Caribbean Spanish is characterized by an array of syntactic peculiarities with respect to the position and lexicalization of the subject NP. The frequent use of overt subject pronouns, reduction in inflectional morphology, absence of subject-verb inversion in obligatory context are some features which set the Caribbean dialect apart from other varieties (Barbosa, 1995; D'Introno, 1989; Gonzalez-Escribano, 1991; Henríquez-Ureña, 1940; Jiménez-Sabater, 1984; Kany, 1970; Lipski, 1977; Lipski, 1988; Morales, 1986; Morales, 1988; Suñer, 1986; Toribio, 1993) . It has been proposed that these distinctive features are the result of a single underlying parametric shift, involving INFL. The focus of this paper is to tease apart the different properties of the dialect and explore the viability of a unified account. In this paper I argue that the subjects in infinitival construction is a phenomenon independent of other subject-related phenomena. Previous proposals have considered the character of these infinitives as exceptional, suggesting the possibility that they may be attributable to the transition from a pro-drop to a non pro-drop grammar that seems to characterize Caribbean Spanish. I argue that the scope of the overt subjects of infinitives is less extended than previous accounts suggest, and that the overestimation of the phenomenon was due to a misanalysis of distinctive pronouns as subjects. The analysis in this paper concludes that the Caribbean dialect's difference is limited to properties related to the transition towards a non pro-drop grammar, and optionality of verb movement.

The first description of the Dominican dialect was compiled by Henríquez Ureña, a noted Dominican philologist and pioneer in the study of Hispanic American dialectology. Henríquez Ureña points to several syntactic characteristics of subjects in the Dominican dialect:

a) tendency to place the subject of a gerund or infinitive before the verb, and,

b) overt subjects in impersonal constructions

These two properties are illustrated in examples (1) and (2):

(1) a. Al yo salir, …

at I leave-INFL

'As I was leaving…'

b. Yo jugando, …

I play-INF

'me playing…' (from Henríquez Ureña (1940))

c. Al María llegar, …

at María arrive-INF

'As Maria arrived…'

d. Al niño volver a su casa, …

at child return to his house

'As the child was returning to his house, …'

(from Jiménez Sabater (1984))

(2) a. Ello es fácil llegar.

it is easy arrive-INFL

'It is easy to arrive'.'

b. Ello es constante, si es que pueda ponerse en duda.

it is constant, if is that could put-INF-SELF in doubt

'It is constant, if it could be doubted'.

c. Ello no hay dios …

It is no god

'There is no god …' (from Henríquez Ureña (1940))

d. Ello hay poca naranja.

it is little orange

'There is little orange left.' (from Jiménez Sabater (1984))

Other varieties of Spanish would require verb-subject inversion, and null expletive pronouns. Compare (1a) with (3) and (2a) with (4), their respective counterparts in Standard Spanish:

(3) Al salir yo,

at leave-INF I

'As I was leaving…'

(4) Es fácil llegar

is easy arrive-INF

'It is easy to arrive'.'

Jiménez Sabater (1984) describes additional related characteristics of the syntax of subjects in Dominican Spanish: overuse of subject pronouns and lack of subject inversion in questions. In (5a), the overt subject pronoun lacks the emphatic interpretation that it would have in other dialects. Example (6a), with SV order, is clearly ungrammatical in other dialects. The respective standard counterparts require null subjects and VS order in questions, as shown in (5b) and (6b).

(5) a. Tú no estudias porque tú no quieres

you not study because you not want

'You don't study because you don't want to'

b. No estudias porque no quieres

not study because not want

'You don't study because you don't want' to

(6) a. ¿Cuándo ella llega?

when she arrives

'When does she arrive?'

b. ¿Cuándo llega ella?

when arrives she

'When does she arrive?'

