DIANE LIL'LO-MARTIN

T W O K I N D S OF N U L L A R G U M E N T S SIGN LANGUAGE*

IN A M E R I C A N

1. I N T R O D U C T I O N

In some languages, such as Italian and Spanish, the subject of a sentence can be non-lexical, as in example (1). 1 (1) Mangia una mela. (Italian) (He-~she-) eats an apple.

Although the subject is not phonologically overt in these sentences, it is understood as a definite pronominal such as 'he' or 'she'. English, on the other hand, does not allow such NULL ARGUMENTS in tensed clauses. In this paper, I will investigate the appearance of null arguments, (that is, those subjects and objects which are not specified phonologically as a separate overt lexical pronoun or noun phrase), in yet another language, American Sign Language (ASL). It will be seen that there are two kinds of null arguments in ASL, one corresponding to the rtull arguments found in languages like Irish, and the other corresponding to the null arguments found in languages like Chinese. I will show why two analyses are needed for null arguments in ASL, and how this bears on the analyses of other languages with and without null arguments. I will also examine some other syntactic constructions which have been reported to be connected to null arguments, and explain why ASL does or does not manifest these constructions. Throughout, I am assuming in general the

* This research was supported in part by National Institutes of Health Grants #HD13249, NS15175, N819096, and by National Science Foundation Grant #BNS83-09860 to Drs. Ursula BeUugi and Howard Poizner at the Salk Institute for Biological Studies. I would like to thank the following people for their support, consulting, assistance, comments, and discussions of this research: Ursula BeUugi, Sandra Chung, David Corina, Carol Georgopolous, Edward Kiima, Lucinda O'Grady, Maureen O'Grady, Carol Padden, Sally Rice, Leslie Saxon, Dennis Schemenauer, and three anonymous NLLT reviewers. Illustrations were made by Frank A. Paul, copyright Dr. Ursula Bellugi, The Saik Institute. The following notation is used in the English translations. When a pronoun is not given in the original language (i.e., it is null), but is needed for a grammatical English translation, it will be included in the translation within parentheses, e.g. (she). If a pronoun is needed in English, and does not appear as a pronoun in the original language but is signified by verb agreement, it will appear in the translation within parentheses preceded by a hyphen, e.g. (-them), if it is a subject, or followed by a hyphen, e.g. (He-), if it is an object.

Natural Language and Linguistic Theory 4 (1986) 415-444 ~) 1986 by D. Reidel Publishing Company

416

DIANE

LILLO-MARTIN

t h e o r e t i c a l f r a m e w o r k of G o v e r n m e n t C h o m s k y (1981) a n d s u b s e q u e n t w o r k s .

and

Binding,

as o u t l i n e d in

I n A m e r i c a n Sign L a n g u a g e , for a l a r g e set of v e r b s , s u b j e c t a n d object are not necessarily distinguished from each other by word order or c a s e m a r k i n g . R a t h e r , t h e y a r e m a r k e d b y t h e m o v e m e n t of t h e v e r b in r e l a t i o n t o specific p o i n t s in s p a c e . T h i s s p a t i a l l y e x p r e s s e d s y n t a c t i c system has been called 'verb agreement' by researchers working on A S L . 2 T h i s p a p e r will e x a m i n e t h e null a r g u m e n t s of v e r b s t h a t use this s y s t e m to m a r k a g r e e m e n t , a n d null a r g u m e n t s of v e r b s t h a t d o n o t m a r k a g r e e m e n t . I t will b e s h o w n t h a t t h e null a r g u m e n t s to t h e s e t w o t y p e s of v e r b s a r e d i f f e r e n t i a l l y d i s t r i b u t e d a n d in f a c t s h o u l d b e e x p l a i n e d in d i f f e r e n t ways. Specifically, if a n i n f l e c t i o n a l m a r k e r is p r e s e n t , t h e effect, e v e n w h e n t h e p r o n o u n ' a g r e e d w i t h ' is null, is t h e s a m e as if an o v e r t pronoun were present, indicating that the empty category should be p r o n o m i n a l , pro. H o w e v e r , w h e n t h e r e is n o i n f l e c t i o n a l m a r k e r , t h e n t h e a p p e a r a n c e of null a r g u m e n t s is m u c h m o r e l i m i t e d , a n d t h e e m p t y c a t e g o r y will b e a n a l y z e d as a n o n p r o n o m i n a l ( W h - t r a c e ) e m p t y t o p i c . 3

2. V E R B A G R E E M E N T ' IN A S L A m e r i c a n S i g n L a n g u a g e is t h e v i s u a l - g e s t u r a l l a n g u a g e u s e d b y m o s t of t h e d e a f c o m m u n i t y in t h e U n i t e d S t a t e s a n d p a r t s of C a n a d a . T h e

2 Researchers who have discussed ASL verb agreement in various terms include Coulter (1979), Fischer (1974), Fischer and Gough (1978), Frishberg and Gough (1973), Kegl (1976), Lacy (1973), Liddell (1977), Meier (1982), and Padden (1983), and Shepard-Kegi (1986), among others. For anoverview see Klima and BeUugi (1979), or Wilbur (1979). The analysis to be presented here resembles in many ways an analysis of null arguments in ASL proposed independently by Shepard-Kegl (1986, pp. 480-491). Both analyses conclude that ASL is both a pro-drop language and a discourse-oriented language, with the same distinction between agreeing verbs and nonagreeing verbs serving to distinguish the two types of null arguments. However, many details of the analyses differ. Importantly, Shepard-Kegl assumes a different analysis of the internal structure of the ASL sign than the one assumed here. This analysis takes as morphologically significant many of the properties of ASL signs (such as the handshape, and aspects of the location and movement) which are considered here merely formational. With respect ~o the analysis of null arguments, the main difference is that !n some cases, Shepard-Kegl does not consider the verb movement to spatial loci as the agreement (AGR) which sanctions null arguments. Rather, the position of the signer's body with respect to the spatial loci, which Shepard-Kegl calls a clitic, sanctions empty arguments. Shepard-Kegl's analysis of the internal structure of the sign is highly complex, leading her to posit that the single inflected sign which I would represent by oGIVEb consists of 43 morphemes (p. 136). Under my analysis, this complexity is unneeded and unwarranted, at least for the analysis of the null argument structures discussed here. (See Lillo-Martin 1986b for further discussion of the similarities and differences between these two accounts.)

NULL

ARGUMENTS

IN AMERICAN

SIGN

LANGUAGE

417

position taken in this p a p e r is that when the surface effects of modality are stripped away, A S L will be seen to follow m a n y of the patterns proposed as universals for h u m a n language. For this reason, the study of A S L can be significant for proposed theories of universal grammar. Primary to an understanding of the v e r b a g r e e m e n t system of A S L is the notion of nominal association with loci in the signing space. Loci in space can be identified as associated with particular NPs. For referents which are pliysically present in the signing situation, the places which they occupy are generally understood as their associated loci. Thus, the locus of first person association is the signer's own chest (or the space directly in front of it); second person is the addressee's chest, and third person association, for those third persons in the immediate surroundings, is with the places at which these third persons are located. 4 For referents which are not physically present, abstract loci in the signing space in front of the signer's body are associated with each referent. 5 This is accomplished by signing the NP at some arbitrary locus in space, or making the sign and then pointing to the locus with the index finger, or b y gazing in the direction of the locus while making the sign. Pronominal reference is m a d e by indicating the loci associated with the intended referent. Thus, first person pronominal reference is m a d e by the signer pointing toward her own chest, and second person pronominal reference is m a d e by pointing toward the addressee's chest. Third person pronouns, when the referents are actually present, are likewise m a d e by pointing toward the appropriate persons. When the referents are not present, the abstract loci associated with each referent serve as the locations for the direction of the pronominal signs. Hence, after a signer has associated John with a locus on her right side, she can use a pronominal sign indicating this locus as a pronoun for John. T h e association of abstract loci with non-present referents remains throughout a discourse until a new framework is established. 6 An important fact about this system is that the n u m b e r of different

See Meier (1986) for discussion of the lack of distinction between second and third person pronominals in ASL. 5 Abstract loci are also often used for discussing hypothetical situations, or past or future situations, even when the referents are present. See Bahan and Petitto (1980) for discussion of the physical spacing of the abstract loci. The /1/ handshape (index finger only projecting from a closed fist) is used for personal pronominal reference; /B/ (all four fingers extended and closed together) for possessive reference; and/A/(closed fist with thumb extended) for reflexives. Note, however, that the distribution of the /A/ 'SELF' sign is not exactly parallel to English reflexives. See, for example, Fischer and Johnson (1982).

