Null Subjects a Reanalysis of the Data

30
Null subjects: a reanalysis of the data MELVYN DOUGLAS COLE Abstract In the face of problems with previous syntactic theories, this article 1 estab- lishes a new concept relating to the incidence of null thematic pronouns, focusing on null thematic subjects. The data indicate that, in rich agreement languages with null subjects, prima facie, recovery of such subjects is achieved first by reference to discrete agreement and, if this fails, by refer- ence to an antecedent in context. Subsequently, preferred interpretations are resorted to and then overt pronouns. The argument is developed to show that every language has its point of morphological maximality. This is the maximum point up to which identifying subject verb agreement can occur in a language and this level of verbal agreement must be present for thematic subjects to be null. This point is scalar, being represented by morphology for gender, person and number in Tarifit, for person and number in Spanish, for person in Bengal and null morphology in Chinese. A further necessary feature for thematic null subjects to occur is that an an- tecedent in context completes their recovery. Whether this occurs or not is put down to the contextual weakness and strength of languages. This theory has the consequence that the licensing of thematic subjects is redundant and that null expletive subjects need to be treated as a separate empty category. 1. Introduction The incidence of covert pronouns has been one of the more notable areas for generative grammar research over recent decades. The pre-eminent early contribution, that of Rizzi (1986), saw covert pronouns as licensed by a governing head and recovered by the features of that head (e¤ec- tively AGR S in the case of null subjects). This approach faced problems as Chinese and Japanese, inter alia, have null subjects without verbal agreement for person and number. These initially appeared to be resolved Linguistics 47–3 (2009), 559–587 DOI 10.1515/LING.2009.019 0024–3949/09/0047–0559 6 Walter de Gruyter

Transcript of Null Subjects a Reanalysis of the Data

Page 1: Null Subjects a Reanalysis of the Data

Null subjects: a reanalysis of the data

MELVYN DOUGLAS COLE

Abstract

In the face of problems with previous syntactic theories, this article1 estab-

lishes a new concept relating to the incidence of null thematic pronouns,

focusing on null thematic subjects. The data indicate that, in rich agreement

languages with null subjects, prima facie, recovery of such subjects is

achieved first by reference to discrete agreement and, if this fails, by refer-

ence to an antecedent in context. Subsequently, preferred interpretations

are resorted to and then overt pronouns. The argument is developed to

show that every language has its point of morphological maximality. This

is the maximum point up to which identifying subject verb agreement can

occur in a language and this level of verbal agreement must be present for

thematic subjects to be null. This point is scalar, being represented by

morphology for gender, person and number in Tarifit, for person and

number in Spanish, for person in Bengal and null morphology in Chinese.

A further necessary feature for thematic null subjects to occur is that an an-

tecedent in context completes their recovery. Whether this occurs or not is

put down to the contextual weakness and strength of languages. This theory

has the consequence that the licensing of thematic subjects is redundant and

that null expletive subjects need to be treated as a separate empty category.

1. Introduction

The incidence of covert pronouns has been one of the more notable areas

for generative grammar research over recent decades. The pre-eminent

early contribution, that of Rizzi (1986), saw covert pronouns as licensed

by a governing head and recovered by the features of that head (e¤ec-tively AGRS in the case of null subjects). This approach faced problems

as Chinese and Japanese, inter alia, have null subjects without verbal

agreement for person and number. These initially appeared to be resolved

Linguistics 47–3 (2009), 559–587

DOI 10.1515/LING.2009.019

0024–3949/09/0047–0559

6 Walter de Gruyter

Page 2: Null Subjects a Reanalysis of the Data

by C-T. J. Huang’s (1984) operator variable analysis, according to which

object and subject empty categories in those languages are variables

bound by a preceding overt or covert topic. However, Xu (1986) and Xu

and Langedoen (1985) observed that this analysis falls foul of the require-

ments of subjacency, the bijection principle and the strong crossover con-

dition in certain environments in Chinese2. Also, Y. Huang (2000) points

out that the main idea upon which it is based, namely that null objects ofcomplement clauses in languages lacking verb object agreement cannot

have a matrix clause subject as an antecedent, is not always the case.

C-T. J. Huang (1989) also produced his generalized control rule, which

required pro/PRO to be controlled in its control domain if it has one.

The control domain is closely associated with the idea of an accessible

SUBJECT and means that, in Chinese, pro/PRO must be identified by a

nominal in a higher domain. Y. Huang (2000) observes that the rule does

not work universally in Chinese, citing, inter alia, the following example:

(1) Xiaohong de meimei shuo Ø xihuan tan gangpin.

Xiaohong GEN younger-sister say like play piano

‘Xiaohong1’s younger sister2 says that (I/you/he/she1/2/3/we/they)like(s) to play the piano.’

Finally, Chinese has an as yet unexplained expletive null subject, as in

Example (2):3

(2) Haoxiang ta mei jin guo cheng.

Seems he not go-to PAST town

‘It seems that he did not go to town.’

In the face of this problem, Jaeggli and Safir (1989) argued that null sub-

jects are permitted in morphologically uniform languages, namely those

with either only derived or only underived forms in their verb paradigms

(like Chinese). This was found to be unsatisfactory in the face of evidence

that, inter alia, morphologically uniform Norwegian, Swedish and Dan-ish lack null subjects.

Speas (1994) proposed that the expression or otherwise of null subjects

is determined by whether the j features in IO have some specification. If

they lack such specification, they must be given value through the spec/

head relationship. Languages with poor agreement do not allow null

subjects, since null subjects do not provide value to such agreement. In

languages lacking subject verb agreement (like Chinese), null subjects

may occur because there is no agreement to give value to. Speas accountsfor the lack of null subjects in Swedish by observing that, while it has no

verb subject agreement, it does possess some adjectival agreement. How-

ever, the logical connection between adjectival agreement and the expres-

560 M. D. Cole

Page 3: Null Subjects a Reanalysis of the Data

sion or otherwise of subject or object pronouns is unclear. Furthermore,

as Neeleman and Szendroi (2007) observe, Afrikaans lacks agreement

altogether and also null subjects, while Speas herself states that Papia-

mentu, Duka and Guaymı, which prima facie have no subject verb agree-

ment and no null subjects, require further research before the theory can

be properly substantiated in relation to them. To these one can add Song-

hay, as shown by Kameyama (1985), and Fon and Vata, as reported byLaw (1993), which also lack subject verb agreement and null subjects.

Tomioka (2003) proposes that null subjects occur in languages that

allow (robust) bare NP arguments. In his view, languages can delete NPs

providing this does not strand determiners. However, Neeleman and

Szendroi (2007) rightly question why it should be only the stranding of

determiners by NP deletion that blocks null subjects, and why such null

subjects should not also be blocked by the fact that NP deletion would

also strand functional heads like sentence function particles wa, ga and o

in Japanese. Also, by the same logic, why should not NP deletion and by

extension null subjects also be wrongly ruled out by the stranding of the

classifiers with which the count nouns in Chinese are obligatorily merged.

Facing these di‰culties, Neeleman and Szendroi (2007), not wishing to

ignore the language internal evidence that agreement must play a part in

the incidence of null subjects, take the view that languages fall into two

categories. These are those where agreement basically determines whether

null subjects occur and radical pro drop languages, where the incidenceof null pronouns is orthogonal to agreement. Basically Neeleman and

Szendroi propose that radical pro drop can occur in a language where

pronouns are marked agglutinatively for case or some other morphologi-

cal element. This proposal is shown to produce correct solutions in 20

languages, but it faces problems.

Firstly, it is not clear to what extent quantitative considerations are

involved in whether or not a language is a rampant pro drop one. If, as

may be assumed, they are involved, the absence of verb subject agreementin non-rampant pro drop Italian leaves no justification for the fact that,

in the following example in that language, the null object can be referen-

tial as well as generic, depending on the context.4

(3) L’atteggiamento di Giolitti in quell’ occasione lascio Ø

the attitude of Giolitti on that occasion left

daverro perplessi.

really perplexed

‘The attitude of Giolitti on that occasion left Ø really perplexed.’

