16)a.

[awante aniya:t:t

i

ur9.99i

his s i s t e r - n s lept

His s i s t e r

has

gone

to sleep.)

b . [::tante

n i y t ~ i

uralJ }i

s e l f s

c .

[swaniam aniyat t i ]

uranlJi

s e l f s

I f

15b),

16b,c)

are

embedded

in

a matrix that con-

ta ins

an

antecedent,

the

r e s u l t

i s

grammatical,

thereby

showing

that and are required

to have

an

ante

cedent in

the

same sentence.

17)a. ta

an

aanaye nu::} i

entra] k u t t i

paraiiflu

that

child s aid

b .

[ tan te/swan t

aniya

l ~ i

ura99i

e n n ~ awam oo p.)i

that

he-d

f e l t

He f e l t

that s e l f s ~ s t r

had

gone

to

s l e e p .

Gi van the d e f i n i t i o n

o f anaphora

i n 14)., wha

t

we rous

t

do in order

to account

for

this behaviour of

and

Bwa-

i s to s t i p u l a t e that they have

the

feature

~ n p h o r i c .


14/55

14

1.4.

Disjoint Reference

While

noncoreference and obligatoriness of antecedents

separate pronouns from

pronominal and nonpronominal ana

phors, the

phenomenon

of

dis jo in t

reference

separates

non

pronomina l anaphors

from

pronouns and pronominal anaphors.

The principle i s

the one

tha t allows

(lab,

c) in English,

while

blocking

(18a).

(18)a.

* O car admires

him.

b.

Oscar admires

his

wife.

c. Oscar sai

d

Mary admired

hint

The same phenomenon

i s

found in Malayalam as wella

(19)a.

* Mechan w n ~ aaraaghik'k'uDQu.

Mohan-n him

worships

Mohan

worships him.)

b.

moo han [awan te

bhaaryayeJ aal aaghi

k k unnu

his

wife-a

Mohan worships ~ i s wife . )

c. moohan

paranii.u

~ 1 ; e ; l r 1 awane aaraaghik 'k 'unn

u

n n ~

said

Mary-n him t t

Mohan s i d

tha t

Mary worshipe d

hiD .)

As in En,,?;lisll, pronouns cannot have th i r antecedents

in the same

minimal

P or S that

contains

them.

ow

we

find the

same

behaviour

in pronominal anaphors, but

not

in

nonpl-onominal

anaphors

I

8. For the

l i t r tur

on ~ i s j o i n t Reference

and

Nvncorefe

renee,

see Reinhart

1976

f Lasnik

(1976), Chomsky (1980).


15/55

20 a.

meohan

tannc

aaraaghik k unnu

b.

maohan

[ta

nte

bhaaryaye]

a .raaghik k unou

c.

meohan

parannu

[meeri

tanne

aaFaadhik k unnu

enr.raJ

-

--

(21) moohan

swayam aaraaghik k unnu

s e l f

Mohan worships himself.)

I

shall assume tha t

pronouns

have

the feature

[+ pronominal] , and th t i s

th i s

feature th t i s res

ponsible for disjoint reference.

The

principle

of dis jo in t

reference can

then

be stated as ei ther (22a) or (22b):

(22)a. Pronomals are free in

the i r

minimal governinf

cate,qory.

b. Pronominals cannot

have

thei r antecedents

within their minimal

c lause nuc leus .

(22a) and (22b) wil l be respective

f o r m u l t i o ~ s

of

disjoint

reference

in and

l ~ x i l

functional ~ r m m r

and they

do

not

make

the

same

empirical predict ions .

I

shall not go into these issues

here.

The l ssif i t ion

of Malayalam pronouns, pronominal

anaphors,

and nonpronominal anaphors can

now

be

given

as

follows

(2J)a. Pronouns [-anaPhoric, p r o ~ o m i n a l

b. Pronominal anaphors [+anaphoric, +pronominaI]

c.

Nonpronominal anaphors

[+anaphoric,

-pronominal]

d. Nouns

[-anaphoric, p r o n o m i n ~


16/55

I t

may be

instruct ive

to point out

a con tra st b et

ween

n g ] ~ i s h and Malayalam wi

th respec t

to

the phenomenoll

of disjoint

reference in

inf ini t ival

clauses. Thus,

as i s

well known,

24a)

and not

24b) i s

possible

in

English.

24)a. John expects

that

he would win.

b.

John expects him

to

win.

In M a l a ~ l a m on the 0ther hand, the pronoun-antece

dent

re la t ionsh ip

i s

possible

in

both

f ini

te

and

in.f ini

ti--

val st ructures

25)a. moohan [awan bllgghimaan aa ).a

eooaJwicaariccu

he-n

in te l l igent

i s that thought

Mohan thought

that

he was

in te l l igent .

b.

moohan [awan

bugghimaan aawaan] aagrahiccu

become-inf.

desired

Mohan wanted him

to

become

in te l l igent .

Given the

fact

that rnoohan is not

contained

in the

m i n i n ~ l

S

that contains awan, 22) and

2,3)

to ge th er c or

rectly predict the

pronoun-antecedent

re la tio n in 25).

Some addit ional statement

will

have to be made

about

the

contrast

between

24a)

and

24b) in

Enflish.

9. See the discussion of sentences

l ike

Johnwas sur -

prised for him to be l e f t out , and

I t surprised

John for him

to

be l e f t out- in section 2.4.


17/55

17

1.5. The C-command

Condition

We

shal l see

in this sect ion tha t and Bwa

in

Malayalam,

but not

the

pronouns,

obey th e fo llowing

universal

principle

about

t i le

antecedents

of

anaphors

26 Anaphors

must

be

c-commanded

by t h e i r

n t e c e d e n ~ s

I hava a l r e a dy shown tha t pronoulls do no t 0

bey

2 6 ) .

In

fact ,

as

examples 8a), 9a),

and

l la) demonstrate,

the

pronoun

can asymmetrically c-command i t s antecedent

in

Malayalam.

What I must now show i s tha t taan and 8wa

must

obey 2 6 ) .

27)a.

tante

aniya tiye

s e l f s

s i s t e r - a

ut l l

nul: .i

child

pinched

The child pinched s e l f s

s i s t e r . )

b.*

kuttiyute

aniyat t iye

c h i l d s

taan n u ~ : } i

s e l f - n

28)a.

~ aanaye

n u ~ i

enna]

s e l f - n

elephant-a pinched

that

ut t

child

raajaawinoo1;a

king- . ;

paraiffiu

said

The child

told the

king

that s e l f pinched

the

elephant.)

b.*

[ k u 1 ; ~ i

aanaye

nU:J i

erma]

t.aan

raajaawinoo11

child-n

s e l f - n

paraffilu


18/55

29

[[iante

k U ~ ~ ~

aanaye

nUfli

s e l f s

child elephant-a

pinched

18

e n n ~ J

th t

mother

~ m O O h a n

Mohan

ummaweccaJ

61riiyoo1;al

paranfiu

kissed-

r e l . p a r t . woman-d s a i d

Mother to l

the womarl whom

Mol,an kissed th t

mother s/

*Mohan s/ *woman s

chi ld pinched

the elephant . )

~ w a n t a n l ku t i]

aanaye

Dull i ~ amma

s e l f s

[tnOOhan urnmaweccaJ s ~ r i i y o o ; ~

parannu.