418

DIANE

LILLO-MARTIN

referents is theoretically unlimited. If there are three individuals who figure in a particular narrative that the signer is relating, then three loci will be established in space. If there are five or seven or ten individuals, the signer could divide up the space with that many distinct loci, but this tends not to be done because of perceptual and memory factors. This system of indexed pronouns thus more resembles overt referential indices than it does the categorical pronouns of spoken languages (Lacy, 1974; Lillo-Martin and Klima, 1986). The ASL verb agreement system makes use of the association of nominals with loci described above. ASL has a class of verbs which takes agreement morphology, and a class which does not. When agreement morphology is present, it is manifested in the following way. Agreement is signaled by changing the movement dimension of the verb root. Under this modulation, for a large set of verbs, the sign begins at the locus which the subject occupies, and terminates a t the locus of the object. With present referents, for example, to sign, I hate you, the basic sign 7 H A T E is articulated so that the movement begins at me, and moves toward and ends at you: 1HATE2. For referents which are not present, the verb is executed between the abstract loci. Thus, if John is associated with a locus to my right, and Mary is associated with a locus to my left, then movement of the sign root (HATE) from the locus on my right to the locus on my left means, John hates Mary. Similarly, if the verb moves from my left to my .right, it means, Mary hates John. This verb agreement serves as the strongest kind of cue for grammatical relations (Wilbur, 1979). s An example of the locus association and verb agreement morphology system is illustrated in Figure 1 below. The utterance consists first of associating 'the dog' with a locus on the signer's left, called a, and 'the cat' with a locus on the signer's right, called b. Verb agreement morphology is then illustrated with the verb BITE. In the figure, the movement of the verb BITE is from the left to the right; thus, the sentence means, '(The dog) bites (the cat)'. If the verb moved from right to left, it would mean, '(The cat) bites (the dog)'. Not all verbs, however, can be marked for agreement in this way. With noninflecting verbs, called "Plain verbs" by Padden (1983), SVO word

7 The notation used for the ASL examples is shown in the Appendix. The ASL verb agreement is an inflectional process rather than a process of cliticization, at least using the morphological criteria of Zwicky and Pullum (1983). See Lillo-Martin (1985). For a different view of agreement and cliticization in ASL, see Kegl (1985) and Shepard-Kegl (1986).

NULL ARGUMENTS

IN A M E R I C A N

SIGN LANGUAGE

419

DOG-aINDEX

CAT-blNDEX

oBITEb

(The dog) bites (the cat).

Fig. 1. ASL verb agreement morphology.

order usually serves to mark grammatical relations. Deviations in SVO order, with both plain and inflecting verbs, occur via a productive process of topicalization, which is also marked in ASL by a specific facial gesture and by the rhythmic grouping of the signs (Liddell, 1977). An example of (overt pronominal) word order used with nonagreeing verbs is given in Figure 2 below. This figure assumes the same association of nominals with loci as in Figure 1, with the dog on the left, and the cat on the right. As with the example in Figure 1, this association is assumed to take place some time in the preceding discourse. The overt pronouns ,,INDEX and bINDEX are used with the nonagreeing verb LIKE, in a sentence which means, 'He (the cat) likes her (the dog)'. Notice that the reference (although not the gender) for each of the pronouns is unambiguous, given the preceding association of NPs with loci. As is often the case in languages with rich agreement systems, subject and object NPs in ASL can be" 'phonologically' null (nonover0. For example, once NPs have been assigned to loci, inflecting verbs may show agreement with these loci without the NPs having to be overtly renamed

420

DIANE LILLO-MARTIN

? C
DOG-.INDEX bINDEX LIKE

CAT-bINDEX

,,INDEX

He (the cat) likes her (the dog).

Fig. 2. Nonagreeing verbs, overt pronouns, and SVO order.

either as full NPs or overt pronouns. It is also possible in ASL for an overt pronoun or nominal to c o - o c c u r with the verb agreement. This is also true for some other null subject languages (e.g., Spanish), but not for others, such as Irish (McCloskey and Hale, 1984), and C h a m o r r o (Chung, 1984), in which an overt pronoun cannot c o - o c c u r with certain types of verb agreement. It will be argued in this paper that when agreement is present in ASL the effect is in several ways the same as if an overt pronoun were present, in that structure which otherwise would need an overt pronoun are grammatical, and structures in which island violations would have o c c u r r e d are 'saved'. In this respect, ASL is like Irish (which is described in the following section), although it is unlike Irish, as already mentioned, in that A S L overt pronouns may c o - o c c u r with agreement.

3. NULL PRONOMINAL ARGUMENTS OF A G R E E I N G VERBS

3.1. pro in Irish

McCloskey and Hale (1984) argue that Irish sentences in which an argument is overtly signaled only by agreement behave in several ways as if a pronoun corresponding to the argument were actually present. T h e y postulate t h a t there is, in these cases, a syntactically real though nonovert

NULL

ARGUMENTS

IN AMERICAN

SIGN

LANGUAGE

421

argument, (not merely a semantically real one), which they call the 'inflectional argument'. In Irish, for example, a variety of suffixes and enclitics attach to basic pronouns to make other kinds of pronouns: These elements also attach to the inflectional argument. The inflectional argument behaves like an overt resumptive pronoun with respect to the syntax of relative clauses and constituent questions. The inflectional argument can function as the head of a relative clause. It can even be conjoined with a lexicallyspecified NP. All of these facts, and additional facts about government, semantics, and binding requirements, support the existence of a syntactically real nonovert pronoun: The presence of this form is sanctioned by a particular agreeing verb form, and its behavior is like an overt pronoun. McCloskey and Hale therefore suggest that this null element is the empty category pro, because that is the empty category which is predicted to behave exactly like an overt pronoun. It will be argued in this paper that the facts of ASL likewise indicate the presence of an inflectional argument in sentences containing an agreeing verb. I propose that this empty category is pro, sanctioned by verb agreement and identified by an INFLection which is marked [+AGR]. 3.2. pro in A S L In ASL, null subjects and objects ca n occur in tensed finite clauses. Some examples with plain (2-3) and inflecting (4-5) verbs follow.

Plain Verbs:
hn

(2)

aJOHN aFLYb bCALIFORNIA LAST-WEEK. ENJOY SUNBATHE[dur].

John flew to California last week. (He's) enjoying a lot of sunbathing.

(3)A. B. Did you eat my candy? YES, EAT-UP.

Yes, (I) ate (it) up. Inflecting Verbs:

(4)A. B. Did John send Mary the paper? (In which John has been established at a and Mary at b.) YES, oSENDb.

Yes, (he-) sent (it) to (-her).

422 (5)a.
a JOHN

DIANE

LILLO-MARTIN

KNOW-WELL PAPER FINISH oGIVEb. John~ knows (hei-) gave the paper to (-her). JOHN KNOW-WELL PAPER FINISH bGIVEa. John~ knows (she-) gave the paper to (-him).

b.