Next, Finnish has agglutinative case endings on pronouns and lacks so-

called radical pro drop, thus plainly running counter to the theory of

Null subjects 561

Page 4: Null Subjects a Reanalysis of the Data

Neeleman and Szendroi. Lao is also a counterexample. Enfield (2007)

lists its pronouns as follows:5

1st person 2nd person 3rd person

Singular Bare kuu3 mung2 man2Familiar haw2 too3 laaw2

Polite khooj5 caw4 phen1

Formal khaa5-phacaw4 thaan1 thaan1

Plural Bare phuak4-kuu3 suu3 khaw3, man2

Familiar haw2Polite cu-haw 2 (inc) cu-caw4 khacaw4, phen1

cu-khooj (exc) phuak4-caw4

I am informed by Enfield (personal communication) that the two ele-

ments phuak and cu (see above) that precede some singular pronouns to

make plurals are independent words meaning group although cu has

undergone slight phonological modification (loss of a final glottal stop).

Even if one sees the element cu as a separate morpheme in an overall pro-noun, say cu-haw2 or cu-caw4, rather than a separate word, it does not

mark plural throughout all persons. So Lao pronouns are not generally

agglutinatively marked for plural. Neither are they marked at all for

case. However, despite the lack of agglutinative marking on Lao pro-

nouns, they can generally remain unexpressed in any position, so the

language is radical pro drop and consequently a counterexample to

Neeleman and Szendroi’s theory.

Aiton, a Tai language spoken in Assam, allows subject and objectpronouns to be dropped without agreement6 and appears to be a radical

pro drop language in Neeleman and Szendroi’s terms. Its pronouns are as

follows:

1st sing 2nd sing 3rd sing 1st plural 2nd plural 3rd plural

kau2 ma m2 man2 hau2 suu1 khau1

Since Aiton’s pronouns are not marked for case and are not agglutina-tively marked for plural, it is another counterexample to Neeleman and

Szendroi’s proposal. Vietnamese also does not appear to conform to

Neeleman and Szendroi’s ideas either. Cooke (1968) lists its pronouns as

follows:

may/may/ bay mınh no toi ho˙

to

ta tao toa moa ng§ƒi chung

He observes that there is no marking for case. Plural is frequently not

overtly expressed, but where it is, there are various ways of doing so.

With some pronouns, it can be done by preposing chung or cac, meaning

562 M. D. Cole

Page 5: Null Subjects a Reanalysis of the Data

group. Thompson (1965) refers to chung as a restrictive complement in

phrases with pronouns as head. It can also appear independently as a

third person plural pronoun. Some pronouns such as minh and ta can be

either singular or plural. The only specifically plural pronoun other than

chung, namely ho˙

, has no formal relationship with any singular form.

Again, there does not seem to be morphological marking for case or any

other characteristic that justifies the presence of radical pro drop in thesense of Neeleman and Szendroi.

Finally, this review turns to Holmberg (2005), who addresses some dif-

ficulties caused to the syntactic theory on null pronouns by the fact that,

under the version of minimalism outlined in Chomsky (2001), the F fea-

tures in I are uninterpretable and enter the derivation unspecified. This

means that an unspecified pronoun cannot be specified by these features.

Using evidence from Finnish, Holmberg rejects the view of Manzini and

Roussou (1999), Alexiadou and Anagnostopoulou (1998) and Manziniand Savoia (2002) that verb subject agreement acts as a pronoun in what

have traditionally been viewed as null subject constructions. He concludes

on the basis of Finnish that covert subject pronouns appear in Spec IP

and that narrow syntax is oblivious to whether they are expressed. This

conclusion seems not totally at odds with evidence in Cole (2000) that

syntax is not the sole determinant of the incidence of null subjects. First,

Example (4) in Spanish, where the verb tenıa is ambiguous between 1st

and 3rd person singular, is satisfactory even without an overt subjectpronoun in the 2nd sentence, because the null subject is recovered by ref-

erence to Juan, its antecedent in the initial sentence:

(4) Juan llegaba. Ø Tenıa las llaves.

Juan arrive-1/3S-IMP have-1/3S-IMP the keys

‘Juan was arriving. He had the keys.’

This recovery of null subject pronouns in Spanish by reference to an ante-cedent in context is what Nakamura (1987) and Yan Huang (2000) re-

spectively observe to occur in Japanese and Chinese, which lack person

and number agreement on the verb.

Further evidence that syntax is not the sole factor involved in null pro-

nominal subjects comes from the observation of Samek-Lodovici (1996)

that null thematic subject pronouns are only possible in Italian with a

topic antecedent and thus that a null subject in the second sentence of

Example (5) is unacceptable because it lacks such a topic antecedent.

(5) Ogni mattine, la mostra e visitata da Gianni.

Every morning, the exhibition is visited by Gianni.

Null subjects 563

Page 6: Null Subjects a Reanalysis of the Data

Piu tarde lui/egli/*Ø visita la universita.

More late he visit-3S.PRES the university

‘Every morning, the exhibition is visited by Gianni. Later, he visitsthe university.’

The idea that null subjects occur in individual circumstances in languages

is further developed in this article. Indeed, here, a system will be outlined

of the recovery of thematic null subjects by reference to contextual ante-

cedents, supported by morphology in certain languages.

Since previous syntactic theories are inadequate, this article re-

examines the incidence of thematic null subjects on the basis of new datafrom a variety of languages. Section 2 outlines the strands of the problem

and an initial approach to investigating it. It examines the incidence of

overt and covert pronominal subjects in languages where null subjects

occur widely, including those with significant verb subject agreement and

those completely lacking such agreement (other than for honorifics). It

then illustrates the absence of thematic null subjects in languages with no

agreement and in certain languages with rich agreement. The first conclu-

sion is that thematic null subjects do not occur in any language regardlessof circumstances. There is a step-by-step process in the recovery of the-

matic null subjects in languages with rich agreement that proceeds first

by reference to discrete morphology, then if the morphology is not dis-

crete, by reference to an available antecedent in context and, thirdly, to

recovery by virtue of a preferred interpretation. Finally, if none of these

methods adequately recovers thematic null subjects, an overt pronoun

must be resorted to. Section 3 examines the general principles that any

new approach to thematic null subjects must embody and shows thatany language has a point of morphological maximality up to which

agreement recovers null thematic subject pronouns. The degree of mor-

phological maximality is scalar, varying from Chinese, where it is null,

to Tarifit, where it is for person, number and gender. In Section 4, this

concept is related to the previously established procedure for the recovery

of thematic null subjects to show that, while their recovery by agreement

up to the point of morphological maximality is necessary, it is ultimately

recovery by reference to an antecedent in context that determines whetherthey occur. Section 5 concludes that a significant di¤erence between lan-

guages with thematic null subjects and languages lacking them is that

the former are contextually stronger than the latter. Section 6 addresses

some consequences of this theory, including the facts that the licensing

of thematic null subjects is redundant and that expletive null sub-

jects need to be dealt with as a separate issue. Section 7 concludes. The

article does not address null thematic pronouns in other than subject

564 M. D. Cole

Page 7: Null Subjects a Reanalysis of the Data

position, or expletive null subjects, which are regarded as a separate area

of research.

2. Basic problem and overall approach and language survey

2.1. Basic problem

The following four axes in relation to thematic null subjects require cor-

relation:

(a) Italian and Spanish have thematic null subjects and rich identify-

ing morphology;

(b) Swedish lacks thematic null subjects and identifying morphology;

(c) Japanese has thematic null subjects, but no identifying morphol-

ogy;(d) Icelandic lacks thematic null subjects, but has rich verb morphol-

ogy.

The argument will set out from the position, observed by Gilligan (1987),that thematic null subjects are the norm in most languages. A good

approach to determining why they occur accordingly seems to be to ob-

serve the conditions under which overt pronouns are necessary in so-

called null subject languages and to compare this with reasons why they

are necessary in non-null subject languages.

2.2. Pragmatic use of overt pronouns in so-called null subject languages

Overt subject pronouns occur crosslinguistically for the purpose of focus

in languages where null subjects occur. Example (6) in Turkish and Ex-

ample (7) in Japanese illustrate.7

(6) Ben Cuma gunu Istanbul-a git-ti-m

I Friday day Istanbul-DAT go-PAST-1S

‘I went to Istanbul on Friday.’

(7) Kare-wa dokusho-ga sukida.he-TOP reading-NOM like-PRES

‘He likes reading.’

These examples only incorporate overt pronouns ben and kare for the

purpose of focus and would be quite adequate otherwise with covert pro-nouns.