(Reading

as in

29

I t

must be

mentioned

tha t there are

cer tain

possess-

iva constructions in which the c-command restr ict ion

seems

to be

relaxed. Compare

the following

examples:

(Jl)a.

moohante

wiswaasarn

Mohan s

b e l i e f

s e l f

brav

i s

tha t

i s

(Mohan s b e l i e f i s

tha t s e l f i s

brave.)

b.* rnoohante makan

h i i r n ~ ~ enna paranfiu

Mohan s son

s e l f brave i s

that said

(Mohan s son said

that

s e l f i s brave.)

The fac t tha t

31b i s ungrammatical

suggests tha t

the

relaxation

of

the

c-command

r e s t r i c t i o n

in

(J la)

i s

a

special

property of

nouns

l ike wiswaasam . belief ,

abhip.raayam opinioll

lt

,

laoonal

f e e l i n g . e t c .

a l l o f

which are nouns that a s s e r t proposit ions.


19/55

9

Now

a t

Some

level

of

representat ion,

we

may say

t ha t

x s be. l ief

t ha t

S ,

X iS be l i e f

i s

tha t S ,

and

x believes that

S

have para l le l

structures ,

l1amely,

the

one

in

which

x

is

the

subject of

bel ieve,

and

ts

complement.

TIle

technical

detai ls

tha

t map

x

s bel ie f

i s

tha t

S

onto

believe

(x,S) are not

quite clear to

me,

but i f ,

a t the

relevant

l evel

o f representat ion, Mohan s

bel ie f

i s

tha t

. . . . . i s

represented as

having the

same

structure

as

Mohan

believes that . . . . .

, we

have an

expla-

nation for

the

contrast

between

(Jla)

and

(Jib).

Note

tha t

t

is only when

the

head (pelief) i s

pre-

dicative tha t i t s possessive (Mohan s)

i s allowed to be

the antecedent of

Compare

31a) with (32),

rakf?iccu

saved

32

moohante wiswaasam

tanne

Mohan s bel ief / fa i th se l f

(Mohan s faith saved se l f . )

One may,

in

fact , suggest tha t

Mohan s bel ief but

not Mohan s son is a clause nucleus (cf . Bresnan ( in press))

or a

lexical

S (cf .

the notion of lexical

VP in Chomsky

(in

press , even

though configurationally, both are NP s.

A clause

nucleus

may

be

defined,

following Bresnan,

as

consist ing of a predicate argument

s t ructure .

Alternately,

one may define a lexical S as consist ing of a lexical VP


20/55

20

and the

NP that

i s associated

with i t

10

Thus,

the

con-

t r a s t between (J la)

on the one hand,

and

(J1b)

and

32

on

the

other, may

be represented a t the relational or

lexical level

as follows.

JJ a.

=

la

clause

Mohan believe clause

L

~ l

i s brave

b. = 1 b

c

=

32

clause

ohan s son

say

clause

L

se l f is brave

X J

clause

clause se l f save

Mohan

believe

x

A

I f the

suggestion

given above i s correct , then the

10.

See

the discussion of

lexical VP and lexica l S in

section

J.

f-command and I-command may be

thought

of as notions para l le l to c-command except

tha t

they

are defined a t the

levels

of f(unctional)

s t ruc-

ture

( in

lex ica l i s t functional grammar

and

l (exical)

s tru ctu re (in G

respectively.


21/55

21

condition o f c-comrnand i n 2 6 ) ,

which

i s a property o f

the

ca t e g o r i a l o r

configura

t i o n a l

l e v e l ,

should

be repla .c-

ed

by

a c o n d i t i o n

o f If-command

o r

I-command .

which

would beAcondition

a t ~ l e v e l

o f f - s t r u c t u r e c f . Kaplan

and Bresnan in press

or

o f l e xic al and l e x i c a l S.

I s h a l l n o t pursue th ese i s s u e s any f rth er i n t h i s t h e s i s .

1.6.

Subjecthood o f Antecedents

The

ai m o f

t h i s

s e c t i o n i s

to show

t h a t

the follow-

ing p r i n c i p l e holds i n Malayalama

34) Antecedents o f anaphors

must

be

s u b j e c t s .

Th e p r i n c i p l e i s i l l u s t r a t e d by

th e

fo llo w in g exam-

ple s l

35)a.

k ~ l l

tante/swantarn

a n i y a t:tiye

n u l l i

-

c h i l d - n

s e l f s

s i s t e r - a

pinched

The

c h i l d

pinched

s e l f s s i s t e r .

)

b .*

k u t t i y e t,ante/swantam a n i y a 1 t i

Du:J.li

c h i l d - a

s e l f s s i s t e r - n pinched

I n

35a),

k u t t i i s the

s u b j e c t ,

and in J5b),

the

o b j e c t . Hence,

34)

allows

k u t t i

in J5a)

but n o t i n 35b)

to

be the antecedent o f ~ t o r s w a t a ~

The same

p o i n t

i s

i l l u s t r a t e d

by

J6a,

b


22/55

36 a [taan aanaye

s e l f elephant-a

n u ~ ~ i

~ raajaawa

pinched t h a t king-n

22

m i r i y o o ~

p r nfiu

minis ter-d s a i d

The

king told the minister

that s e l f (king

*minister)

pinched

the elephant.)

b [taan aanaye

nu:r i

raajaawinoo

king-d

m n ~ r i

parannu

minis ter-n

s a i d

The

minister

told

the king

that

s e l f

(minister/

*king) p inched the elephant.)

Is t possible to characterise the

phenomenon

i l l -

ustI ated

in these sen tences i n tt rms

of

a

condi t ion on

the case of the antecedent of the NP?

The answer

i s no,

by

c a s e what we

mean

is

overt case.

Thus, even though

the

antecedent

i n

35a

and

36

happens

to

be

in

the

nominative

case, this i s not

a

requirement, as

dative

sub-

jects can

be antecedents of anaphors. Consider, for

exam

pIe ,

wha


23/55

2)

J7)b.

j o o ~ i k k a

meeriye

~ n t e s w a n a m w i i ~ ~ i l wecc

Jo h n -d Mary-a

s e l f s

h o u s e - l

a t

umm

w e k k ~ m

k iss

place-wants

John wants to

kiss

Mary

a t s e l f s

J o h n s /

*Mary s house.)