Note the difference in the direction of movement for the verb GIVE in (4a) (oGIVEb) and (4b) (bGIVEo), which correlates with a difference in meaning, specifically in the determination of subject and object. Both inflecting and plain verbs can have null subjects and objects, though the appearance of null arguments with plain verbs is more limited. In section 4.2 1 will discuss the null arguments of plain verbs, and show that they have a different explanation from the null arguments of inflecting verbs. In this section I will be concerned with the occurrence of null arguments with inflecting verbs, in structures for which null arguments do not occur with plain verbs. I will provide evidence that those sentences with an inflectional argument and no overt pronoun are comparable to sentences with plain verbs and an overt pronoun. 9

3.2.1. pro as a Locus-Assigning Pronoun One way in which the inflectional argument is like an overt pronoun is that the inflectional argument can associate NPs with loci in space. As mentioned above, there are several ways in which to indicate the association of an NP with a point in space. Recall that one such way was to produce the sign, and then point with the index finger to some locus. An additional method, not mentioned above, is to produce the sign (in neutral space) and then to produce an inflecting verb for which the onset location is a locus in space which is not already associated with a different NP. That onset locus is now associated with the nominal, and can participate in further verb agreement and pronominal reference. In sentence (6), for example, the subject verb agreement with point a on P R E A C H indicates that the nominal M O T H E R is associated with point a. Thus, the subject pronoun of the second sentence, which agrees with point a, is understood as referring to MOTHER. This is parallel to the use of an overt index to associate an NP with a spatial locus.

9 R e c a l l that b o t h inflecting and plain verbs can also o c c u r with o v e r t pronouns.

NULL

ARGUMENTS

IN AMERICAN

SIGN

LANGUAGE

423

(6)

B R O T H E R SICK M O T H E R aPREACHb[cont], alNDEX aTELLb CLEAN ROOM. My brother is sick of mother preaching at (-him). Now she's telling (-him) to clean his room.

3.2.2. pro as a Resumptive Pronoun The inflectional argument also patterns like an overt pronoun in topicalization, which is a highly productive process in ASL. When a constituent is topicalized, it is marked by a particular combination of raised eyebrows and slight backward head tilt, and a temporal lengthening of the topicalized sign (Liddell 1977). The topicalized constituent usually appears at the beginning of the sentence. Sentences (7-8) show examples of topicalization. 1
t

(7)

a T H A T aBOOK, ~JOHN aREAD. That boo~, John read (-i~).

t

(8)

aEXERCISE CLASS, 1INDEX HOPE bSISTER S U C C E E D bPERSUADEc c M O T H E R aTAKE-UP. (Padden, 1983) The exercise classy, I hope my sister manages to persuade my mother to take (-i~)

Sentence (7) shows topicalization from a main clause, and sentence (8) shows that topicalization is possible from an embedded clause containing an inflecting verb. However, when a constituent is topicalized from an embedded clause containing a plain verb, a resumptive pronoun is required to fill the gap. Compare sentence (8) above with (9) below, and compare (10) with (11). In (8) and (10), the verb of the clause frOm which the topic has been extracted is an inflecting verb, and the sentences are fine without an overt pronoun. However, in (9) and (tl), the verb is plain, and must have an overt pronominal argument.
t

(9)a. b.

aTHAT aCOOKIE, 1INDEX HOPE bSISTER S U C C E E D bPERSUADE~ ~ M O T H E R E A T alNDEX.

t

*aTHAT aCOOKIE, 1INDEX HOPE bSISTER S U C C E E D bPERSUADE~ , M O T H E R E A T That cookiei, I hope my sister manages to persuade my mother to eat i~.

~o I n t h e s e a n d f u t u r e e x a m p l e s I a m u s i n g t e r m s s u c h as ' e x t r a c t i o n ' a n d ' e x t r a c t i o n s i t e '

without meaning to necessarilyimply a movement analysis.

424 (10)

DIANE t

LILLO-MARTIN br

a T H A T ~MAN, bJOHN SAY cMARY FINISH cGIVE~ BOOK. That man~, John said Mary already gave a book to (-hin~).
t br

(ll)a. b.

a T H A T MAN, aINDEX
t

bJOHN SAY cMARY D O N ' T - K N O W

br

*aTHAT aMAN, bJ O H N SAY cMARY D O N ' T - K N O W That maw, John said Mary doesn't know hin~.

Although an overt resumptive pronoun is not always required under topicalization, in some structures with nonagreeing verbs this pronoun is necessary. However, in the otherwise comparable structures with verb agreement, the overt pronoun is optional. The presence of agreement is sufficient to make an overt pronoun unnecessary. A third category of structures which exhibit parallels between overt pronouns and the inflectional argument are those structures traditionally subsumed under the Island Constraints of Ross (1967), and more recently accounted for by Subjacency ~nd the Empty Category Principle. ASL obeys the Island Constraints. 11 However, as in many languages, overt resumptive pronouns can 'save' sentences which would otherwise be island violations. In ASL, the inflectional argument itself can also serve to 'save' a sentence from being an island violation. Sentence (12) is a violation of the Wh-island Constraint subcase of subjacency, since the topicalized element must cross over more than one bounding node (two S-nodes and an S-node). It furthermore violates the Empty Category Principle, since the trace in the most deeply embedded clause is not properly governed. However, the sentence can be saved when a resumptive pronoun marks the site of extraction, as in (13). Example (14) shows that with agreeing verbs, the inflectional argument can 'save' the sentence.
t

(12)

*[~ [Top aMOTHER,], Is IINDEX D O N ' T - K N O W [g 'WHAT' Is t, LIKE]]]]. *Motherl, I don't know what ~ likes.
t

(13)

aMOTHER, lINDEX D O N ' T - K N O W 'WHAT' INDEX LIKE. Motheri, I don't know what she~ likes.

11 Fischer (1974) shows this for the C o m p l e x NP Constraint, and P a d d e n (1983) for the

Coordinate Structure Constraint.

NULL

ARGUMENTS t

IN

AMERICAN

SIGN

LANGUAGE

425

(14)

aMOTHER, 1INDEX DON'T-KNOW 'WHAT' aSEND1. Motheri, ! don't know what (shei-) sent (-me).

Sentence (15) shows a violation of the Sentential Subject Constraint subcase of subjacency (the topicalized element must cross two S-nodes). Again, (16) shows that a resumptive pronoun can save the sentence, and (17) shows that agreement can also save the sentence.
t br

(15)

*[g [Top oBILLi], [s[s bMARY KNOW t,], NOT^NECES SARYII. *As for Bil~, that Mary knows ~ is not necessary.
t br

(16)

~Bill, bMARY KNOW INDEX, NOT'INECESSARY. As for Bill, that Mary knows him is not necessary.
t br

(17)

aBILL, bMARY bGIVE~ PAPER, NOT^NECESSARY. As for Bill, that Mary gives (-hin~) the paper is not necessary.

ASL obeys the Coordinate Structure Constraint. This can be seen in sentence (18), in which an argument has been topicalized from one conjunct but not the other, and the result is ungrammatical.
t

(18)

*[~ [Top aMOTHER~], [s 1INDEX tHITb SISTER], [s clNDEX TATTLE,, t,]]. *His motheri, I hit my sister and he told 4. (Padden 1983)

ASL does, in addition, allow Across the Board extractions (Ross, 1967; Williams, 1978), as in (19) with noninflecting verbs.
t

(19)

[~ [Top ~THAT ~MOVIE], [s bJOHN LIKE t,], [s calLL DISLIKE t~]]. That movie, John likes ~ (but) Bill dislikes ~.