A further general use of overt pronouns in languages in which null sub-

jects occur is for change of topic. The following examples illustrate:

Null subjects 565

Page 8: Null Subjects a Reanalysis of the Data

Japanese:

(8) John-wa mat-tei-ta. Mary-ga ie-ni

John-TOP wait-PROG-PAST Mary NOM house-LOC

tsui-ta. Kare-ga doa-o ake-ta.

arrive-PAST he-NOM door-ACC open-PAST

‘John was waiting. Mary arrived at the house. He opened the door.’

(9) John-wa mat-tei-ta. Mary-ga ie-ni

John-TOP wait-PROG-PAST. Mary-NOM house-LOC

tsui-ta. Ø doa-o ake-ta.

arrive-PAST door-ACC open-PAST

‘John was waiting. Mary arrived at the house. She opened the

door.’

Spanish:

(10) Juan esperaba. Juanita llego a casa. El

Juan wait-3S.IMP Juanita arrive-3S.PAST to home. he

abrio la puerta.

open-3S.PAST the door

‘John was waiting. Juanita arrived home. He opened the door.’(11) Juan esperaba. Juanita llego a casa.

Juan wait-3S.IMP Juanita arrive-3S.PAST to home.

Ø Abrio la puerta.

(she) opened the door

Examples (9) and (11) are both satisfactory. In the first, the null subject is

recovered by reference to previous context in the form of Mary and, in

the second, it is recovered by reference to verb agreement as third person

singular. Substitution of kare and el respectively for the null subject in

otherwise identical Examples (8) and (10) provides a change of topic

from the one context would have implied had there been a null subject

as in Examples (9) and (11).

However, while focus and change of topic are important purposes forusing overt pronouns in languages where null pronouns occur widely,

such use of overt pronouns is only for purely pragmatic purposes. In

fact, overt thematic subject pronouns must occur in certain circumstances

in many so-called null subject languages, an issue which will form the

basis for further discussion.

2.3. Obligatory incidence of overt thematic subjects in null subject

languages

Before proceeding to the main discussion, a number of examples repre-

senting axes (A) to (D) set out in Section 2.1 will be analyzed in order

566 M. D. Cole

Page 9: Null Subjects a Reanalysis of the Data

to illustrate some crosslinguistic patterns of the required use or otherwise

of overt subjects. Languages under axis (A) will be discussed at greater

length than the other types so as to establish a crosslinguistic generaliza-

tion that will be an essential factor in subsequent discussion.

2.3.1. Languages with null thematic subjects and rich agreement. Swa-

hili has null thematic subjects and discrete subject agreement on all verbs,

leaving no reason for overt subject pronouns to be required in this lan-guage. Example (12) illustrates.8

(12) Ø A-li-m-busu Halima.

3S.SUBJ-PAST-3S.OBJ-kiss Halima

‘He kissed Halima.’

Spanish presents a less simple picture, as the following examples show:

(13) Ø Llegue a casa.

Arrive-1S.PAST to home

‘I arrived home.’

(14) Juan y yo llegabamos a casa. *Ø/El

John and I arrive-1PL.IMP to home. ?/he

tenıa las llaves.

have-1/3S.IMP the keys.

‘John and I were arriving home. I/he had the keys.’

(15) Juan llegaba a casa. Ø Tenıa las

John arrive-3S.IMP to home. (he) have-1/3S.IMP the

llaves.

keys.

‘John was arriving home. He had the keys.’

A large majority of its verb forms have unique subject verb agreement,

but the 1st and 3rd persons singular of the imperfect indicative tenseamong others are syncretic. In Example (13), the null subject is recovered

by unique 1st person singular agreement. In Example (14), a null subject

is not acceptable since tenıa is ambiguous between 1st and 3rd person sin-

gular, so an overt pronoun (El ) is needed. Example (15) is, however, ac-

ceptable, even though tenıa is again ambiguous, because the subject of

tenıa is taken to be Juan, the subject of the verb in the 1st sentence in

that example. Thus the null thematic pronoun is recovered by reference

to an antecedent in context. This is impossible in Example (14) becauseJuan y yo, the subject of llegabamos in the initial sentence, being 1st per-

son plural, is not an antecedent from which the identity of the subject of

the subsequent tenıa can be established. European Portuguese, with a

Null subjects 567

Page 10: Null Subjects a Reanalysis of the Data

similar pattern of syncretic forms to Spanish, exhibits a similar pattern of

the incidence of overt and null thematic subjects.

Serbian has syncretic forms in the 2nd and 3rd person singular of the

imperfect and aorist tenses. The following examples in the imperfect illus-

trate the position:9

(16) Ø beste u prirodi.

be-2PL-IMP in country.

‘You (plural) were in the country.’

(17) Ti I Marko beste u prirodi. *Ø/On

You and Marko be-2PL-IMP in country. ?/he

uzivase.

enjoy-2/3S.IMP‘You and Marko were in the country. He was enjoying himself.’

(18) Marko bese u prirodi. Ø uzivase.

Marko be-2/3S-IMP in country. enjoy-2/3S.IMP

‘Marko was in the country. He was enjoying himself.’

Example (16) is satisfactory with the null subject recovered by unique

agreement. However, Example (17) is unacceptable with a null subject be-

cause uzivase does not distinguish between 2nd and 3rd person singular,

so an overt pronoun (on) has to be inserted. Example (18) is satisfactory

without an overt pronoun, even though uzivase is not di¤erentiated for

person, because the null subject is recovered by reference to an antecedent

in context, namely Marko, the subject of the verb in the first sentence in

the example. This is impossible in Example (17), because ti i Marko, the2nd person plural subject of bevte, gives no indication whether a null sub-

ject for uzivase in the subsequent sentence is 2nd or 3rd person. Macedo-

nian and Bulgarian also have syncretic forms in the 2nd and 3rd person

singular of the aorist and imperfect tenses and show a similar incidence

pattern for overt and null thematic subjects.

In Amharic, the simple imperfect tense is frequently prefixed by a com-

plementizer like be- ¼ if, so that the basic form ttedars (you [masc]/ she

come[s]) becomes bettedars ¼ (if you [masc] /she come[s]). Any overtpronoun precedes such a complex form. Object pronominals may occur

as su‰xes or infixes and may change according to the preceding vowel

or consonant of the verb to which they are infixed or su‰xed. Examples

showing situations where overt subject pronouns must occur follow.10

(19) Anta-nna Hanna eddelann-occ n-accehu. Ø Basa?at-u

You-and Anne fortunate-PL be-2PL on-hour-DEF

be-ttedars-u, Ø

if-2PL-arrive.IMP-2PL

568 M. D. Cole

Page 11: Null Subjects a Reanalysis of the Data

t-agan-u-t-all-accehu.

2PL.SUB-find-PL-3MS.OB-MVM-2PL.SUB

‘You and Anne are fortunate. If you arrive on time, you will catchhim.’

(20) Anta-nna Hanna eddelann-occ n-accehu. *Ø/erswa

you-and Anne fortunate-PL be-2PL. she

basa?at-u be-ttedars, Ø

on-hour-DEF if-arrive-2M/3F.IMP

te-ssallam-all-accehu.

2PL.SUB-win-MVM-2PL.SUB

‘You and Anne are fortunate. If she arrives on time, you (pl) willwin.’

(21) Hanna eddelanna n-acc. Ø basa?at-u

Anne fortunate-F be-3F. (she) on-hour-DEF

be-ttedars, Ø sera e-sat’-at-alla-hu.

if-arrive.2M/3F.IMP, job 1S-give-her-MVM-1S

‘Anne is fortunate. If she arrives on time, I shall give her the job.’

The null subject of the main clause verb in the 2nd sentence of each ofExamples (19)–(21) and the null subject of bettedarsu in Example (19)

are all recovered by unique subject verb agreement. However, bettedars,

the verb in the subordinate clause of the 2nd sentence of Examples (20) to

(21) is ambiguous between 2nd person masculine singular and 3rd person

feminine singular. Example (20) is unacceptable with a null subject for

ambiguous bettedars because the 2nd person plural Anta-nna Hanna in

the initial sentence is not an antecedent capable of identifying that sub-

ject, so an overt subject (erswa) is necessary. Example (21) is, however,acceptable with a covert subject for bettedars, because it is recovered in

the absence of unique subject verb agreement by reference to Hanna, the

antecedent in the previous sentence in the example. A similar pattern is

available in Arabic where the ambiguity is between the 3rd person femi-

nine singular and 2nd person masculine singular of the present. Thus, the

languages so far considered show that, first of all, unique subject verb

agreement is relied upon for the recovery of null thematic subjects. Then,

in the absence of such unique agreement, they are recovered by referenceto an antecedent in context. Finally, if neither unique agreement nor an

acceptable antecedent is available, overt pronouns are resorted to. The

recovery of null subjects by reference to an antecedent in context in the

presence of syncretic verb agreement in rich agreement languages has

also been observed to occur in Estonian, Turkish, Hindi/Urdu, Bengali,

Imbabura Quechua and Catalan.11 Italian provides a slightly more com-

plicated picture. It has syncretic forms in the 1st, 2nd and 3rd person

Null subjects 569

Page 12: Null Subjects a Reanalysis of the Data

singular of the present subjunctive, and in the 1st and 2nd person singular

of the imperfect subjunctive. Let us observe some examples:12

(22) Pavarotti1 dice che mangia gli spaghetti.