Even

though dative

su b je c ts

ca n

be antecedents

o f

anaphors,

dative i n d i r e c t obj e c t s cannot

a

38 j o o ~ i

meerikka

~ n t e s w a n a m w i i ~ t i l weCC3

John-n Mary-d

s e l f s

house-l a t

oru pus takam ko

t tu

on e book gave

John

gave M ary a

book a t s e l f s

John s/*Mary s)

house.)

Therefore, w conclude t h a t the condition

governing

the antecedents o f

anaphors

cannot

be

s ta t e d i n terms

o f

a

condition on the ove r t case

o f the

an teced en ts.

The

next

question

i S i s

pos s i bl e to

s t a t e the

condition

i n

terms

o f

semantic ro le s such

as

ag en t and theme? Once ag ain ,

the

answer

i s

no. The

c r u c i a l

examples a re to be found in the

i nt e ra c t i on b e ~ e e n anaphor binding and c a u s a t i v i s a t i o n

and p a s s i v i s a t i o n . I s h a l l

assume

t h a t passive

i s

a

rule

t h a t

promotes

an

o b je c t

to

Subjecthood.

Consider

the

e f f e c t

o f the

s h i f t

o f subjecthood on anaphorsl

11.

For

the

d e ta i ls o f

p a s s i v i s a ~ i o n

an d

c a u sa tiv isa tio n

in

Malayalam,

see

Mohanan in pr ess) .


24/55

4

(J9)a.

j o o ~ i

meeriye tante/swantam

w i i ~ ~ i l

wecca D u ~ l i

John

Mary-a

3 e l f s

house

a t pinched

(John

pinched

Mary

a t

s e l f s

(John s/*Mary s)

house.)

b. j o o ~ i y a a l

meeri

John-instr Mary-n

~ n t e / s w a n ~ w i i ~ ~ i l wecca

s e l f house-l

a t

D U ~ f a p p e ~ F

p i n c h ~ p a s s . p a s t

(Mary

was

pinched by John a t Mary s/*John s

house.)

Since the seman tic

roles of

jooQi and

meeri

are

pre

sumably

the same ;.n 39a and (.39b), a ~ o n j i

t i o n

on

the

semantic r o l e s o f antecedents

w i l l not

l e able to account

for

the

contrast . The same

point holds for

causativisation,

in which a new

subject

i s

i n t r ~ m l c e d

and the original

subjec t

i s

e i ther

changed

to an object 01 in to an

i n s t r u -

mental

d jun t l

(40)a. k u ~ t i ~ n t e s w a D ~ a m w i i ~ ~ i l wecc;

urauui

child-n

s e l f s house-l a t s l e p t

(The

c h i l d

s lept

a t s e l f s house.)

b. rom k u ~ t i y e

mother

c h i l d a

~ n t e ~ s w a D : t a m wii

~ ~ i l weco.

urakki

s e l f s house a t s l ~ e p

caused

The

mother

made

the

child sleep

a t

s e l f s

(mother s/*child s) house.)


25/55

41

a.

auseeppa

ouseph-n

j o o ~ i y e k k o ~ ~ a meeriye ~ a n t e s w a D ~ a m

John-a with

Mary-a s e l f s

w i i ~ t i l weCC3

umma

weppiccu

house-l

a t

kiss

place-causa-past

Ouseph made John

kiss

Mary a t s e l f s

(Ouseph's/*John's/*Mary's) house.(cf.37a

On

the

basis of these facts , we

are

j u s t i f i e d in

con

cluding tha t i s

the

subjecthood of antecedents, not

the ir

case

or semantic role

that

govel ns

the

antecedent

anaphor

relat ion.

1.7. Long Distance Anaphora

As

the

reader

must have

already

noticed, anaphors in

Walayalam can

have

antecedents

which are not in the

same

clause f i n i t e or n on fin ite ), in contrast to the s i t u a t i o n

i n , say, E n ~ l i s h Except

in marginal

cases l i k e

'rhey

think

th at p ic tu re s

of

each other

are

on sale , anaphors

in Envlish do not

cross

clause boundaries. ~ h u s 42a i s

u n ~ r a m m a t i c a l

while

the

c o r r e s p o n d i n ~

sentence in Malaya-

lam, (42b) , i s perfect ly grammatical:

(42)a.

John thought t h a t himself

was a

fool.

b.

aan

w i 9 ~ i

aal .a

eDDd

jo0t:l

i

wicaariccu

s e l f fool-n

i s

t t

John

thought

The

antecedent

can be

removed from

the anaphor

by

any

number

of clauses, as

demonstrated

by t he fol lowing :


26/55

6

43 sEsrs[ta

an

aanaye Dul l i

e n ~ ]

aroma

s e l f

elephant pinched that

mother

h n o o ~ parannu ennaJs raajaawini

father said

tha t

king

oonni eooaJ

S

f e l t tha t

man riye

r ~ w i ~ w s i p p i u

minister queen believe-caused

The queen convinced

the

minieter that the

king

f e l t tha t

the

mother

told the

fa ther tha t

s e l f (queen/*minister/king/mother/*father)

pinchAd the elephant.)

Queen , k i n g , and mother

are subjects,

and

there-

fore,

the pronominal

anaphor

taan

can

be coreferential

with any

of them but

not

with

the

direct object

mdnister

or

the

indi rec t

object

f a t h e r .

With

respect

to

the

p o s s i b i l i t y

of long

distance

anaphora, pronominal

anaphors d i f f e r

crucial ly from non-

pronominal anaphors. The former can

have

long distance

anaphora whatever be the

grammatical

function of the ana-

phor;

the l a t t e r ,

on

the

other

hand,

i s

allowed to have

long

distance

anaphora only

when the

anaphor

i s contain

ed in the

subject.

I f

i s

contained by a nonsubject,

the

nonpronominal

anaphor must have i t s

immediate

subject

i . e . , the subject of the same clause) as

i t s

antecedent.


27/55

7

44 a. [[fiante/swan1am sUhra:t:t3 aanaye

n U ~ J i ~ D n ~

s e l f s fr iend-n elephant pinched that

aroma h n o o ~

mother father-d

parannu o o ~

s a i d t h a t

raajaawina

king-d

tnonni

n n ~

man:triye

f e l t that

m i n i s ~ e r

r ~ i wiswasippiccu

queen

believe-caused

(The queen

convinced the minister that the

king f e l t t h a t the

mother told

the father

that s e l f s (queen s /*m inis ter s /king s /

mother s/*father s) fr iend pinched tne ele-

phant .

b. ~ a a n a ian

te

suhrat. tine

nuJ..:J.i

enIJC1] aroma

elephant

s e l f s

fr iend-a

pinched

that mother

acchanoota parannu

fa ther sa id

e L l ~

r j w i n ~

t oomJi

tha t

king

f e l t

e rr;; l]

r ~ i

t t

queen

m a n ~ r i y e

w i ~ w s i p p i u

minister

believe-caused

The queen

convinced

the minister tha t

the

king

f e l t

tha t the

mothsr

told

the

father

tha t

the elephant pinched s e l f s (queen s/

m i n i s t e r s /

k i n g s / m o t h e r s J f a t h e r ~ s /

elephant s)

f r iend.)