In sentence (20a) we see that with across-the-board extractions, overt resumptive pronouns can occur at both extraction sites. In (20b) there is verb agreement with null arguments in both conjuncts.
t

(20)a.

. T H A T .WOMAN, bBILL FORGET .INDEX cJOHN KNOW INDEX. That woman,, Bill forgot her~ but John knows her~.
t

(BUT)

b.

. T H A T aWOMAN, bBILL bSLAPo (BUT) cJOHN ,KISSa. That woman,, Bill slapped (-her) but John kissed (-her).

426

DIANE

LILLO-MARTIN

Null arguments with agreement can behave like overt resumptive pronouns in across the board extractions. Sentence (21) shows that a resumptive pronoun can occur in one conjunct with agreement in the other. whq (21) WHO a JOHN LIKE bINDEX (BUT) cMARY cHATEb? Who~ does John like (hin~) but Mary hate (-hirr~)? If the sites governed by verb agreement and containing no overt pronoun were simply deletion sites, then (21) should be bad, as a CSC violation, just as (18) is. However, I suggest that in (21) verb agreement is sanctioning the inflectional argument, which is again functioning in the same way as the double overt pronouns in example (20a), and the double null arguments in (20b). 3.2.3. pro as a Crossover-Evading Resumptive Pronoun One additional area in which parallels between the inflectional argument and overt pronouns can be found are crossover-like sentences. Sentences which display Strong Crossover are ungrammatical in ASL. In (22), for example, an object gap is c-commanded by an A-binder in subject position as well as an ,~-binder in COMP, and the sentence is bad in ASL, as it is in English. (22) whq *WHICI~IBOY a.cINDEX bINDEX EXPECT 1INDEX LOVE ti, WHICH?

T.

* Which boyi does hei expect me to love ~?

Example (23) shows that this structure is fine (again as it is in English), if the pronoun in subject position is not coreferential with the moved element. (23) whq WHICH BOY .cINDEX alNDEX EXPECT tlNDEX LOVE, WHICH.'?

WMch boyi does hej expect me to love ~?

However, ASL can evade a crossover violation by base-generating a pronoun in object position, and moving the Wh-word from subject position, leaving an overt resumptive pronoun in the subject's place. Although a crossover analysis of (24), schematized in (25a), would be theoretically ungrammatical, the sentence would not be ruled out under a

NULL ~_RGUMENTS IN A M E R I C A N SIGN L A N G U A G E

427

subject extraction analysis, schematized in (25b). This sentence is grammatical in ASL. (24) Which boy~ does he~ expect me to love hiram? (i.e., Which boyi expects me to love him~?) (25)a. *i t b. i i i i I i
whq

WHICH BOY ..clNDEX bINDEX EXPECT IINDEX LOVE blNDEX WHICH?

5'

The same evasion strategy can be used with agreeing verbs. Although all three pronouns need to be overt with the nonagreeing verbs in order for the sentence to be grammatical, with the agreeing verbs the object pronoun governed by agreement need not be overt. This is shown in example (26). (26)
whq

WHICH BOY ._clNDEX blNDEX EXPECT 1INDEX 1BEATb, WHICH?

Which boyl does hei expect me to beat up (-him~)? (i.e., Which boyl expects me to beat him~ up?) This is another construction in which the inflectional argument functions comparably to an overt pronoun. In summary, the distribution of overt pronouns in ASL is paralleled by the distribution of nonovert pronouns with agreeing verbs. I argue that this supports an analysis which incorporates a syntactically real nonovert argument, of the pronominal category. This element, pro, is sanctioned by the presence of overt morphological verb agreement. 4. T H E O C C U R R E N C E OF N U L L A R G U M E N T S WITH
NONAGREEING VERBS

In the preceding sections I have established that the inflectional argument in ASL acts like an overt pronoun. In so doing, I included example sentences in which a null argument was not allowed with nonagreeing verbs. However, it is not the case that null arguments are never allowed with nonagreeing verbs (cf. examples (2) and (3) above). The contrast of (2) and (3) with the ungrammatical sentences in the

428

DIANE

LILLO-MARTIN

preceding sections raises the question of exactly when null arguments are allowed with nonagreeing verbs. I will argue here that there is an analysis of null arguments with nonagreeing verbs which accounts for these differences. In example (27) below, taken from a signed narrative, null arguments appear with both agreeing and nonagreeing verbs. (27) ONE DAY, , , D A U G H T E R N O T H I N G CAD-O, D E C I D E W A L K b.cWOODS, bWALKc alNDEX SEEd dFLOWER, PICK-UPa, SEE, ,WATERFALL, cWALKe, FASCINATED,, LOST[d:resultative]. One day, the daughter had nothing to do, so (she) decided to take a walk in the woods. (She) walked around, and saw there some flowers, and picked (-them) up; (she) saw a waterfall, and walked (-near it); and (she) was so fascinated (-by it) that (she) became lost.

In this example, DECIDE, WALK, SEE, PICK-UP, FASCINATE, and LOST all appear with at least one null argument. D E C I D E and LOST are both verbs which do not take verb agreement. SEE, PICKUP, and FASCINATED agree only with their objects (indicated by the subscripts at the end of the verbs). The objects of the two SEE verbs are overt; the objects of PICK-UP and FASCINATED, being governed by agreement with 'the flowers' and 'the waterfall', respectively, are pro. WALK is what Padden calls a 'spatial' verb; it agrees with points representing location rather than points representing NP arguments. Notice that all of the null pronouns (except for the two pros), refer to the same person: the daughter. It is also clear that in this passage, the daughter is the topic of the narrative. This fact will play a role in the analysis suggested. 4.1. Huang" s Account of Chinese Null Arguments Chinese has no verb-subject or verb-object agreement, yet it does allow null arguments. Huang (1984) gives the following account for null arguments in Chinese. I will discuss this account briefly, and then in section 4.2 claim that an analogous explanation can be used for the null arguments that appear with aonagreeing verbs in ASL. Some accounts describe Chinese as a discourse-oriented language, as opposed to a sentence-oriented language like English (for example, Li and Thompson, 1976). There are a clustering of properties in Chinese which led Tsao (quoted in Huang, 1984) to posit a parameter to dis-

NULL

ARGUMENTS

IN AMERICAN

SIGN LANGUAGE

429

tinguish discourse-oriented languages from sentence-oriented languages. Among the properties of a discourse-oriented language is a phenomenon called 'topic chaining', by which the topic of a sentence can be deleted under identity with a topic in a preceding sentence. Huang assumes these languages have "a rule of coindexation, in the discourse g r a m m a r . . . (in the LF' module of grammar following LF), which coindexes an empty topic node with an appropriate preceding topic." Another property of discourse-oriented languages is that they are 'topic prominent', as opposed t o ' s u b j e c t prominent'. Topic prominent languages (at least on the surface) do not seem to have the requirement that all sentences must have subjects. These lan.guages have topiccomment structures which don't seem to be derived from any other kind of underlying structures, and they have no overt pleonastic elements. Furthermore, in discourse-oriented languages there are anaphors which are bound in the discourse rather than at the sentence level (as well as sentence-bound anaph0rs). This range of facts, then, seems to indicate a systematic difference between discourse-oriented and sentence-oriented languages. Huang uses this 'independently motivated' parameter to account for the appearance of null arguments in Chinese. He argues that these empty categories are not pronominals, but variables left by the movement of an empty topic which is then coindexed with an appropriate preceding topic by the discourse rule described above. Huang gives a list of examples (his (65), repeated here as (28)) showing positions in which empty categories might be found, and he explains which of these could be interpreted as null pronouns in Chinese and which null topics. As will be seen, the same distribution is found for ASL. (28)a. b. c. d. e. f. e came. John saw e. e saw e. John said that e saw Bill. John said that Bill saw e. John tried e to come.