Pavarotti1 say-3S.PRES that eat-3S.PRES the spaghetti

‘Pavarotti says that he eats spaghetti.’

(23) Maria ed io1 siamo amici. Bisogna

Maria and I be-1PL.PRES friends. be-necessary.3S.PRES

che Ø1 sappia la verita

that (I) know1/2/3S.PRES.SUBJ the truth

‘Maria and I are friends. It is necessary that I know the truth.’

(24) *Maria1 ed io siamo amici. Bisogna che Ø1 sappia la verita.

(25) Maria1 ed io siamo amici. Bisogna che lei1‘Maria and I are friends. It is necessary that she

sappia la verita.

know the truth.’

(26) Maria e un amica. Bisogna che

Maria be-3S.PRES a friend. Be-necessary-3S.PRES that

Ø Sappia la verita.

(she) know-1/2/3S.PRES.SUBJ the truth.

‘Maria is a friend. It is necessary that she know the truth.’

In Italian, a thematic null subject is, prima facie, recoverable when it isthe subject of a verb with unique subject agreement like mangia in Exam-

ple (22). In Example (23), present subjunctive sappia is ambiguous be-

tween 1st, 2nd and 3rd person singular and there is no antecedent in con-

text by reference to which its subject can be recovered since the subject of

the initial sentence is 3rd person plural. The example is acceptable, how-

ever, without an overt pronoun, but only with a preferred 1st person in-

terpretation.13 For a 3rd person interpretation to work, it is necessary to

insert an overt 3rd person pronoun as in Example (25). Example (26) isacceptable without an overt pronoun but with a 3rd person singular inter-

pretation of the null subject because this is recovered by reference to

Maria, its antecedent in the 1st sentence in that example. A pattern of

preferred interpretation is also possible with syncretic forms in the first

and second person of the imperfect subjunctive. Consequently, one can

see that, in Italian, null thematic subjects may be recovered by reference

to unique agreement and, in its absence, by reference to a suitable ante-

cedent in context. In the absence of both, it may have a preferred inter-pretation. If the preferred interpretation is not the one required, an overt

pronoun must be used. This mirrors the position in all other languages

examined, except that, in Italian, preferred interpretations come into

570 M. D. Cole

Page 13: Null Subjects a Reanalysis of the Data

play after context has failed to di¤erentiate the subject of syncretic verb

forms and before resort is finally had to overt pronouns.

2.3.2. Languages with no agreement and null thematic subjects. Lan-

guages like Chinese, Japanese and Korean have null subjects and no sub-

ject verb agreement (other than for honorifics in the case of Japanese and

Korean). Overt pronouns occur in subject position in these languages forthe same two pragmatic reasons of focus and change of topic as in null

subject languages with rich agreement morphology. The languages also

obligatorily use overt subject pronouns in other situations parallel to

those for which they are used in rich agreement languages like Spanish

and Serbian, as the following Chinese examples show:14

(27) Zhang he ta de fu mu canguan le bowugan.

Zhang and he of parents visit PAST museum.

Ran hou Ø qu le juyian.

after that go PAST theater.

‘Zhang and his parents visited the museum. Then they went to the

theater.’(28) Zhang he ta de fu mu canguan le bowugan

Zhang and he of parents visit PAST museum.

Ran hou ta qu le juyian.

after that he go PAST theatre.

‘Zhang and his parents visited the museum. Then he went to the

theatre.’

In Example (27), the null subject in the 2nd sentence is recovered by ref-

erence to the 3rd person plural antecedent Zhang he ta de fu mu in the 1st

sentence. In Example (28), the same antecedent gives no clue as to the

identity of a 3rd person singular feminine subject in the second sentence,

so overt pronoun ta has to be inserted. Examples are also available to

demonstrate a similar situation in Japanese and Korean.So, in these languages that lack subject verb agreement (other than for

honorifics), covert pronouns are permitted where they are recovered by

reference to available antecedents, but otherwise overt pronouns are re-

sorted to. Thus, null subjects occur in parallel circumstances to those

in which they occur in languages with rich agreement, like Spanish and

Serbian, insofar as, here, as there, if agreement does not properly recover

them (which, here, being absent, it cannot do), an antecedent in discourse

may be resorted to in order to do so.

Null subjects 571

Page 14: Null Subjects a Reanalysis of the Data

2.3. Languages that lack null subjects

The situation in so-called non-null subject languages will now be exam-

ined, drawing parallels and contrasts with that in so-called null subject

languages.

Norwegian and Swedish have no person/number verbal morphology,

but, unlike Korean, lack null subjects. The following Swedish exampleillustrates the position.15

(29) Mannen kom hem. Forst oppnade *Ø/han

the man come-PAST home. first open-PAST. hedorren, sedan . . .

the door then . . .

‘The man came home. First he opened the door, then . . .’

Here, the null thematic subject in the 2nd sentence has an antecedent incontext which might recover it in the form of the subject of the preceding

sentence, mannen, but the example is still unacceptable without the inser-

tion of overt pronoun han.

Evidence that even rich agreement does not guarantee null thematic

subjects further complicates the picture. The following example in Ice-

landic illustrates.16

(30) �er komi¶ alltaf of seint. Viti¶ �er/*Ø

you come-2S.PRES always too late. know-2S.PRES you

ekki hvenœr vi¶ byrjum a¶ vinna herna.

not when we come-1PL.PRES to work here?

‘You are always late. Don’t you know when we come to workhere?’

Example (30) is unacceptable with a covert subject in its second sentence,

even though this has an antecedent in the preceding sentence and the

verb of which it is subject, namely viti¶, has unique subject agreement.Accordingly, insertion of the overt thematic pronoun ber is necessary for

acceptability. Thus, thematic null subjects are not possible in this lan-

guage, although, in various circumstances, the means to recover them ap-

pear prima facie to exist. In fact, it is remarkable that Icelandic, with

nearly totally discrete verbal morphology, which might have recovered

thematic null subjects on a wide scale, lacks them completely. Sigur¶sson

(1993) explains this by pointing out that Icelandic lost thematic null sub-

jects at a certain stage in its history without concomitant loss of agree-ment and concludes that such agreement was never anaphoric in this

language and that the ability of antecedents in context to identify null

thematic subjects has been lost.

572 M. D. Cole

Page 15: Null Subjects a Reanalysis of the Data

Thus, there is prima facie a disparate mixture of circumstances to cate-

gorize. In Swahili Example (12), Spanish Example (13), Serbian Example

(16), Amharic Example (19), and Italian Example (22), thematic null sub-

jects occur with unique verb forms. In Icelandic Example (30), such a

thematic null subject cannot occur even with rich agreement. In Spanish

Example (15), Italian Example (26), Serbian Example (18), Amharic Ex-

ample (21), and Chinese Example (27), a null subject occurs with a syn-cretic verb form but is recovered by reference to an antecedent in previous

context. While Swahili does not require context to identify null thematic

subjects since all verbal a‰xes are unique, Chinese still has such null sub-

jects without any person or number morphology to identify them. Also,

while Example (27) shows that Chinese has null subjects recovered by

reference to an antecedent in context, Example (29) shows that Swedish,

also with no verbal agreement, does not.

3. Morphological maximality and determination of the occurrence of null

subjects from context

3.1. Recovery by agreement or from context or both

While hitherto agreement has generally been seen as the driving force be-hind null subjects, it is clearly not so in languages where it is absent in

any relevant form. Accordingly, what is required is a concept that allows

null subjects to occur in Japanese for precisely the same reasons as in

Bengali, Spanish or Swahili and also encompasses the fact that thematic

null subjects occur in agreement rich Spanish, but not in agreement rich

Icelandic. While a purely agreement-based approach is clearly not satis-

factory, there is clear evidence that an approach based entirely on contex-

tual identification is also unsatisfactory, including Examples (31)–(35)from split ergative Pashto.17

(31) Ø Mana xwr-em.

apple eat-1M.S.PRES

‘I eat the apple.’