Naturally, these syntactic differences have given rise to some speculation about their origin and explanation. With respect to the uninverted subject of infinitives, Henríquez Ureña notes the existence of similar examples in the medieval play La Celestina:

(5) Tiempo de yo vivir, …

time of I live-INF

'Time for me to live, …'

He suggests that the uninverted subject in Dominican Spanish was 'archaic' in origin: "son restos de la antigua determinación del pronombre" (Henríquez-Ureña, 1940: 230). This proposal is rejected by Jiménez Sabater, who sees pronoun overuse as expressing a general tendency to identify each verb: "una tendencia a personalizar todo verbo" (Sabater, 1984: 164). Jiménez Sabater suggests the need for a unified explanation of the various subject-related phenomena in Dominican Spanish.

The idea that the different subject-related phenomena are manifestations of the same underlying process has guided most accounts of the Dominican dialect, and of the closely related Puerto Rican dialect (Lipski, 1977; Lipski, 1988; Morales, 1986; Morales, 1988; Suñer, 1986; Toribio, 1993) . Morales (1986) explored the possibility that the overuse of subject pronouns and the subject + infinitive construction would be the result of linguistic convergence between English and Spanish in Puerto Rico. The results of her research indicated no association between degree of language contact in the Puerto Rican speakers, and their use of the subject of infinitive construction. For that reason, she rejected the hypothesis that the infinitival construction in Puerto Rican Spanish is a case of transfer: "un caso de estructura anglicada". Instead, she suggests that the syntactic changes operating in Caribbean Spanish are forcing the dialect to have fixed SVO order, and that if language contact plays any role, it is the role of reinforcing the already existing trends. She concludes that " [en] el español del Caribe [ …] se encuentra también un proceso de cambio sintáctico que parece llevarlo a la redundancia pronominal general" (Morales 1988: 86).

Generative analyses of the infinitive subjects in Dominican Spanish look at the erosion in the verbal agreement as the source of a dialect-internal explanation for the observed syntactic changes in subjects of infinitives and non-inversion in Wh-questions. Suñer (1986) considers the problems brought by the subject of infinitive constructions for both the Case theory and Binding theory. The subjects of infinitives are problematic for the Binding Theory because the pronouns in subject position behave as if the infinitival clause was a binding domain, even when these clauses do not fulfill all the requirements of a binding domain. The definition of binding domain requires a SUBJECT, and no accessible SUBJECT seems available for the subject NP of the infinitive clause. The pronoun in (7), which represents schematically the infinitive construction, is free to corefer with a higher NP.

(7) NPi … [prep [ pronouni INFL V …]

[-Tense]

[-Agr]

The other problematic aspect of the dialect is the fact that nominative Case is assigned to the pronoun in the absence of +Agreement. Suñer proposes that the weakening of the Agreement part of INFL allows for nominative to be assigned not only by government from a lexical head, but in purely structural terms in the subject position. In her account, the loss of inversion rules is correlated with a special mechanism for assigning nominative case: "Word order is less flexible and as a consequence nominative Case is not necessarily assigned under government" (Suñer, 1986: 179). A language specific, 'marked' rule, would specify that infinitival clauses embedded under a prepositional phrase can assign nominative 'structurally' to the subject position.

In contrast to Suñer, Lipski (1988) does not consider the construction in Caribbean Spanish 'marked' in the typological or linguistic sense of the term, but 'maximally unmarked', instead. As evidence, he points out that these appear in Child Language and in Spanish-based creoles. He suggests that the wide spread of the phenomena in Caribbean Spanish (henceforth CS) is a consequence of the different parametrization in which [+AGR, -TNS] is not morphologically realized yet active and able to participate in nominative assignment. It is clear that this type of approach could solve both the problems identified by Suñer: strong AGR could count as a SUBJECT for the purpose of computing the binding domain, and it could be active for the purpose of assigning nominative.