Huang claims that the distribution of null arguments in all languages (illustrated in (28)) can be accounted for "as the result of the interaction of a number of independently motivated and generalized principles of U.G. [Universal Grammar]" These principles are: "(a) the principle of

430

DIANE LILLO-MARTIN

recoverability, (b) the assumption that a zero pronoun is a pronoun, (c) the assumption that the agreement-marking A G R on a verb qualifies as a potential 'antecedent' of a zero pronoun, (d) the binding theory of Chomsky (1981), in particular the condition of disjoint reference (D JR) or Condition (B), and (e) the Generalized Control Rule (GCR)." The principle of recoverability "says that every empty pronoun must be identified" (this identification is how the reference of the pronoun is determined). The condition of Disjoint Reference is given in Huang's (59), and repeated here as (29a), and the Generalized Control Rule (Huang's (61)) is repeated here as (29b). (29)a. b.

Disjoint Re[erence (D JR) A pronoun must be free in its governing category. Generalized Control Rule (GCR) Coindex an empty pronominal with the closest nominal element.

By the interaction of these principles and language-particular factors, the availability of each of the null elements (e) in (28) can be determined. For languages such as Chinese, which have the discourse coindexation rule, and for ASL sentences with plain verbs, if these are subject to a similar discourse rule or some other mechanism permitting empty topics, the following pattern emerges. The empty category in each of (28a,b,c), and (28e) can be an empty variable, bound to a zero topic, but not an empty pronoun. It cannot be an empty pronoun in (28a or c) because there is no agreement or other nominal element for the empty pronominal to be coindexed with. It cannot be an empty pronoun in (28b or e) because (i) there is no AGR for it to be coindexed with, and (ii) it cannot be coindexed with the subject (John in (b) or Bill in (e)) because this would violate the D JR. (An empty pronominal in (e) cannot be coindexed with the matrix subject, although this would not violate the D JR, because this is not the 'closest' nominal element.) In (28d), the empty category can be an empty topic, or it can be an empty pronoun coreferentiai to John. Since there is no AGR in the complement clause, the empty element is free to look in the matrix clause for its closest nominal element, and coindexation with the matrix John would not violate the D JR. In (28f) the empty category can only be an empty pronoun coreferential to John. The empty variable reading is ruled out by an independent principle of grammar, the Empty Category Principle.

NULL ARGUMENTS

IN A M E R I C A N

SIGN LANGUAGE

431

The above distribution is valid only for nonagreeing verbs. For agreeing verbs, the A G R will be present and available as an antecedent for an empty category, as stated in assumption (c) above. Huang gives the following pattern for languages which have subject-verb agreement, such as Italian. In languages like Italian, which are not discourse-oriented, none of the empty categories in (28a-f) can be variables. In (28a), however, t h e empty category can be an empty pronoun, coindexed with the agreement element. The same is true for (28d), which is coindexed with the agreement element of the embedded clause. However, (28b, c, and e) are ill-formed in languages like Italian. 12 The agreement element in Italian is coindexed with the subject (to ensure that the subject and the subject agreement share person, number, and gender features); if the empty category in object position were coindexed with the A G R or the subject it would violate the D JR. There is no agreement in the embedded clause for the empty category in (28f) to be coindexed with, but it can be coindexed with either the A G R or the matrix subject John, making it an empty pronominal whose reference is determined by the matrix subject (i.e., PRO). 4.2. A Null Topic Account of A S L Null Arguments of Plain Verbs The above account will work equally well for ASL. ASL has been called discourse-oriented and topic prominent, for much the same reasons as Chinese. 13 It has topic-comment structures, discourse-bound anaphors, and no overt pleonastic elements such as it or there. Examples are given in (30). (Compare (30a,b) to (31a,b), which are from Huang.)
t

(30)a.

MEAT, 1INDEX LIKE LAMB. As for meat, I like lamb. (Padden, 1983)
ynq

b.

A. aJ O H N ~FORCEb bMARY bGOc? Did John force Mary to go? neg B. NO, aSELF aGOc. No, himsel[ went.

12 However, see Rizzi (1986) for discussion of arbitrary pro in object position in Italian. 13 Several authors have claimed that ASL word order can only be described in terms of topic-comment at the sentence level, which is n o t what I am claiming here. (One such author is Ingrain (1978); see Coulter (1979) for a review.) This section, rather i is claiming a topic proimnence structure at the discourse level, following Huang's discussion.

432
c. d.

DIANE LILLO-MARTIN

a J O H N SAY S E L F ,,GOt.

John~ said hei was going (non-emphatic). 14

Y E S T E R D A Y , RAIN.

Yesterday it rained.
(31)a. neichang huo,

Chinese
de
zao.

that
xingkui

fire
xiaofangdui lai

fortunately fire-brigade come COMP early

T h a t fire, fortunately the fire brigade came early. (Huang, 1984 [561) b. A. John-i salam-il ponae-ass-ni?

Korean

John-NOM m a n - A C C s e n d - P A S T - Q
Did John send the man? B. ani, caki-ka cikcap o-ass-ta.

no self-NOM in-person c o m e - P A S T - D E C L
No, self came in person. (Huang, 1984 [57]) Given the criteria H u a n g uses, ASL would qualify as a discourseoriented language, and therefore it should not be surprising for i t to have the discourse coindexation rule which allows null topics to appear, thus accounting for examples like (27). Given this rule, null pronouns should then be able to o c c u r in A S L with nonagreeing verbs in just the same places in which null pronouns o c c u r in Chinese. This seems to be the case. Examples in A S L of (28a-f), with plain verbs, are given in (32a-f), T h e y display the same range of acceptability for empty variables and empty pronominals as discussed above for Chinese (indicated by + markers under the categories Variable and Pronominal). (32)a. b. c. d. e. f. e THINK. J O H N L I K E e. e L I K E e. J O H N SAY e L I K E BILL. J O H N SAY B I L L L I K E e. JOHN TRY e THINK. Variable + + + + + Pronominal + +

14 See footnote 6 above regarding the ASL sign glossed SELF.

NULL A R G U M E N T S IN A M E R I C A N SIGN L A N G U A G E

433

Thus, in every case that null topics are allowed theoretically, null arguments of nonagreeing verbs are allowed in ASL, and when null topics are not allowed theoretically, null arguments of nonagreeing verbs are also not allowed, except in the case of (32f), which allows PRO. This automatically follows if the null arguments of nonagreeing verbs are indeed the result of null topics. In ASL, for agreeing verbs, the analysis given for (28a,d,f) in Italiantype languages will hold. However, ASL differs from Italian in having object agreement as well as subject agreement. If the object agreement can identify (be coindexed with) an empty category in object position just as the subject agreement can identify an empty category in subject position, then the empty elements in object position in (28b,c,e) should be interpretable as valid null pronominals. In ASL, it is true that such null pronominals can appear, and their reference is determined by the object agreement. Thus, for agreeing verbs in ASL, all of (28a-f) are grammatical. ASL sentences exemplifying each of these sentence-types are given in (33a-f). (33)a.
b.

e a C A T C H BALL.
` ' J O H N aHITb e.

c. d. e. f.

e`'HITb e.