28/55

28

rill-ana

swan1am suhrattine

n

1 i

elephant s e l f s

fr iend pinched t h a t

~ ~

a c c h a n o o ~ d parannu e n D ~ r a a j a a w i n ~

mother

fa ther sa id tha t king

oonni e n D ~ ]

r a a ~ i

man riye wiswasippiccu

f e l t

that queen

minister

believe-caused

(The queen

convinced the minister t h a t

the

king

f e l t

that

the

mother told the father

tha t the elephant pinched s e l f s (*queen s/

*minis

t e r

s /*king s /* fet

t ~ e r

s/*mo

t h e r s /

elephant s)

fr iend.)

Note

t h a t i n

(44,), the

NP 6wantam

6uhrattine i s the

object of

the

clause. Hence, 3wa- i s forced tc find i t s

antecedent

in t ~ subject of i t s

own clause.

In

(44a),

on

the

other hand,

since

swau tam.,j uhr-att3

i s

the sUbject

o f

the

clause,

there

i s no such

l oc a ii ty r e st ri c ti o n on ante-

cedentship.

44b

shows

t h a t

there

i s

no

r e s t r i c t i o n

on

long distance

anaphora for

taan.

The immediate technical problem

th at a rise s i s

the

exact formulation of the

condition

tha t allows long

dis

tance

anaphora for

6wa-.

In

(44a),

where

shows

long

distance anaphora,

i s

immediately dominated

by

t

Bub-

j e c t Immediate

domination,

however, cannot be the r ight

condition,

as

i l l u s t r a t e d

by (45)1


29/55

29

45)a. [ [ ~ w a n t a m sUhrattinte] b h a a r y a Y U ~

amma]

s e l f s f r i e n d s wife s

mother

aanaye nU:l- i e o o ~ raajaawin_

l

oo

0 0 i

erl

r a a ~ i

m a n ~ r i y e wiswasippiccu

Th e

queen

c o n v i ~ c e d

the m i n i s t e r

that the

king fe l t

tha t

se l f s q u 6 e n s / ~ i n g s )

f r i e n d s w i f e s mother pinched the e l e p h a n t .)

t

erma

b .

, . .

~ O P

N P s u b j)

V

Poss

P S ~

P O ~ N

s w n ~ m

s u h r ~ ~ ~ n t ~ h r y y u ~ e

a r r ~ a

aanaye n u ~ ~ i

In 45),

6wa i s

s e v e r a l

nodes

removed from the s u b j e c t

NP and yet

i t shows long

d i s t a n c e

binc1i ng. Hence immediate

domination

cannot

be

the

condition p e r mi t t I n g long distance

b i n d i n g . On tt l )

other

hand, the c ondi t i o n tha t be domi-

nated b y the s u b j e c t

i s inadequate,

as shown b y

46).


30/55

46 a. f}wan:tam suhra:ttine Du: .iya] kU tti] aanaye

s e l f s fr iend pinched-REL chi ld elephant

~ i c c u e ~

raajaawina

toonni

enga r ~ i

beat

t h t king f e l t t h t

queen

man1riye w i ~ w s i p p i u

minis ter believe-caused

The

queen convinced

the minister

tha t

the

king f e l t

that

the

child

who pinched s e l f s

(chi ld 's /*queen 's /*minister 's /*king 's) fr iend

beat the elephant . )

v

ennd

b .

s

NP

NP

S

COIVIP

~

raajaawina

tccnoi

eDua raal i. mantriye

wis-

wasippiccu

S

COIVIP

NP(subj)

NP

S ------------ . : . NP

~

NP V

PosS- - - N

swaniam suhra tine nU:I: iya k u t ~ i a

naye

a ticcu

What the

contrast

between 45 and 46 i l l u s t r t e s

i s

that

long

distance

anaphora

i s

possible

only

8wa-

i s

dominated by the subject with no intervening nodes which

ara not NP's. In (46), even

though

the subject dominates

swa- t

there

i s

a n intervening S node which makes 8wa- in-

capable

o f

long distance anaphora.


31/55

)1

The

c o n t r a s t

between (45) and46) can be accounted

fo r

by using the notion o f NP -containment defined thus,

47 0 NP-c onta ins 1 i ) a

i s

o r i i )

0

domi

na

te s

(

wi

th

no

i n t e r v n n i n g

non-NP

nodes.

12

12. The

n o t i o n

o f

NP-containment, I think, ie

u s e f u l

i n

o t h e r

a r e a s

o f

grammar

a s

w ell. Thus, i n order to

account f or c on tra sts such

as in

i) and i i ) , th e

notion

o f

weak c-command derived from

Higginbotham

(1980

i s proposed

i n Mohanan 1981).

i ) a . Everyone i s upset by h is

f a i l u r e s .

b.

?

Everyone s

fai.lures u p s e t him.

c . ?

Everyone s

fa1;her s f a i l u r e s u p s e t

him.

d.

*

Fa i l u r e s

o f

everyone

u p s e t

him.

i i ) a . Who i s

u p s e t

by h is f ail u re s?

b.

Whose

f a i l u r e s

upset him?

c .

? Whose f a t h e r s f a i l u r e s upset him?

d.

*

Fa i l u r e s

o f whom

up set h is

mother?

The p ri n c i p l e s

t h a t

account fo r the c o n t r a s t a re given

belowl

i i j ) Strong Cross Overa Q uantified antecedents

must weakly) c-command

pronouns.

i v)

0

weakly c - commands 3 a) oJ c - comrnands

o r

b

the

node

th at d ir e ctly

dominates

0< weakly c-commands .

Given the

notion

o f NP-containment, iv ) can be

r e f o r

mulated as follows

v weakly c-comroands

i f f t he b ra nc hi ng node

tha

t - c o n t a i n s 01 domina

t e s

and 0 .

does

no

t

domina te

J

The notion

o f

NP-containment

i s

a l s o u s e f u l to account

fo r the follo win g c o n t r a s t ,

pointed

ou t i n Mohanan 1981).

v i ) a .

? His n e t h e r

hates

John.

b.

The-Professor

on

h is

committee

hates

John.

The r e l e v a n t param etricised p ri n c i p l e

t h a t

accounts

fo r the c o n t r a s t i s as follow s.

v i i ) Pronouns

cannot weakly)

c-command t h e i r

antecedents.


32/55

3

In (45),

the

subject NP-contains ~

in

(46),

does

not. We

can now fo rmulate the

principle governing

long

distance anaphora

of

as follows

48)

I f

8wa-

is

not

NP-contained

by

the

subject,

must

find i t s

antecedent

in i t s minimal clause

nucleus.