(32) *Ze Ø xwr-em.

I (it) eat-1M.S.PRES

‘I eat it (e.g., the apple).’(33) *Ø Mana we-xwar-a.

apple PFT-eat-3F.S

‘I ate the apple.’

Null subjects 573

Page 16: Null Subjects a Reanalysis of the Data

(34) Ma Ø we-xwar-a.

I (it) PFT-eat-3F.S

‘I ate it (e.g., the apple).’

The verb in present tense Examples (31) and (32) agrees in person, num-

ber and gender with the subject and only null subjects (not null objects)

are possible. On the other hand, the agreement on the verb in past tense

Examples (33) and (34) is object related and only null objects (not null

subjects) are possible. This shows such a clear connection between the in-

cidence of null subjects and the presence of agreement that agreement

cannot be ruled out as an element in the determination of their incidence.Further evidence that agreement is a factor in the incidence of null sub-

jects comes from Breton, Celtic languages generally, Standard Arabic

and Moroccan Lebanese and Beni Hassan Arabic, Alvdalsmalet and

Angami. Stump (1984) points out that, in Breton, thematic null subjects

occur only when verb subject agreement is present and do not occur

when it is absent. Furthermore Doron (1988) refers to a complementarity

between verbal agreement and overt subjects in Celtic languages gener-

ally. Kenstowicz (1989) provides similar evidence in Bani-Hassan Arabic,showing that, in expressing the past, the perfect (þ tense þ person) form

can have a null subject, while the participial (þ tense � person) form can-

not. Aoun, Benmamoun and Sportiche (1994) point out that Standard

Arabic has both SV and VS orders in simple clauses, In SV order the

verb agrees with the subject for person, number and gender whilst in VS

order the verb agrees with the subject for person and gender but not for

number. On the other hand Moroccan and Lebanese Arabic have verb

subject agreement for person, number and gender for both orders. Tocomplement this, Benmamoun (2000) points out that in all these three

Arabic languages, null subjects can only occur where there is verb subject

agreement for number as well as for person and gender. Further evidence

comes from Alvdalsmalet, which Sigur¶sson (1993) reports as allowing

null subjects only in the 1st and 2nd person plural, the only places where

rich agreement occurs. Sigur¶sson points out that the first, second, and

third persons singular of the verb in this language are identical and that

the third person plural is most frequently the same as them. Finally, evi-dence comes from Angami, where Giridhar (1980) points out that only

stative verbs expressing mental states, processes and attributes take sub-

ject agreement and only with such verbs do thematic null subjects occur.

The failure of either a totally agreement based or a totally context

based solution to work seems to reflect the division of languages made

by Tsao (1977), cited in C-T. J. Huang (1984), into sentence oriented

ones, like English, Spanish and Italian, and discourse oriented ones, like

574 M. D. Cole

Page 17: Null Subjects a Reanalysis of the Data

Chinese, Japanese and Korean. In fact, Li and Thompson (1976: 50)

earlier distinguished the following four possibilities:

Subject prominent English

Topic prominent Chinese

Subject and topic prominent Japanese

Neither subject nor topic prominent Tagalog

They saw discourse oriented languages as topic prominent, having topic

NP deletion and the syntactic content of empty categories determined

from outside the sentence. Rather than just categorizing languages as dis-

course or sentence oriented, it seems that, in order to achieve a unifying

theory, these kinds of di¤erences need to be seen as points on a spectrum

rather than oppositions. In fact, Li and Thomson’s view that Japanese is

both subject and topic prominent suggests some blurring of the dividing

line between [þdiscourse orientated/�sentence orientated] languages and[þsentence/�discourse oriented] ones. This gives further credence to the

idea that what is needed is a bridging operation between languages like

Spanish and Italian, which are construed as being sentence orientated,

apparently requiring predominantly unique subject verb agreement to de-

termine whether thematic null subjects occur from within the sentence,

and Chinese, Korean and Japanese, each of which, to its own degree, is

discourse orientated and allows the content of empty categories to be

determined to a considerable degree from outside the sentence. Such abridging operation must involve a concept that incorporates both agree-

ment and context in a solution that can apply as well to rich agreement

Spanish as to context driven Chinese. This will be seen as a two part

problem. Initially, the agreement facet will be addressed, but this itself

will throw light on the context aspect, which will be subsequently dis-

cussed.

3.2. Situation in Bengali

A key to the involvement of subject verb agreement in the crosslinguisticincidence of null thematic subjects seems to lie in Bengali, a null subject

language with subject verb agreement for person but not for number.18

However, it has personal pronouns, singular and plural, for all persons

and modes of address. Like other null subject languages, it uses such pro-

nouns for focus, as in the following example:19

(35) Ami bajare jacchi

‘I(focused) am going to-town.’

Null subjects 575

Page 18: Null Subjects a Reanalysis of the Data

Overt pronouns are also used in Bengali for change of topic as Examples

(36) and (37) illustrate.

(36) Budhbar-e amra Calcutta-e gelum.

on-Wednesday we to-Calcutta go-1-PAST.Brihoshpothibar-e Ø gari kinlum. Shonibar-e Ø bari

on-Thursday car buy-1PAST. on-Saturday home

firlum.

come-1PAST

‘On Wednesday we went to Calcutta. On Thursday we bought a

car. On Saturday we came home.’

(37) Budhbar-e amra Calcutta-e gelum.

on-Wednesday we to-Calcutta go-1PAST.Brihoshpothibar-e Ø gari kinlum. Shonibar-e ami

on-Thursday car. buy-1PAST. on-Saturday I

bari firlum.

home come-1PAST

‘On Wednesday, we to Calcutta went. On Thursday, we bought a

car. On Saturday, I came home.’

In Example (36), amra, the 1st person plural subject of the 1st sentence is

an antecedent providing identification for the null subject of the 2nd and

3rd sentences in the example. In Example (37), to change the reference of

the subject of the 3rd sentence to 1st person singular, insertion of overt

subject pronoun ami is necessary. However, as in other null subject lan-

guages, such pragmatic use of overt pronouns gives no definition of theincidence of thematic null subjects. Examples (38)–(40) provide a better

indication of this.

(38) Iqbal Calcutta-e budhbar-e gœlo.

Iqbal to-Calcutta on-Wednesday go-3PAST.

Brihoshpothibar-e Ø gari kinlo.

on-Thursday car buy-3PAST

‘Iqbal went to Calcutta on Wednesday. On Thursday he bought a

car.’

(39) Shonibar-e Iqbal Calcutta-e jabe. Ø gari

on-Saturday Iqbal to-Calcutta go-2/3FUT. car

kinbe Ø Restaurant-e jabe.

buy-2/3FUT. to-restaurant go-2/3FUT.

‘On Saturday, Iqbal will go to Calcutta. He will buy a car. He willgo to a restaurant.’

(40) Shonibar-e Iqbal Calcutta-e jabe. Ø Gari

on-Saturday Iqbal to-Calcutta go-2/3FUT. car

576 M. D. Cole

Page 19: Null Subjects a Reanalysis of the Data

kinbe. Tumi restaurant-e jabe.

buy-2/3FUT you-S to-restaurant go-2/3FUT

‘On Saturday, Iqbal will go to Calcutta. He will buy a car. Youwill go to a restaurant.’

In Example (38), the person of the null subject is recovered by reference

to unique subject verb agreement, but recovery of number comes from theantecedent in the form of Iqbal, the subject of the initial sentence. In Ex-

ample (39), the verbs kinbe in the 2nd sentence and jabe in the 3rd are

ambiguous between 2nd and 3rd person, but a null subject is still possible

because, with an intended 3rd person interpretation, it can be recovered

by reference to the antecedent, Iqbal, the 3rd person singular subject of

the 1st sentence in the example. In Example (40), an overt pronoun must

occur as the subject of the 3rd sentence if a second person interpretation

is intended, since the verb jabe is indistinct between 2nd and 3rd personand does not give such an interpretation. So, in Bengali, where only

person is the norm for the morphological recovery of the subject, a null

subject may occur, just as in Serbian, Amharic, or Spanish, where it is

recovered by reference to that morphology and, in the absence of that,

by reference to a suitable antecedent in context. However, where it re-

mains unrecovered, an overt pronoun is required.