`'JOHN SAY e `'HITb bBILL. ,,JOHN SAY bBILL bHITo e . `'JOHN T R Y e `'CATCH BALL.
5. ' G E N U I N E ' N U L L O B J E C T P R O N O U N S

In the text of his paper, Huang claims that none of the languages he discusses has null arguments of the pr6'nominal category in object position - what he calls 'genuine' null object pronouns. Rather, he claims that the only kind of null objects allowed are those derived by the null topic device, i.e., those of the variable category. Huang mentions one language with object agreement, Pashto, and indicates in a footnote that he is not including such languages in his discussion of the positions available for null pronouns. I claim that when one does include these languages, one can find null object arguments of the pronominal type. Huang claims that the only empty category which can appear in (30e) in the object position is a null topic. However, as I suggested above, if a language with object agreement can do for null objects what subject agreement can do for null subjects (.i.e., be rich enough to identify them,
\

434

D I A N E LILLO-MARTIN

without violating the Disjoint Reference Rule or the Generalized Control Rule), then languages like ASL could be accounted for. This richness is found if the language has object agreement identifying a null object pronoun. (In section 6 1 will discuss how this might be done). In this case the empty category could be a pronominal, with the normal range of interpretations possible. Thus it seems clear that there are real null object pronouns, when there is verb-object agreement. As further support for his topic-deletion analysis, Huang provides some data from additional syntactic constructions which show systematic differences between English and Chinese. In each of these constructions, the English data show no differences between subjects and objects. In the Chinese data, however, there are subject-object asymmetries. Huang proposes that these differences are due to the fact that Chinese allows null arguments, whereas English does not. In the rest of this section I will present data from ASL, paralleling that given by Huang, which shows subject-object asymmetries that reinforce his analysis of topic deletion and the identification of null pronouns, as well as the nonavailability of null object pronouns when there is not a rich enough agreement. However, the ASL data will also show that when there is object agreement, the subject-object asymmetries again disappear, but this time in the opposite direction from English. This confirms that the distinction between agreeing verbs and nonagreeing verbs in ASL is significant. Huang cites Strong Crossover examples showing that "in Chinese, strong crossover involving an embedded object E C [empty category] is prohibited as in English, but strong crossover involving an embedded subject E C is permitted, contrary to the cases in ~nglish."'~ sentences In (22) and (26) above, I showed that strong crossover-like sentences involving object ECs are bad in ASL with nonagreeing verbs, although they are grammatical with an agreeing verb. In (34), we see that even with a nonagreeing verb, this kind of sentence is grammatical in ASL when it involves a subject E C (compare (34) to Huang's [73a], reproduced as (35)).

I5 That both cases are ungrammatical in English can be seen in (i) and (ii) below. Sentence (i) shows strong crossover involving an embedded subject empty category, and sentence (i) shows it involving an embedded object empty category.

(i) (ii)

*John,, hei said ti saw Bill. *John, he, said Bill saw t,.

NULL

ARGUMENTS

IN AMERICAN

SIGN LANGUAGE

435

(34)

,JOHN, ,INDEX TELL1 DISLIKE bMARY. John~, hei told (-me)t~ doesn't like Mary. (Johr~ told me he~ doesn't like Mary.) Zhangsani, tai shuo e~ mei kanjian Lisi. Zhangsan he say no see Lisi Zhangsan~, he~ said that [he~] didn't see Lisi.

(35)

Huang claims that the reason the sentences involving empty subjects are grammatical is that the empty categories can be real null pronominals. identified through the Generalized Control Rule by the matrix subject ta. However, the empty objects cannot be real null pronominals. If the empty objects were pronominals, they would need to be identified by some c-commanding nominal. Since there is no agreement, the closest such nominal is the embedded subject. The objects cannot, however, be identified with the subject because this would violate the condition of disjoint reference. Thus the difference boils down to the difference between (30d) and (30e) above. For Chinese, and for ASL nonagreeing verbs, the empty subject in (30d) can be a pronominal, but in (30e), the empty object cannot be a pronominal. Likewise, Huang provides examples which show that extraction of the subject of a sentential subject is grammatical, violating the Sentential Subject Constraint. Although he does not show this, extraction of the object should still be bad, because the closest nominal to the empty object category would be subject, and coindexation of the empty object with the subject by the Generalized Control Rule would again violate the condition of disjoint reference. This range of facts obtains in ASL. In (15) above, I showed that extraction of the object of a nonagreeing verb in an embedded sentential subject was bad. In (36), I show that extraction of the subject, even of a nonagreeing verb, in an embedded sentential subject is OK (compare (36) to Huang's [i] of his footnote 31, reproduced in (37)). 16
t br

(36)

~BILL, KNOW bMARY, NOT^NECESSARY. As for Billi, that ~ knows Mary is not necessary.

16 Huang relegates the SSC data to a footnote in the published version of his paper because of an alternative analysis of SSC violations which relates them to the Condition on Extraction Domain rather than subjacency.

436 (37)

DIANE LILLO-MARTIN Zhangsan~, [[ei kan zhebu dianying] bu heshi]. Zhangsan see this movie not appropriate Z h a n g s a n i , that [hei] sees this movie is not appropriate.

T h e data showing subject-object asymmetries for nonagreeing verbs in ASL supports Huang's analysis for these cases, and the fact that null pronominal a r g u m e n t s can be identified by NPs as well as by agreement. 17 However, the parallel sentences showing no asymmetries for agreeing verbs io ASL, both examples being grammatical, supports my claim that there are two types of null arguments in ASL: those that appear with nonagreeing verbs, and those that appear with verb agreement. By Huang's own kinds of arguments, both kinds of empty category must be relied on to account for all of the data in ASL.

6. W H A T Is THE NULL SUBJECT PARAMETER? Various proposals have been made as to how the Null Subject Parameter should be stated. From the evidence that I have presented here, it is clear that this is not simply one parameter with two discrete settings. T h e r e are at least two types of Null Subjects (Arguments): those present in languages like Italian and Irish, and those present in languages like Chinese. ASL manifests both types, and they pattern in distinct ways within this one language. T h e data from ASL supports the view that one type of null argument arises from topic deletion, and that this possibility is tied to the independently motivated Discourse-Oriented versus Sentence-Oriented parameter. Those languages which are set as discourse-oriented will have the discomse coindexation rule discussed in section 4.1, and the null arguments will fall out as a result of that. 18 What about the second type of null argument? Most proposals regard-

~7 Huang also cites data from the Complex NP Constraint and the Left Branch Condition, showing similar kinds of subject-object asymmetries. However, these constructions are

more complicated in ASL, and more work on them is needed before it would be clear whether they would also support Huang's analysis for topic deletion. Is Recent works on Chinese and Japanese null arguments have challenged various aspects
of Huang's analysis (for example, Hasegawa, 1986; Hoji and Saito, 1986; Li 1985; Xu and Langendoen 1985; Whitman, 1985). Many of these authors have suggested that at least some of the null arguments (including objects) that Huang analyzed as variables are

actually pronominal. The ASL data indicate that some distinction between null arguments
of agreeing and nonagreeing verbs must be maintained.

NULL ARGUMENTS

IN AMERICAN

SIGN LANGUAGE

437

ing these null arguments h a v e b e g u n with the idea that null arguments are allowed only when there is sulticiently rich agreement that the content of the empty category can be recovered. This is known as Taraldsen's Generalization (after Taraldsen, 1978), and is also called the Identification Hypothesis by Chung (1984) and Jaeggli (1982). It is in fact often the case that null arguments often co-occur with rich inflection. However there are both instances of languages with rich agreement which do not allow null arguments (e.g., German, Dutch), and languages which do allow null pronominal arguments even without rich agreement. For example, Chamorro has null objects even without object agreement; Chung (1984) argues that these are null pronouns and not variables from Wh-topic movement. Nevertheless, in ASL, at least, some reference to verb agreement must be made, since the null pronominal arguments crucially depend on the presence of argreement. I will suggest how this might be done in sections 6.2 and 6.3. Another point which most proposals have addressed (directly or indirectly) is whether the empty NP in subject position is subject to the Empty Category Principle (ECP), and hence must be properly governed. 19 Although some accounts have assumed so, others have claimed (including Huang's account above), following ~homsky (1982), that empty pronouns need only to be identified, but not necessarily properly governed. This question is addressed in section 6.1.