1.8. General Properties

of

Anaphora

in

Malayalarn

The special features

of

anaphora

in

Malayalam that

deserve

some

theoret ical

at tent ion

can

now

be summarised

as follows.

irst

anaphora in

Malayalam

does not

exhibi t

the generalisation

that anaphors are bound where pronouns

are

free .

That i s

to

say, unlike what

has

been

claimed

for

English, the domain in which the

princip le of dis-

jo in t reference operates i s not ident ical

to the

domain

in

which

anaphors

are

required

to

find

thei r

antecedents

,1)

The domain in which

dis jo in t reference

applies in Malayalam

is

t minimal S,

NP containing

e

pronominal

the

domain

in which -the anaphor

i s

required to f ind ts antecedent

is

the

ent i re sentence. As

a

resul t

one

finds

that a

pro-

noun

and

an anaphor

in

the

same

s t ructura l

posi t ion

can

have

the

same

antecedent.

13-

cf.

Chomsky 1979);

Fiengo

Higginbothom (forthcoming).


33/55

wiQ9hi aart a

fool-n

is

49)a. k

U

l ti a m m a y o o ~ ~ ~ n w n aanaye n U l ~ i ennaJ

c h i l d mother s e l f

he elephant pinched

t h a t

p r ffiiu

said

The

child

told

the

mother

th at s elf/h e

pinched

the

elephant.)

b . kalJ..anil ~ w a n ] a r n : t a n t e a w a n t e naaya

t h i e f d

s e l f s

his dog-n

e na]

t h a t

manassilaayi

understood

The t h i e f

re alise d th at

s e l f s / h i s

dog

i s

a

fool.)

In both 49a and 49b ,

the

pronoun can have the

subject

of

the

matrix

as

i t s

antecedent, which

i s

what

and 8wa- are required to

do

in

these

cases.

The second property i l l u s t r a t e d y anaphora in Mal

ayalam

i s

that of long distance anaphora. Both

taan

and

swa-, as shown in 1.7., can cross any number of f i n i t e

clause

boundaries

to find an

antecedent

whether or

not

there are

intervening

subjects qualified to be

antecedents

themselves.

The third property,

demonstraten in 1.6. , i s

that

the

antecedents

of anaphors in

IVlc llayalam

are

required

to

be subjects. e

found tha t the notion subjec t can

not

be ident i f ied

wi

th unique configura t ional prcJperties


34/55

such as

of

S case

features such as the

nominative

or semantic roles such

as agenthood . This raises an in -

terest ing ques tion regarding

the

ident i f icat ion of

ante-

cedents

in Malayalam anaphora.

The fou rth p roperty found

in the

anaphor

~

is

tha t shows

long

distance anaphora only when is

NP contained

in

the subject . When NP contained

in

the

object swa mus t have i t s

immediate

subject

as

i t s ante-

cedent

i e

must

find i t s antecedent in i t s

minimal

c lause nuc leus as in

the

case of English

reflexives

and

reciprocals.


35/55

5

2.

ANAPHORA

IN KANNADA CHINESE YORUBA AND ICELANDIC

In

this

section, I

shal l

show

that the four proper

t ies

of

anaphora in Malayalam

are

not accidental language

specif ic

quirks,

but

are

found

to occur

again ana

again

in various other languages l ike Kannada, Chinese, Yoruba,

and

Icelandic.

Therefore, an

adequate

universal

theory

of anaphora must incorporate the r ight propert ies which

would

derive principles of

thi.s

kind in individual gra-

mmars

2 1

A h

K d

14

nap ora l.n anna

The

anaphor

laanu

in Kannada,

a

s i s t e r Dravidian

language, shows

very

much the same properties as the Mal-

ayalam

taan,

as

Shown by

50).

50

s s s ~ n u aanayennu k i l ~ i { i a l e g u s amma

se l f

elephant

pinched-that mother

maga ige

hee iQ.aleDQu]s

r a a ~ i cin:tisi{ialeUQu]

daughter told that

queen

thought that

aa hevgasu nanna

h n g ~ i y n n u

nambisi9a1u

that woman

my

wife

believe-caused

That woman convinced

my

wife

th t

the queen

thought

th t

the mother

told

the daughter that

se l f woman/*wife/queen/mother/*daughter) pin

ched the elephant . )

50 shows that ~ n u must have a subject as

i t s

ante-

14. I

am

grateful

to Sreevas

Mandalarn

for

the

data.


36/55

6

cedent, and tha t

allows

long

distance

anaphora, i . e .

can find i t s

antecedent

across f ini te

clause

boun-

daries. The domain of dis joint

reference in

Kannada, as

in

Malayalam, is

the

minimal

S,

NF

as

shown

51 .

51 a. moohan [awanu malagalengu] p r r l h i s i ~ n u

Mohan

he to

sleep

prayed

Mohan prayed

for

him

to

s leep.

b.

amma [raaniyige awalu k i ~ ~ l n g ~

mother

queen she to pinch

praarthisigalu

prayed

Mother

prayed

for

her to

pinch

the queen.

c.

aroma

[raal iYige ~

k i ~ a l e n g u praartjlisiq.alu

mother queen she to pinch

prayed

Mother

prayed for

her

to pinch the

queen.

The following

sentences demonstrate

more clear ly that

in

Kannada,

as

in

Malayalam,

i s

not

the case

tha t

anaphors

are

bound where pronouns are freeJ

52

moohan

Mohan

bugghimwanta engu

in te l l igen t

tha t

t i l i

dukondi danu

-

thought

Mohan

thought

tha t

was

inte l l igent .


37/55


38/55

8

Like Malayalam swa-, z i j i allow s long distance ana

phora

only

when NP-contained by the subj ect .

55 S a l l y

xiangxin

[ J ohn

gaosu

B i l l [Mary s h a s i l e zi

i]]

S a l l y

b e l i e v e d t h a t John t o l d

B i l l

t t

Mary

k il l e d s e l f Ma r y / * Bi l l / * J o h n / * Sa l l y ) . ) c f . 5 4 b )

a . ~ [

5 6)/Jo hn x ian gx in L Bill gaosu

Sam

z i j i

de

t a i t a i

believes told s lf s wife

s h a s i l e

JaCk]

]

k i l l e d

J o h n b e l i e v e s tJlat B i l l

t o l d Sam

t h a t s e l f s

B i l l s / * S a m s / J o h n s )

wife

k i l l e d

J a c k .)

b .

John xiangxin [ B i l l gaosu Sam

[JaCk

b e l i e v e s t o l d

s h a s i l e

z i j i

de

t a i t a i ] J

k i l l e d

s e l f

s

w ife

John believes t h a t B i l l t ol d Sam t h a t Jack

k i l l e d s e l f s

Jack s/ *B i l l s/ *S ar n s/ *John s)

wife.