3.3. Morphological maximality, antecedents and null subjects

Thus, Bengali, with verbal morphology for person only, provides a bridge

between Chinese, where thematic null subjects are always recovered byreference to the features of an antecedent and morphology plays no part,

and Italian, where they are morphologically recovered for person and

number. In fact, if we examine the morphological recovery of thematic

null subjects in all languages where they occur, it can be seen to form a

scale, at one extreme of which is Chinese, where recovery by reference to

verbal morphology is nil and the null subject is recovered by reference to

context alone. Then, in Bengali, a null subject is morphologically recov-

ered for person by verb subject agreement, except when such agreementis syncretic, when such a subject may be recovered by reference to con-

text. Next come Spanish and Italian, where a null subject is recovered

for person and number by subject verb agreement, except when such

agreement is syncretic, when such a subject may be recovered by reference

to context. Finally, in Tarifit, the norm is for the null subject to be recov-

ered for person, number and gender by subject verb agreement, as in Ex-

ample (41).20

Null subjects 577

Page 20: Null Subjects a Reanalysis of the Data

(41) Kurizzn1 y-nna qa Ø ur y-ssin

everyone1 3M.S-say that not 3M.S-know-PAST

ad y-ghinni.

to M.S-sing

‘Everyone1 said that he1 does not know how to sing.’

This appears to give a unified account for null subject languages, in so far

as we can see a scalar progression in the degree to which null thematicsubjects are recovered by reference to morphology. At one end of this

scale is Chinese, where context is all-prevailing in the recovery of null

thematic subjects and subject verb agreement, being absent, plays no

part, and at the other end is Tarifit, where, prima facie, recovery of such

null subjects relies on agreement for person, number and gender. How-

ever, is this the end of the scale, or should one not see the features of an-

tecedents in context themselves as broken down into further unspecified

factors X, Y, Z by reference to which null subjects can be recovered, justas context:

(i) recovers number when a null subject is used with unique forms

in such a language as Bengali, which only has morphology for

person? or

(ii) recovers gender in such languages as Italian or Spanish whichonly have morphology for person and number?

E¤ectively what one can say is that every language has its own morpho-

logical maximality in terms of subject agreement. A null thematic subject

can occur provided that, in the first instance, subject verb agreement is

unique up to the point of morphological maximality. In Chinese, thepoint of morphological maximality is null. Nil morphological features

are required to satisfy this null morphological maximality and so a null

thematic subject may occur in the first place provided those features not

represented by morphological maximality are available by reference to an

antecedent in context. So, in Chinese, a null thematic subject may occur if

it is satisfactorily recovered by reference to an antecedent in context. In

Bengali, morphological maximality is represented by person, so a null

thematic subject may occur if unique agreement recovers person, pro-vided that features not represented by morphological maximality are re-

covered by reference to an antecedent in context. In Spanish and Italian,

morphological maximality is represented by person and number. A null

thematic subject may occur in the first place if unique subject verb agree-

ment recovers its person and number, and features not represented by

morphological maximality are recoverable from an antecedent in context.

In Tarifit, morphological maximality is represented by person, number

578 M. D. Cole

Page 21: Null Subjects a Reanalysis of the Data

and gender, and any features beyond that are recovered by reference to

an antecedent in context.

In Arabic, morphological maximality is represented by person, number

and gender. This creates a slight problem since, in this language, gender is

only morphologically recovered in the second and third person singular

and plural. The concept therefore needs some modification to allow null

subjects to occur in the 1st person singular and plural without agreementfor that person being specified for gender. The only practical way to

achieve this without other undesirable results is to specify the e¤ective

facts. Morphological maximality must be across a whole language for

person and number, but can be across only certain persons in a language

for other features. Thus, in Arabic and other languages with similar verb

subject agreement characteristics, morphological maximality is for person

and number in the 1st person, but for person, number and gender in the

second and third persons.Having dealt with this di‰culty, we can now say that a null thematic

subject may occur if unique verb subject agreement recovers features up

to the point of morphological maximality and all other features are recov-

erable by reference to an antecedent in context.

3.4. Recovery from context

In the previous section, a framework was established in which agreement

could be incorporated into an overall definition of the incidence of null

subjects even in a language like Chinese, which lacks agreement morphol-

ogy. Incidental to this, it was established that context always plays some

part in the recovery of null subjects. Let us now pinpoint exactly what

part that is.

So far we have said that

(A) in an example like the following in Spanish with syncretic

agreement:

Juan llegaba a casa. Ø

John arrive-1/3S.IMPERF to home. (he)

Tenıa las llaves

have-1/3S.IMPERF the keys

‘John was arriving home. He had the keys.’

null subjects are recovered by reference to an antecedent in context;

(B) in Chinese, null subjects are totally identified by reference to ante-

cedents in context; and

Null subjects 579

Page 22: Null Subjects a Reanalysis of the Data

(C) in Bengali, Spanish and Tarifit, where there is unique verbal

agreement, context recovers all features of thematic null subjects

beyond the point of morphological maximality

While it is plain in all these circumstances that context is a determinant of

the incidence of thematic null subjects, what is needed is a single state-

ment of its contribution in this respect that will cover all its occurrencesin all languages. It is clear that full recovery of thematic null subjects

from context is needed in Chinese, and it appears necessary in the circum-

stances of A above, where agreement does not recover the null thematic

subject up to the point of morphological maximality. It seems self-evident

that it fully recovers them in the circumstances of C above too, but can

we say that it is a requirement or is it just the case that contextual recov-

ery is only necessary beyond the point of morphological maximality? The

following Spanish example gives the answer:

(42) *[Juan y Juanita]1 llegaron a casa. Ø1

Juan and Juanita arrived to house.

Abrio la puerta.

open-3S.PAST the door

‘Juan and Juanita arrived home. (He/she) opened the door.’

Assuming no deictic context, this is unacceptable because the null subjectin the 2nd sentence is recovered as 3rd singular by verb subject agree-

ment, but is co-indexed with a 3rd person plural antecedent. If contextual

recovery were merely required beyond the point of morphological maxi-

mality, then such co-indexation would be acceptable. This is because the

failure of the singular agreement feature to accord with the plural ante-

cedent would be of no consequence because number is a feature within

morphological maximality and thus would not need to be contextually re-

covered. To ensure that contextual recovery does not contradict morpho-logical recovery, we must say that, for there to be a 3rd person singular

null subject, it must be fully contextually recovered. Thus, as well as be-

ing recovered up to morphological maximality (with a tolerance for

syncretisms) by agreement, a null thematic subject must be fully recov-

ered by reference to an antecedent in context. A formulation of this

approach follows:

A null thematic subject may occur when it is both(a) recovered by reference to an antecedent in context; and

(b) recovered by agreement up to the point of morphological unifor-

mity (with a tolerance for sycretisms)

580 M. D. Cole

Page 23: Null Subjects a Reanalysis of the Data

4. An essential di¤erence between null subject and non-null subject

languages

A clear di¤erence is now apparent between languages where thematic null

subjects occur and those where they do not. In languages where thematic

null subjects occur:

(i) they are recovered up to the point of morphological maximality

or as far as possible towards it, given syncretisms, by subject

verb agreement.

(ii) recovery as at (i) is supplemented by reference to an antecedent incontext.

(iii) where a combination of (i) and (ii) fails to recover a thematic null

subject, it may have a preferred interpretation (as in Italian).

(iv) where neither (i) and (ii) nor (iii) provide an identity for a the-

matic null subject, an overt pronoun must be substituted.

In prototypically non-null subject languages like Swedish with no verb

agreement:

(i) recovery by subject verb agreement is up to the point of morpho-

logical maximality which is nil, as in Chinese, and this is not sup-

plemented by reference to an antecedent in context; so

(ii) overt pronouns must be used.

There is an essential di¤erence between the two patterns. In languages

where null thematic subjects occur, such as Spanish, Italian, Bengali and

Chinese, their recovery by subject verb agreement up to the point of mor-

phological maximality, (or as far as possible towards it, given sycretisms)

can be supplemented by recovery by reference to an antecedent in context

before overt pronouns are resorted to. However, in prototypical non-null

subject Norwegian and Swedish, agreement can recover a null subject up

to the point of morphological maximality, which is null. However, such asubject is not recovered by reference to an antecedent in context, so resort

is had immediately to overt pronouns, which must always be used.