6.1. Proper Government vs. Identillability Rizzi (1982) assumed that the empty category in subject position needs to be properly governed. This is because Rizzi claimed that the null element in subject position of null subject languages is an anaphor (trace), bound by the agreement in INFL. In his account, Null Subject Languages have the option of marking their INFL node [+pronominal]. When this option is chosen, the INFL so marked has clitic-like properties. It can be a proper governor of an empty category in subject position, it has a pronominal interpretation, and it must absorb Nominative Case. Therefore, when INFL is so marked, a lexical subject cannot

~9 Empty categories must be properly governed, according to the ECP. A proper governor can be a lexical category or an antecedent that meets certain structural conditions. This is i0 some senses a manifestation of the Identification Hypothesis as a principle of grammar.

438

DIANE LILLO-MARTIN

occur because it would not receive Nominative Case and would thus violate the Case Filter. 2 Jaeggli (1982), on the other hand, proposed that the empty category in subject position of a pro-drop language is PRO, and hence, must necessarily be ungoverned. Chomsky (1982) proposed that this empty category is pro, which, although it can be governed (unlike PRO), need not be properly governed. He contended that null pronouns must be 'recoverable', 'identified', or 'determined', but not necessarily properly governed. Is it true that null arguments need to be properly governed, or i s identification correct? The answer to this question requires first a rigid definition of 'identify'. Chomsky (1982, p. 85) says, "the simplest approach is to understand 'local determination' as 'government by AGR'." He then goes on to describe how this might be implemented using the Case-assigning property of INFL. Huang (1984), as discussed in section 4.1 above, talks about the principle of recoverability, and he defines identification in terms of the Generalized Control Rule (given in (29b) above)i Huang explains that 'the closest nominal element' to a given category C will be a c-commanding NP or A G R separated from C by the fewest number of clause boundaries (hence no distinction is made between subjects and objects within the same clause). In most cases, structures in which null arguments are properly governed (as by an agreement in INFL) are also structures in which null arguments are identified by a c-commanding NP or AGR. However, an identifiability requirement without proper government would also allow some additional sentences which the proper government requirement would exclude. This consequence can be seen by examining the nu!l arguments of nonagreeing ASL verbs. Recall that most of the time, such null arguments will be null topics. However, in (28d), it was proposed that an embedded null subject with a nonagreeing verb in the embedded clause can be a pronominal coreferential to the matrix subject. The analysis given in section 4, from an identifiability standpoint, shows why this interpretation is possible. However, this null pronominal will not be

20 The Case filter, referring to abstract Case (not necessarily overt morphologically realized case), can be formulated thus: *NP where NP is phonetically realized. [-Case] Subject NPs usually receive their Case (Nominative) from INFL, and object NPs usually receive their Case (Objective)from the verb.

NULL

ARGUMENTS

IN AMERICAN

SIGN LANGUAGE

439

properly governed. Thus, under a proper government approach, the structure would be predicted to be ungrammatical. The sentences are, however, good with that interpretation, in both Chinese and ASL. Most of the time, with an identification account such as Huang's, the identifying element of a null pronoun will be the agreement which governs it, just as in the proper government account. However, there is a class of sentences which display null pronouns that are not properly governed. These null pronouns are grammatical, and can be identified by a close c-commanding NP. Therefore, the identifiability account must be preferred to a proper government account. 6.2. The Structure of Agreement If AGR is what permits some INFLs to identify subject null arguments, how can other null arguments be best accounted for? In their discussion of Irish null arguments, McCloskey and Hale (1984) propose the structure given in (38) (their [59]) for Possessor Agreement in NPs. "(38) NP

NP

1st

mo

teach

This structure is not unlike some proposed for verbal object clitics in Romance languages, such as that in Rizzi (1982, p. 134) adapted in (39). (39) VP

/ CL loj

~ V concosco

NPj

[e]

440

DIANE

LILLO-MARTIN

Except for order, these structures both resemble the structure often assumed for agreement with null subjects, given in (40). (40) S (= IP)

NP

INFL

VP

AGR i

TNS

/\
una mela

NP

pro i

mangia

I therefore propose that the structure for object verb agreement be like that in (39), the structure for possessor agreement be like that in (38), etc. Abstracting away from category and order, we have (41). (41) XP

Xa

Xb
In the structure (41), A G R and Y are coindexed. For descriptive purposes, let us call [ X , X P ] 'Xa', and [X,X] 'Xd. I propose that the f-features from A G R percolate up to X~, which then governs, can assign Case and/or a theta-role to, and must agree with, Y. Which of these things will happen depends on the nature of the head, Xb. If it is a verb, X,, does everything that the verb is lexically marked to do, including assigning Objective Case and/or theta-roles. If X,, is INFL, it will assign Nominative Case if Xb is [+tense]. Thus, the null subject parameter is extended to allow many kinds of null arguments, each usually identified by a coindexed nominal category. 6.3. The Place o1:Morphological Agreement in the Null Argument Parameter The Null Argument Parameter captures the spirit of the Identification Hypothesis given in the introduction to section 6 by associating the

NULL

ARGUMENTS

IN AMERICAN

SIGN

LANGUAGE

441

a p p e a r a n c e of Null A r g u m e n t s with an identifying nominal category. H o w e v e r , it is well-known that the association between rich m o r phological a g r e e m e n t and null arguments is in m a n y cases incomplete. T h e task the child faces with respect to the licensing of null arguments is not to notice whether there is p e r s o n - n u m b e r agreement, but to decide whether the various A G R s (in INFL, in V, etc) in his language are 'rich e n o u g h ' to be identifiers. 21 Since 'rich' is ambiguous between syntactically rich enough to allow null arguments and overtly morphologically rich, I will use 'strong' to refer to syntactic agreements which can identify null pronouns. Thus Finnish will have 'strong' agreement, and Estonian will have ' w e a k ' agreement, even though morphologically both are almost equivalent in 'richness' (Milsark, 1985). Th#s in Spanish, subject a g r e e m e n t in I N F L is strong, and therefore null subject pronouns are possible. For Spanish, that setting will be consistent, so that all verbs will be m a r k e d with agreement, and A G R will always be strong, as long as the I N F L is [+tense]. In Spanish, when the distinction between second and third person a g r e e m e n t on the v e r b has been lost by final consonant deletion (as in dialects of C u b a n Spanish), null arguments are still allowed. In languages like Spanish, a lack of overt, unambiguous v e r b a g r e e m e n t in some cases does not affect the possibility for pro drop. H o w e v e r , in ASL, when a v e r b does not take v e r b agreement, it cannot have the kind of null arguments that v e r b a g r e e m e n t sanctions. As I have shown, the A S L g r a m m a r must be able to refer to the presence of overt, distinctive a g r e e m e n t marking. A possible analysis is that morphological a g r e e m e n t is generated in the lexicon and a g r e e m e n t - m a r k e d elements are freely inserted at D-structure. T h e r e will then be filters at PF to eliminate any sentences in which the morphological features of coindexed items do not match. Those forms which have overt morphological a g r e e m e n t will also have a phonetic matrix under A G R . In this way any reference that needed to be made to f-features during the syntax could be made. A S L would then have a restriction t h a t only A G R with phonetic matrices can be strong, although Spanish would have no such restriction. 22 21 It is possible that a markedness hierarchy will come into play here, so that the occurrence of null pronominal objects implies a greater likelihood for the existence of null subjects in a given language, for example. Much more cross-linguistic work would be needed before such a generalization could be recognized and codified. 22 This complication in the licensing of null arguments in ASL might then cause them to be harder to learn for children learning ASL as their native language, than {or children learning a language like Spanish. In fact, deaf children learning ASL do seem to have problems, using null argument structures relatively later than children learning English or