2.3.

Anaphora i n Yoruba

6

Yoruba has

an

anaphor Qun and a pronoun which show

i n t e r e s t i n g

p r o p e l t i e s .

The

anaphor

Qun

must

take an

n t ~

cedent

in the

sentence, and

the antecedent,

l i k e

the ana-

16.

I

am

g r a t e f u l to

Douglas Pulleyblank fo r

c o l l e c t i n g

th e da ta f o r me .


39/55

39

phors in

Malayalam, Kannada, and

Chinese,

may be any

subject

higher

up in the sentences

cannot take an

object

antecedent.

The

pronoun

,

on

the other

hand, can

take any

antecedent

except

a subject , and as in the case

of pronouns in general, need not take an

antecedent.

57 a.

Tolu so fun

Segun

p oun sanra

told

tha-t

s e l f

i s f a t

Tolu

told

Segun that se l f Tolu/*Segun/*some

one

else i s fa t .

b.

Tolu

sofun

Segun

p

o

sanra

pron

Tolu

told S

egun

that

*Tolu/S egun/solneone

else i s fa t .

58 a . A r

. S -

pe Tolu

sofUn

egun pe oun

sanra

thought tha t

told tha

t s e l f

Ade thought

that

Tolu told Segun that se l f

Ade/Tolu/*Segun/*someone

else is fa t .

1

pe

o u

sofUn Segun

p

6 sanra

.

.

Ade thought

that

Tolu

told

Segun

that

*Ade/*Tolu/Segun/someone

else

is

fat .

The generalisation

that

underl ies

t he se sen tences is

obvious

I an

anaphor must

have

a subject

antecedent, and

a

pronoun

cannot

have

a

subject antecedent.


40/55

40

s

Subj . Obj . S

ubj

Gbj.

S

pronoun

Obj . S

ubj

Obj .

S

anaphor

s

Subj .

Discoursew.-..-o l

oun

The

behaviour of

/ follows from the

princi.ples

formulated

in

section 1.

In

order to account

for

the

behaviour

of

0,

shal].

se t

up

th e follow ing

princi .ple

for Yorubaa

60

Yoruba

Pronound cannot have subject antecedel lts.

2.4.

Anaphora

in

Icelandic ?

Pronouns in Icelandic exhibit dis jo in t reference as

shown by

sentences l ike (61).

61

Jon ha t hann

(John h ates him.)

In

addit ion to

th e g en eral dis jo in t refe re nc e. Ic e-

landic

pronouns

also show a subject

obviation similar

to

what

found in Yoruba.

The difference between the two

languages

i s tha t

in Yoruba,

obviation

applies

across

17.

The ent i re discussion of Icelandic i s based

on

the

examples provided by Thrainsson 1976). His solutions,

however,

are quite different from

mine.


41/55

tensed clauses,

while

in I celandic, t i s restr ic ted

within

t e r ~ e

clauses. Adopting Chomsky s ori gi na l in

s ight o f

the

tensed S

c o n d i t i o n ,

on e may form ulate the

obviation

p r i n c i p l e

in

I c e l a n d i c

a s follows

62

ce landic

Pronouns cannot have s ubject antecedents

in the minimal tensed clauses

t h a t

con-

ta in thenl

Examples t h a t i l lus t ra te

62 )

a re given

belowl

63)a.

Jon

syndi H a r a l d i

f o t

hann

showed

c l c t h e s fo r him

John

showed

clothes

fo r

him

* Jo h n /Har o ld ) .

b . Jon

r e t t i

H a r a l d i

hans

fo t

handed

h is

c l o t h e s

John

handed Ha rold

h is

* J o h n s /H a r o l d s )

c lo thes

c .

Jon

t e l u r

H ara ld h af a

r ~

hann

e l i e v ~

to

have

shaved

him

John

believes Harold to have shaved

him

*John/*Harold).)

d

ron t e lu r ~ l i a r a l d u r

h a f i

r ~ hann

believes t h a t

has

shaved

him

John b e l i e v e s

tha t Harold

has shaved him

J o h n / ~ H a r o l d .

In

63a) and 6Jb), th e pronoun can have the

o b j e c t


42/55

42

Harold

as

i t s

antecedent,

but not

the

subject John. In

(6Jc), both

John

and Harold are subjects in the minimal

tensed

clause

that

contains hann,

and therefore, nei ther

of

them can

funct ion

as

t s

anteceden t.

In 6Jd , ev en

though John

i s

a subject, does not l i e within

the

mini-

mal

te ns ed c laus e tha t contains hann, and hence

62)

does

not apply to

i t

Compare now the obviation p rin cip le s o f Yoruba

and

Icelandic.

60

Yoruba

Pronouns

cannot have

subject

antecedents.

62 Icelandic

Pronouns

cannot have

subject antecedents

in

the minimal tensed c lause tha t contains

them.

can

collapse

the

two

principles

follows.

64)

Pronouns

cannot

have

subject

antecedents ( in

the

minimal

tensed clause

tha t

contai.ns them.)

At

this point , I would l ike

to

draw the reader s

a t tent ion

to an

in teres t ing

fact of obviation in English

pronouns. t

appears to

be the case tha t English pronoun3,

when

they

occur

as

subjects of in f in i t iva l clauses, cannot

have

matr: x st lbjects

as

antecedents , though

thl.Y

can

have


43/55

4;

matrix

objects as antecedent s.

18

The contras t i s i l lu s -

t rated

by

the

following pai r of sentences.

(65)a.

*

John

was

surprised

for

him

to

be

l e f t

out.

b. I t surprised John for him

to

be l e f t out.

The

principle

responsible for this

contras t

may be

stated as follows.

66 English

In f in i t iva l

subject pronouns

cannot have

matrix

subjects

as ant eceden ts .

I t i s tempting to collapse the obviation

principles

of Yoruba, I celandic, and English in to something l k ~ I

(Subject)

pronouns cannot

have

subject antecedents in

trle

minimal

t ~ n s clause

that

contains them) t Since,

however,

is only the immediate matrix subject ,

and

not

the s ub je cts h igher

up

tt.a t the in.fini

t i

val subj

~ t

shows obviation with,

this

may

not

be the

r ight

move to

m ke

18. This

fact

was

pointed

out

to me by Joan Bresnan.

Even

though mos t speakers re ject (65a), I have also come

across

some who do not. Even in t ~ s cases, however,

the contras t

between the two sentences

is quite

clear:

65b is perfect ly grammatical, While

coreference

is

possible

in

65a

only with

some

ef fo r t .

19.

01serve

t ha t

(66), Which

seeIa.J to be independently

necessary, would also

account

for the fo llowing contrast:

( i) * John believes him to be a fool .

i i John

believes that he is a fool.