The most telling picture is given by comparing how this scheme of

things works in Chinese and Swedish. In both, recovery by agreement,

(which is lacking) is up to morphological maximality, which is null.

In Chinese, a null thematic subject is recovered by reference to an ante-

cedent in context, but overt pronouns are immediately resorted to in

Swedish.One could conclude on this basis that languages in which null thematic

subjects may occur have a certain contextual strength and that those

where they do not occur have a certain contextual weakness. Whether a

Null subjects 581

Page 24: Null Subjects a Reanalysis of the Data

language is contextually weak or strong might then be construed as re-

flected by whether a thematic null subject can be recovered by reference

to an antecedent in context. This approach then lends itself to the solu-

tion of another problem. The reason why null subjects are possible in

Spanish but not in Icelandic, both of which have rich subject verb agree-

ment, could be put down to the fact that, in the former, they can be re-

covered by reference to an antecedent in context, whereas, in the latter,this is not the case. Supporting evidence for this conclusion comes from

the fact that, in Icelandic, as already pointed out, Sigur¶sson (1993) con-

cludes that thematic null subjects were lost without any appreciable loss

in agreement morphology and that, in Old Icelandic, thematic null sub-

jects were recovered by reference to free co-indexing with elements in con-

text. This situation could be interpreted in terms of Icelandic changing

from being a contextually strong to a contextually weak language as it

lost thematic null subjects. Similarly, French and English, languages thatlack thematic null subjects even with the limited verb forms with identify-

ing morphology, could be construed as being contextually weak.

In Pashto, looking back to Examples (31) to (34), null subjects can oc-

cur in intransitive past tense sentences and generally in the present tense,

where there is verb subject agreement for gender, person and number but

not in transitive sentences in the past where there is no such subject verb

agreement. This language can be construed as contextually strong since

null thematic subjects do occur in it. The di¤erence in the incidence ofnull subjects between its tenses can be related to the fact that morpholog-

ical maximality for subject agreement is for gender, person and number,

so that thematic null subjects can occur where they are recovered by this

level of subject verb agreement, namely in intransitive past tense sen-

tences and generally in the present tense. On the other hand, they cannot

be so recovered in transitive sentences in the past tense, where this level of

subject verb agreement is absent, since the agreement there is object

related. This illustrates how the concept developed in this article operatesin a partial null subject language where the presence or absence of null

thematic pronouns varies with the presence or absence of subject verb

agreement.21

5. Consequences of the above theory

The first condition for a null thematic subject to occur is for it to be re-covered by reference to an antecedent in context in a contextually strong

language. If this does not occur, then an overt pronoun is required. From

this we may conclude that the syntactic licensing of thematic null sub-

582 M. D. Cole

Page 25: Null Subjects a Reanalysis of the Data

jects is completely redundant. There are in fact already good argu-

ments against there being a licensing requirement for null subjects. First

there is no reason why a purely phonological opposition (pronounced/

unpronounced) should be the subject of syntactic licensing. This objection

to the syntactic licensing of null subjects is in fact highlighted by Speas

(1994) who points out in relation to head licensing that the licensing con-

dition necessitates that the grammar allow some heads to be designated aslicensors of a category with particular phonological properties (namely

the property of being unpronounced). There are no designations equiva-

lent to head licensing in other components of syntax involving other pho-

nological properties. Furthermore, syntactic licensing would necessitate

that a particular lexical entry, a null thematic subject, is subject to a spe-

cial syntactic requirement by virtue of its phonological status, and there

do not seem to be other cases of this.

This does at first seem to have one rather unfortunate consequence,since with thematic null subjects not requiring licensing, expletive null

subjects seem to have been left isolated. However, there appear to be

some arguments for treating them as a separate empty category. First,

whilst, in languages in which they occur, covert thematic subjects exist

alongside overt thematic subjects, there is generally no overt version

of covert non-thematic subjects in such languages.22 Also, if null non-

thematic subjects were the same entities as null thematic subjects but

without thematic content, then it would be logical to assume that thesame relationship existed for overt non-thematic and thematic subjects.

This works very well for non-thematic it, which with thematic content be-

comes thematic it but there is no thematic pronominal subject equivalent

of expletive subject pronoun there. Furthermore, trying to assimilate non-

thematic null subjects and thematic null subjects into the same entity in

such a language as Spanish or Italian is a little peculiar because the never

expressed non-thematic pronoun suddenly becomes expressible when it

gains thematic content.

6. Conclusion

The aim of this article has been to establish a theory of why thematic pro-

nouns are covert or covert, focusing on subjects. Analysis of an array of

data established that, in languages with wide scale agreement and null

subjects, prima facie, recovery of such null subjects is made first by refer-ence to discrete agreement and if that fails by reference to an antecedent

in context. Subsequently, preferred interpretations are resorted to and

then overt pronouns. Every language is at a certain point on a scale of

Null subjects 583

Page 26: Null Subjects a Reanalysis of the Data

morphological maximality up to which agreement must recover null sub-

jects. It is further concluded that, in languages where null subjects occur,

resort is had to an antecedent in context to complete their recovery,

whereas in languages where they do not occur, this is not possible. This

is put down to the fact that some languages are contextually strong and

others are contextually weak. This proposal has the consequence that the

syntactic licensing of thematic null subjects is redundant. Consequently, itis deemed necessary to treat non-thematic null pronouns as a separate

empty category that may require syntactic licensing.

Further research is also well advanced in relation to the application of

an adapted form of Ariel’s (1988, 1990, 1994, 2001) theory of accessibility

to the concepts of contextual strength and weakness.23

Received 4 November 2003 University of Manchester

Revised version received

7 August 2006

Notes

1. Thanks are due to my PhD supervisor, Professor Nigel Vincent for his advice and en-

couragement long after I had ceased to be his responsibility. They are also due to two

anonymous reviewers. I should also like to express my gratitude to all those who pro-

vided me with examples, including Ketmanee Ausmangokol, whose Thai examples I

have been unable to use. Errors of any kind in this article are my responsibility. Corre-

spondence address: 14 Heathview, Kellbrook Crescent, Salford, Manchester M7 3GH,

United Kingdom. E-mail: [email protected].

2. C.-T. James Huang’s operator variable analysis was a bold and innovative solution to

the problem and the vast majority of the theory upon which the objections were based

appear not to have been expounded in the literature until after the publication of his

article.

3. Example supplied by Feng Shou Gu, University of Manchester.

4. Example taken from Chierchia (1989).

5. The table has been altered from that on p. 77 of Enfield (2007), but only in order better

to reflect his subsequent observation that man2, haw2 and phen1 are used as both

plural and singular. The numbers refer to tones. Tone 1 has a level contour around

the middle pitch range, tone 2 a high-rising contour, beginning around the mid-range

and going to high, tone 3 a low rising contour, beginning around the bottom of the

pitch range and rising sharply, (sometimes pronounced as a low level tone without a

rising o¤set), tone 4 a high-falling contour, beginning at the top of the pitch range

and falling sharply, tone 5 a mid-falling contour, starting at the middle of the pitch

range and falling to low.

6. Information gathered from Morey (2005) and examination of glossed texts Stephen

Morey kindly sent with his personal communication.

7. Turkish example provided by Dr. Evren Erem. Japanese examples supplied by Dr.

Chaoki Taoka.

584 M. D. Cole

Page 27: Null Subjects a Reanalysis of the Data

8. Example from Vitale (1981: 117).

9. Examples and information in Serbian, Macedonian and Bulgarian obtained respec-

tively from Vladislava Ribnikar of Nottingham University, Dr. Vesna Stojanovik of

Reading University and Dr. Ilya Nedin of the School of Slavonic and East European

studies at University College London.

10. Amharic examples supplied by Bodja Gelatcha and Semaw Mekonen, MSc’s at Man-

chester University. Their transcription has been confirmed by Dr. David Appleyard of

SOAS, whose patience in relation to this complex language I am extremely grateful for.

In the examples, I have followed Dr. Appleyard in denoting the component -all- used

in forming the compound imperfect tense as MVM (‘‘main verb marker’’). Amharic

has a simple imperfect tense, marked IMP in the examples, and a compound imperfect

tense. Both have separate functions, for which the reader is referred to Appleyard

(1995).

11. See Cole (1982).

12. Example (22) taken from Huang (2000). Examples (23)–(26) supplied by Dr. Cecilia

Goria of Nottingham University.