442

DIANE LILLO-MARTIN APPENDIX

Upper case English words stand for signs with approximately the same meaning as the English word. aSIGNb Subscripts from the beginning of the alphabet are used to indicate spatial locations. Nouns are marked with a subscript at the beginning of the gloss to indicate the space at which they are signed. Inflecting verbs are marked with a subscript at the beginning to indicate the onset location, and/or a subscript at the end to indicate the endpoint location. 1SIGN2 Subscripts 1 and 2 are used similarly to the subscripts a, b, to indicate first and second person respectively. o.cSIGN Hyphenated subscripts indicate a plural index, in which the hand moves from point a to point c. bINDEX then represents a point in between a and c, referring to one of the referents picked out by the plural. SIGNj Subscripts from the middle of the alphabet are used to indicate abstract coreference. SIGN A line On tOP of a sign or signs indicates that a specific grammatical facial gesture was used during the sign(s). Thus 't' stands for the topicalization marker; 'whq' stands for the Wh-question marker; 'ynq' stands for the yes-no question marker; 'neg' stands for the negation marker; 'br' stands for a brow raise which is sometimes used as a clause marker; 'hn' represents a head-nod. REFERENCES Bahan, Ben and Laura Petitto: 1980, 'Aspects of Rules for Character Establishment and Reference in ASL Storytelling',unpublished, Salk Institute. Chomsky,Noam: 1981, Lectureson Go~rnment and Binding, Foris, Dordrecht. : 1982, Some Concepts and Consequences of the Theory of Ooverament and Binding, MIT Press, Cambridge. and Howard Lmnik: 1977, 'Filters and Control', Lingui.~ticInquiry 12, 155-184. Chung, Smldra: 1984, 'Identittabilityand Null Objects in Chamorro', in Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society. Coulter, C_n~rey: 1979, American. Sign Language Typology, unpublished Ph.D. dissertation, UCSD. Fischer, Susan: 1974, 'Sign Language and Linguistic Universals', in Proceedings of the F~anco-C~rman Conference on French Transformational Orammar, Athaenium, Berlin.
f

SIGN

Italian (who have been studied by Hyams, 1983). However, I have argued elsewhere (Lillo-Martin, 1986a) that the problems the deaf children manifest have to do with their acquisition of the required morphology, and not with the setting of the Null Argument Parameter per se.

NULL A R G U M E N T S IN A M E R I C A N ~i,ON L A N G U A G E

443

and Bonnie Oough: 1978, 'Verbs in American Sigh L~il~age', Sign /.~nguage Studies 1 8 , 17--48. and Robert Johnson: 1982, 'Nominal Markers in ASL', presented st the #inter LSA meeting, San Diego. Frishberg, Nancy and Bonnie Cough: 1973, 'Morphology in American Sign Language', unpublished, Salk Institute. Hasegawa, Nobuko: 1986, 'On the So-called 'Zero Pronouns in Japanese", The Linguistic Review 4(3). Hoji, Hajime and Mamoro Saito: 1986, 'Null Pronouns in Japanese', presented at UC Irvine. Huang, C.-T. James: 1984, 'On the Distribution and Reference of Empty Pronouns', Linguistic Inquiry 15, 531-574. Hyams, Nina: 1983, The Acquisition of Pararneterized Grammars, unpublished Ph.D. dissertation, City University of New York. Ingrain, Robert: 1978, 'Theme, Rheme, Topic and Comment in the Syntax of American Sign Language', Sign Language Studies 20, 193-218. Jaeggli, Osvaldo: 1982, Topics in Romance Syntax, Foris, Dordrecht. Kegl, Judy: 1976, 'Relational Grammar and American Sign Language', unpublished manuscript, MIT. : 1985, 'Clitics in American Sign Language', unpublished manuscript, Northeastern University. Klima, Edward and Ursula BeUugi: 1979, The Signs of Language, Harvard University Press, Cambridge, Mass. Lacy, Richard: 1973, 'Directional Verb Marking in the American Sign Language', presented at the summer LSA meeting, UCSC. : 1974, 'Putting Some of the Syntax Back into Semantics', presented at the winter LSA meeting, New York. Li, Charles and Sandra Thompson: 1976, 'Subject and Topic: A New Typology of Language', in C. Li (ed.), Sub/ect and Topic, Academic Press, New York, pp. 457-490. Li, Mei-Du: 1985, Reduction and Anaphoric Relations in Chinese, unpublished Ph.D. dissertation, UCSD. Liddell, Scott: 1977, An Investigation into the Syntactic Structure of American Sign Language, Ph.D. dissertation, UCSD. Published as American Sign Language Syntax, Mouton, The Hague, 1980. Lillo-Martin, Diane: 1985, 'Agreement, Inflection, and Cliticization', presented at the winter LSA meeting, Seattle. --: 1986a, 'Effects of the Acquisition of Morphology on Syntactic Parameter Setting', in S. Berman, J.-W. Choe, and J. McDonough (eds.), Proceedings of NELS 16, 305-321. --: 1986b, Parameter Setting: Evidence from Use, Acquisition, and Breakdown in American Sign Language, unpublished Ph.D. dissertation, UCSD. - and Edward Klima: 1986, 'Pointing out Differences: ASL Pronouns and Syntactic Theory', presented at the Conference on Theoretical Issues in Sign Language Research, Rochester. McCloskey, James and Kenneth Hale: 1984, 'The SynUtx of Person-Number Inflection in Modern Irish', Natural Language and Linguistic Theorqll1, 487-533. Meier, Richard: 1982, Icons, Analogues, and Morphemes: The Acquisition of Verb Agreement in American Sign Language, unpublished Ph.D. dissertation, UCSD. ~ : 1986, 'Person Deixis in ASL', presented at the Conference on Theoretical Issues in Sign Language Research, Rochester. Milsark, Gary: 1985, 'Which is Rich7', unpublished, Temple University. Padden, Carol: 1983, Interaction of Morphology and Syntax in American Sign Language, unpublished Ph.D. dissertation, UCSD. Rizzi, Luigi: 1982, Issues in Italian Syntax, Foris, Dordrecht.

444

DIANE LILLO-MARTIN

: 1986, 'Null Objects in Italian and the Theory of pro', Linguistic Inquiry 17, 501-557. Ross, John: 1967, Constraints on Variables in Syntax, unpubfished Ph.D. dissertation, M1T. Sfiepard-Kegl, Judy: 1986, Eocative Relations in American Sign Language Word Formation, Syntax and Discourse, unpublished Ph.D. dissertation, M1T. Taraldsen, Tarald: 1978, 'On NIC, Vacuous Application, and the that-t Filter', unpublished, MIT. Whitman, John: 1985, 'A Unified Account of Zero Pronoun Phenomena', presented at the Japanese Syntax Workshop, UCSD. Wilbur, Rounie: 1979, American Sign Language and Sign Systems, University Park Press, Baltimore. Williams, Edwin: 1978, 'Across-the-Board Rule Application', Linguistic lnqu/ry 9, 31-55. Xu, Liejiong, and D. Terence Langendoen: 1985, 'Topic Structures in Chinese', Language 61, 1-27. Zwieky, Arnold and Geoffrey Pullum: 1983, 'Cliticization vs. Inflection: English n't', Language 59, 502-513. Received 15 July 1985 Revised 6 June 1986 Department of Linguistics U-145 341 Mansfield Road, Room 230 University of Connecticut Storrs, CT 06268 U.S.A.