44/55

67

Mary was annoyed to find John to

be surprised

for her

to

be l e f t

out.

i s

c lear th at not a l l languages

choose

to

include

the

obviation

principle in

their

grammars.

As

demonstrated

i n 1.4., Malayalam

grammar

does not con ta in the

principle.

Since

a

number of

geneticall .y unrelated

languages

show

Bome version or the other of (66), however, I

s h a l l

assume

t h a t

i t i s a pa t

o f UG

20

We

ahal l now

turn to the

behaviour

of anaphors

in

Icelandic.

seems

+0

be the case

t h a t

anaphors

Ice-

landic can have

both

subjects and objects

as

an

antecedent,

and they can find t h e i r anteceden ts ou ts ide the domain

of

both

d i s j o i n t

reference and

subject

obviation,

as

shown

21

by the following

examples.

20. Finnish

appears

to

be

another

language

in which ana-

phors

must have

subject

antecedents, and pronouns

canno t. Consider

the fo llowing data (provided by

ur i Cal lson)

i ) a . Juha tappoi Villen puutarhassan

John k i l l e d B i l l

in

s e l f s garden

(John k i l l e d B i l l in John s/*Bill s/*someone

e l s e s garden.)

b.

Juha

tappoi Villen n ~ puutarhassan

i n pron s garden

(John

k i l l e d B i l l i n

*John s/Bil l s/someone

e l s e s

garden.)

21.

Thrainsson s examples do

not

clear ly

demonstrate

tha t

the

antecedent of sig must f-command i t , but then he

does not give

any

examples to the

contrary

e i t h e r .

Therefore, I s h a l l

assume that the antecedent of s ig

should

f-command

i t .


45/55

68)a.

Jon

sYndi Haraldi fOt a s ig

John showed

Harold

clothes fo r him

John

showed

Harold clothes fo r him John/Harold)

b.

Jon

r e t t i Haraldi

s i n

f o t

handed s e l f s clothes

John handed Harold s e l f s J o h n s /H a r o l d s )

c l o t h e s.

c.

on

t e l u r

Haral d

h afa rakaJ s i g

be11eves

to have

shaved s e l f

John believes Harold to have shaved

s e l f

John/Harold).

d . Jon

t e l u r

Haraldur

h a f i

r ~ s i g

believes

t t

has

shaved s e l f

John believes

t t

arold

has

shaved

s e l f

John/Harold)

e . Jon s e g i r

ai

Haraldur t e l j i ~

i l l i

says That b e l i e v e s

t h a t

v i l j i Maria r a k i s i g

wants t h a t Mary

shaves

s e l f

John

says

t h a t Harold believes

t h a t

i l l

~ ~ n t s

t h a t

Mary

shave s e l f Mary/Bill/Harold/John).)


46/55

46

3 .

N PHOR N

In t h i s s ection, I s h a l l r a i s e

orn

o f the

problems

posed

fo r

the binding

p r i n c i p l e s

in

by

the

f acts

o f

anaphora

in languages l i k e Malayalam, Kannada, Chinese,

Yoruba, and I c e l a n d i c . My concern

here i s

mainly with

the

p re se nt ati on o f

the

problems t h m s l v ~ fo r

fellow

researchers in the f i e l d , n o t the

construction

o f an

a l-

ter native

theory

o f anaphora.

The

most

serious

problem

t h a t

the

c u rre n t

formulation

o f the

binding

conditions fa ce s, as fa r as I can

see,

is

r e l a t e d

to the

i d e n t i f i c a t i o n o f the domains

o f

d i s j o i n t

reference and

anaphor binding.

The relevant condi tions

are s t a t e d as

follows

69)a. Anaphors a re bound in t h e i r

governing

category.

b.

Pronominals a re free

i n

t h e i r governing

category.

Whatever be the de fi ni t i on o f

government

and

governing

categor y ,

follows

from th e c on ju nc t

o f

69a

and

69b

t h a t

governed anaphors must

find

t h e i r

antecedents in

the

domain i n

Which

governed

pronominals

e x h i b i t

d i s j o i n t

reference.

That

i s

anaphors

a re

bound

where

pronominals

are free.

s a universal

p rin ci p le , th is

i a i n c o n s i s t e n t

with

the

f acts

o f a na pho ra and d i s j o i n t

reference in Malayalarn, Kannada, Chinese,

Yoruba,

and

I c e l a n d i c . n

a l l

-these

l an gu ag e s, a na ph or s ca n

find


47/55

47

thei r antecedents outside tIle etomain of dis jo in t refere

nce, thereby allow ing

both

pronouns

and anaphors to have

the same antecedent

in the

same s t ruc tura l posi t ion.

What are

the

moves tha t can be made such tha t

these

languages do not consti tute a counterexample

to

(69)7 One

may, for

example, think

of saying

tha t what

I have called

anaphors

in these

languages are

not

in

fact

anaphors, and

tha t therefore. condition (69a)

i s

not

applicable to

them.

22

This proposal has

the

effec t of

making

a dist inct ion

between

those reflexives

which are

anaphors and those

which are not,

thus

rais ing the

follOWing problema

f i r s t

t

forces

us to

t rea t the binding

p rope rtie s o f r ef le xiv es

in

English-type l anguages and non-English-type languages

in unrelated

ways,

which clearly must

be

avoided

possi-

ble . Second, one i s forced

to the

di f f icu l . t t k of defi-

ning anaphor in such a way that t

would

include ref le

xives in English, b ut

would exclude reflexives

in Malaya

lam,

Kannada, Chinese. Yoruba, and Icelandic. Chomsky 1979

Characterises anaphors

as

elements

tha t

are lexical ly spe

cif ied as needing an an-tecedent (p.16) , and (irfp ress) as

NP s that have no capacity fo r in here nt reference (Ch I I I

p.42). Under

e i ther

of these character isa t ions, the

re

f lexives in

the

languages tha t we looked a t in Sections

1 and 2 qualify to

be anaphors.

22. This

solution

was suggested to me by

o m

Chomsky.


48/55

48

For the binding conditions to be meaningful,

the theory should

offer a

universal character isa t ion

of

the class of reflexives to which the principles wil l

apply. t should

a t

l eas t

ident i fy

the properties which

would

make

~ n g u g s

proper candidates

for the binding

condit ions. s

far as

I know, no such proposal exis ts

Even in languages

l ike English,

the prediction

that

anaphors a re bound where pronominals are free i s

not

without

problems. Generally

recognised problem cases

in the l i te ra ture

are sentences

l ike the fo llowing:

(70)a.

They

admire

the ir children.

b.

They

admire

each

other s children.

f i s

false

that anaphors

are

bound

Where

pro

nouns are free, then 69a

must

be revised. Perhaps

a

possible

way

of

approaching

this

task

would be

to

make

th e fo llowing parametric option

available

71 Anaphors

are

bound (in

the i r

governing category).