13. Regarding Examples (23) and (24), seven informants confirmed the first person pre-

ferred interpretation of the subject of sappia and three said that even if a first person

interpretation were intended, this would not be clear without an overt pronoun. This

reflects the view expressed in Cardinaletti (1997) that not all Italians accept such pre-

ferred interpretations.

14. Examples supplied Hongzhu Li and Zhao Wan, students at Manchester University and

Manchester Business School respectively, and confirmed by Wen Jiang, student at

Manchester University.

15. Swedish example supplied by Professor Kersti Borjars of Manchester University.

16. Icelandic example checked with Dr. Katrin Lund of University of Iceland.

17. Examples from C-T. J. Huang (1984).

18. Bengali also has considerable morphology in the second and third person for honorific

level.

19. Examples (35), (36), and (40) supplied by Dr. Gupta, ex-member of German Depart-

ment at Manchester University. All other Bengali examples provided by ‘‘Ike’’ Chow-

dury.

20. Information from Y. Huang (2000: 89)

21. According to Anders Holmberg, Finnish and Marathi appear to be partial pro drop

languages that, in certain circumstances, require pronominal subjects to be overt, even

with rich subject agreement. This requires further research.

22. In Finnish, according to Holmberg (2005), depending on the circumstances, an exple-

tive subject can be overt. Also, Adam Ledgeway (p.c.) informs me that overt expletive

subjects can occur in most Campanian dialects and some dialects of Calabria and

Sicily.

23. This research will need to address the fact recently discovered that Hindi seems to

allow unexpressed pronouns in a manner that is orthogonal to agreement. Informants

are being consulted.

References

Alexiadou, Artemis & Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order,

V-movement and EPP-checking. Natural Language and Linguistic Theory 16. 491–539.

Null subjects 585

Page 28: Null Subjects a Reanalysis of the Data

Appleyard, David. 1995. Colloquial Amharic. London: Routledge.

Ariel, Mira. 1988. Referring and accessibility. Journal of Linguistics 24. 65–87.

Ariel, Mira. 1990. Accessing noun phrase antecedents. London: Routledge.

Ariel, Mira. 1994. Interpreting anaphoric expressions: A cognitive versus a pragmatic

approach. Journal of Linguistics 30. 3–42.

Ariel, Mira. 2001. Accessibility theory: an overview. In Ted Sanders, Joost Schilperoord &

Wilbert Spooren (eds.), Text representation: Linguistic and psycholinguistic aspects, 29–87.

Amsterdam: John Benjamins.

Aoun, Joseph, Elabbas Benmamoun & Dominique Sportiche. 1994. Agreement, word order

and conjunction in some varieties of Arabic. Linguistic Inquiry 25(2). 195–220.

Benmamoun, Elabbas. 2000. The feature structure of functional categories: A comparative

study of Arabic dialects. Oxford: Oxford University Press.

Cardinaletti, Anna. 1997. Subjects and clause structure. In Liliane Haegeman (ed.), The new

comparative syntax, 33–63. London: Longman.

Chierchia, Gennaro. 1989. Structural meaning, thematic roles and control. In Gennaro

Chierchia, Barbara Partee & Raymond Turner (eds.), Properties, types and meaning,

131–166. Dordrecht & Boston: Kluwer.

Chomsky, Noam. 2001. Derivation by phrase. In Micheal Kenstowicz (ed.), Ken Hale: A life

in language, 1–59. Cambridge, MA: MIT Press.

Cole, Melvyn. 2000. The syntax, morphology and semantics of null subjects, Manchester:

University of Manchester dissertation.

Cole, Peter. 1982. Imbabura Quechua. Amsterdam: North Holland.

Cooke, Joseph Robinson. 1968. Pronominal reference in Thai, Burmese and Vietnamese,

(University of California Publications in Linguistics 52). Berkeley, CA: University of

California Press.

Doron, Edit. 1988. Subject on the complementarity of subject and verb agreement. In

Micheal Barlow & Charles Ferguson (eds.), Agreement in natural languages, 201–218.

Stanford, CA: CSLI Publications.

Enfield, N. J. 2007. A grammar of Lao. Berlin & New York: Mouton de Gruyter.

Gilligan, Gary. 1987. A cross-linguistic approach to the pro-drop parameter. Los Angeles,

CA: University of Southern California dissertation.

Giridhar, Puttushetra Puttuswamy. 1980. Angami grammar. Mysore: Central Institute of

Indian Languages.

Holmberg, Anders. 2005. Is there a little pro? Evidence from Finnish. Linguistic Inquiry

36(4). 533–564.

Huang, C.-T. James. 1984. On the distribution and reference of empty pronouns. Linguistic

Inquiry 15. 531–574.

Huang, C.-T. James. 1989. Pro-drop in Chinese: A generalized control theory. In Osvaldo

Jaeggli & Ken Safir (eds.), The null subject parameter, 185–214. Dordrecht & Boston,

MA: Kluwer.

Huang, Yan. 2000. Anaphora: A cross-linguistic study. Oxford: Oxford University Press.

Jaeggli, Osvaldo & Ken Safir. 1989. The null subject parameter. Dordrecht & Boston, MA:

Kluwer.

Kameyama, Megumi. 1985. Zero anaphora: The case of Japanese. Palo Alto, CA: Stanford

University dissertation.

Kenstowicz, Michael. 1989. The null subject parameter in modern Arabic dialects. In

Osvaldo Jaeggli & Ken Safir (eds.), The null subject parameter, 263–276. Dordrecht &

Boston: Kluwer.

Law, Peter. 1993. On null subjects and null arguments. Canadian Journal of Linguistics 38.

1–41.

586 M. D. Cole

Page 29: Null Subjects a Reanalysis of the Data

Li, Charles N. & Sandra A. Thompson. 1976. Subject and topic: A new typology of lan-

guage. In Charles N. Li (ed.), Subject and topic, 457–489. London: Academic Press.

Manzini, Rita & Anna Roussou. 1999. A minimalist theory of A-movement and control.

UCL Department of Phonetics and Linguistics Working Papers in Linguistics 11. 403–440.

Manzini, Rita & Leonardo Savoia. 2002. Parameters of subject inflection in Italian dialects.

In Peter Svenonius (eds.), Subjects, expletives and the EPP, 157–199. Oxford & New

York: Oxford University Press.

Morey, Stephen. 2005. The Tai languages of Assam: A grammar and texts. Canberra: Pacific

Linguistics.

Nakamura, Masaru. 1986. Japanese as a pro language. Linguistic Review 6. 281–296.

Neeleman, Ad & Krista Szendroi. 2007. Radical pro drop and the morphology of pronouns.

Linguistic Inquiry 38. 671–714.

Rizzi, Luigi. 1986. Null subjects in Italian and the theory of pro. Linguistic Inquiry 17. 501–

557.

Samek-Lodovici, Vieri. 1996. Constraints on subjects: An optimality theoretic analysis. New

Brunswick, NJ: Rutgers University dissertation.

Sigur¶sson, Halldor. Armann. 1993. Argument-drop in Old Icelandic. Lingua 9. 247–280.

Speas, Margaret. 1994. Null arguments in a theory of economy of projection. In Elena

Benedicto & Je¤ Runner (eds.), University of Massachusetts Occasional Papers in Linguis-

tics 17, 179–208. Amherst, MA: GLSA.

Stump, Gregory. 1984. Agreement vs. incorporation in Breton. Natural Language and Lin-

guistic Theory 2(3). 289–348.

Thompson, Laurence C. 1965. A Vietnamese grammar. London & Seattle: University of

Washington Press.

Tomioka, Satoshi. 2003. The semantics of Japanese null pronouns and its cross-linguistic

implications. In Kerstin Schwabe & Suzanne Winkler (eds.), The interfaces. deriving and

interpreting omitted structures, 321–339. Amsterdam: John Benjamins.

Tsao, Feng-Fu. 1977. A functional study of topic in Chinese: A first step towards discourse

analysis. Los Angeles, CA: University of Southern California dissertation.

Vitale, Anthony. 1981. Swahili syntax. Dordrecht: Foris.

Xu, Leilong. 1986. Free empty category. Linguistic Inquiry 17. 75–93.

Xu, Leilong & Terence D. Langendoen. 1985. Topic structures in topic prominent lan-

guages. Language 61. 1–27.

Null subjects 587

Page 30: Null Subjects a Reanalysis of the Data