Languages l ike Malayalam, Chinese, Yoruba, and Ice

landic leave out

the

more

res t r ic t ive condition

in

the

brackets, thereby choosing the

more general

condition

anaphors

are

bound , which

i s

in

fact

par t of

the

very

defini t ion of

anaphors.

Languages

l ike English,

on

the

other

hand, choose the fu l le r I ersion of the condi t ion.

A

secontl

problem tha t faces the

theory i s

the

speci-


49/55

f ic a t io n o f what c o n s t i t u t e s a l e gi t i m a t e a n t e c e d e n t in

languages l ike Malayalam. P r i n c i p l e 71) allows any

c-commanding NP to

be

th e

a n t e c e d e n t

o f

an

anaphor,

b u t

in

the Malayalam

type

la ng uag es , o nly s u bje c ts a r e p os s i

b le antecedents

o f anaphors.

L et us

say

t h a t this pheno

menon can

be derived

by p aram etricisin g

the

co n d itio n

on e

s te p fu rth er.

72

Anaphors

must

be bound to

a

su b ject) in the i r

governing

c a t e g o r y ) .

Malay alam, Kan nada, Chinese, and Yoruba

choose

the

more

res t r ic t ive

condition

a bo u t s ub je ct ho od o f antecedents,

languages

l i k e English

and

Icelan d ic

do not .

I f

the pa r a m e t r i c i s a t i on o f the binding conditi.on

as

in

72

is

necessary, ~ q u e s t i o n

t h a t ar i ses

imme-

d i a t e l y

is .

what

c o n s t i t u t e s a

su b ject?

Recent

work

on

nonconfigurational

languages has

made

amply

obvious

tha t the configura

t i o n a l

de f i ni

t io n

o f subj e c t as

NP o f

u and

o f obj e c t as NP

of

VP i s

n ot u n i v e r s a l l y

a p p l i

cab le.

3

I n

o r d e r

to c h a ra c te rise th e

n otio n S Ubje ct

in nonconfigurational

languages,

Chomsky i n

p ress) pro

p05 JS the notion o f lex ica l VP ,

Which

may be s a i d to be

2 ]. c f . Hale

1980),

Nash

1980), Farmer 1980), Simpson

1980), and

Mohanan

in p r e s s ) .


50/55

consist ing of the verb and the

arguments

th t

Bub-

categorises for, on the assumption

tha t

verbs

do

not

subcategorise for subjects. Thus, in

John

gave

Mary

a book ,

the

lexica l

VP consis ts of the unordered

se t

book, give, and

Mary.

Languages may

dif fer

with respect

to having or not

having a

syntact ic

VP

but

l l

langua

ges

on

th is

assumption

have

lexi l

V ~ Even though

Malayalam does

not

have

a

syntact ic

VP

has

a

l ex i -

cal VP

in

the sense outl ined above.

homsky suggests tha t d- and

a-structure

repre-

sentat ions in nonconfigurational languages

may

be looked

upon as pairs of configurat ional and

lexi l

represent-

at ions. Given

th at le xic al

VPls

are

paired with VP-less

configurations

in

Malayalam, the a-structure

of

7Ja)

may be thought

of

as

7Jb).

73)a.

b.

k u ~ ~ i aanaye n u ~ i

child-n

elephant-a pinched

s

P

k U ~ ; i

nom

aanaye

ace

configurational

st ru ture

lexica l

structure


51/55

5

Instead of saying

that

is

only

nonconfigurational

languages that have pair ed a -s tructur e r ep re sent at ions ,

would be

bet ter to

generalise to configurational

languages

as

well,

and

say

that

a-structure

is

univer

sal ly

a

pair of configurational and

l ex i ca l s tructures .

In configura t iona larlguages, the configurational s t ruc-

tures

happen

to , uut

need

not, re f lec t the lexical

struc-

turel

configura t iona l

structure

74

The boy

s

VP

I

pinched the e lephan t

lexical

structure

V

S

rom these assumptions,

follows

that the

universal

def ini

t ion

of objec t

i s

NP of

l ex ica l

VP ,

and

t t t

of

subjec t

NP

of l ex ica l S . f one

accepts

these assump-

t iona,

72)

may be

restated as followsJ


52/55

52

75 Anaphors must be bound

to

the NP of l exica l S)

in

their

governing category).

24

24. There are several residual problems. I f

PRO i s an

anaphor,

75

would p re dic t th at

in those languages

that

leave

out the condition

in thei r governing

category , PRO must

have

an antecedent

in

the sen-

tence. This prediction, as

far

as I know, i s in-

correct . Therefore, would be

necessary

to say

that

PRO

i s not an anaphor. see

also

footnote

7

Another problem would

be the

treatment

of cau-

sat ivas

in Malayalam. Recall

that

under causat ivisa-

t ion,

the

intransit ive

subject

becomes

the

t ransi -

t ive object ,

and

i s

no longer

an

e l ig ib le antecedent

of anaphors. c f. 40 a, b). An ident ical s i tua t ion

i s found with respect to par t ic ip ia l adjunc t c lauses

which are obligatori l ly

controlled

by

matrix

subjects .

Given the formulation in 75), we are forced to say

that the effec t of causa ti vi sa ti on

i s

to

convert

an

NP

of

S

into an of VP a t the lexical J.evel. Such

a move, however, co rresponds to a ru le of

move

tha t

moves

an

NP into a

VP in

configurat ional lan-

guages, and

would

presumably

be ruled out by the

projection principle

c f .

Chomsky

in

press

One may go on to aska

how i s the behaviou r of

anaphors l ike

the

Malayalam wa

and

Chinese

zi j i

which

exhibi t long

distance

anaphora

only when con-

tained

by the

subject ,

taken care

of? In what

precise

terms i s the obviation

in

Yoruba, Icelandic, and

English stated? Straightforward

answero

to

these

questions do not

appear to be

a

t r i v i a l matter .

I

leave these

kno tty quest ions to future

research.


53/55

I f this i s

the

r ight

way of looking

a t anaphora,

what

i t

implies i s tha t binding conditions

apply to

lexical

s t ructure ,

not

configurational

structure. he

notion

bound , which means ltc commanded an

antece

dent , must therefore be

redefined

as

I-commanded

an

antecedent ,

where

I-command a t

the

level

of lex

i ca l structure corresponds

to

c command a t the level

of

configurational s t ructure . This

revision is

perfect

ly consistent with

the analysis

of

examples l ike

(Jla)

and J ib , which independently

suggests tha t the re le -

vant notio n

of

command for

n p o r ~ binding i s not to

be found a t the configurational

structure.

Thus,

we

are lead to conclude

that the principles rove rn ing the

relat ion

between

anaphors

and the i r antecedents are

s ta ted,

not a t the level of

configurational structure,

but

a t

a

level

tha t

represents

grammatical re la t ions .


54

Grammatical relations in Malaya

Documents

Transcript of Grammatical relations in Malaya