This insight (that the lexical subject of infinitive is dependent of infinitival AGR being strong in CS) is developed in a different way in Toribio's dissertation. Toribio argues that the 'differences in word order between CS and Standard Spanish are a reflex of a single underlying change, namely, a change in the nature of INFL' (Toribio, 1993: 23). She argues that a dialect in transition may exhibit properties that belong to both stages in evolution. Some of the properties belong to the grammar of null subject languages, or +pro-drop languages, others to the grammar of non-null-subject or -pro-drop languages.

As this discussion of the literature reveals, the exact nature of the relation between the morphological and the different syntactic changes in the dialect is unclear, and sometimes leads to conflicting assumptions on the part of researchers. One may wonder why evolution towards a non-pro-drop or rigid SVO grammar should result in innovative ways to assign nominative in non-finite clauses. It is not generally a feature of non-pro-drop or SVO languages to allow exceptional licensing of Case in infinitives, as the structure of English attests. Perhaps a more general question may be raised: is correlation of properties evidence of a single phenomenon? A positive answer to this question needs to be empirically justified in each particular case, and attain a certain degree of generality. With respect to CS the specific properties of the dialect have all been attributed to a single underlying change in the nature of INFL. Different proposals about the change in INFL have been suggested: strengthening of AGR (Lipski, 1988) , weakening of AGR (Suñer, 1986) , obligatoriness of subjects (Jiménez-Sabater, 1984) , obligatoriness of SVO (Morales, 1986; Morales, 1988) , resetting of the INFL parameter (Toribio, 1993) . In what follows, the implications of parametrization will be explored independently for each aspect of the dialect. The discussion is organized in three domains: exceptional nominative assignment in infinitival subjects, pro-drop parameter, and the interaction between INFL and verb movement.

As discussed in the previous section, the subject of infinitive construction is problematic for Case theory. The Case filter, which prevents the overt realization of NPs without Case, would presumably rule out examples such as (1), or (8), below:

(8) yo no me acuerdo de [yo ser mala]

I not me remember of [ I be bad]

'I don't remember having being bad' (from Morales (1986))

because the lower subject does not seem to have an available Case assignor. The subject pronoun is nominative, which shows that the preposition de is not the Case assignor. INFL, being negatively specified for Tense and Agreement, is in principle not a possible Case assignor either. What is, then, assigning Case to yo? The prevailing answer is that Spec IP can receive nominative by default. D'Introno (1989) argues that this is the case, and makes default Case dependent on the availability of a governor for the Spec IP position. The various proposals differ on their predictions. D'Introno's and Toribio's approaches limit the distribution of nominative subjects of infinitives to environments where the Spec of IP is governed. Lipski's and Suñer's solution, which do not indicate restrictions on infinitival INFL's ability to assign nominative, predicts nominative subject of infinitives across the board. I argue that both approaches overgenerate. While the observation that there is an abundance of subjects of infinitives in this dialect compared to others is correct, I argue that it needs to be qualified: not everything that seems like a subject is a subject. A review of the data available in all the sources for CS shows that most of the cases are pronominal subjects. Subjects of infinitives have been identified in the following environments: adjunct clauses introduced by a preposition, sentential subjects and infinitive complements of some classes of verbs:

Preposition + Infinitive

(9) Al yo salir, entraba él

At-the I-NOM leave-INF, entered he-NOM

'As I was leaving, he went in'

Complement Clauses

(10) El presidente quiere él hablar con el prisionero

the president wants he to speak to the prisoner

'the presidents wants to speak to the prisoner'

Sentential Subjects

(11) Es absurdo uno trabajar sin descansar

is absurd one work-INF without rest-INF

'It is absurd to work without rest'

(12) Me molesta el yo tener que venir todos los días

me-DAT bothers the I-NOM have that come-INF all the days

'It bothers me to have to come everyday'

(From Pérez-Leroux (1995))

of these, only the examples of the preposition + infinitive construction has been quoted with referential NPs as subjects. I have argued elsewhere (Pérez-Leroux, 1995) that only the preposition + infinitive construction contained 'true' subjects of infinitives. If this is correct, then CS would be just like many other languages which allow nominative cum infinitives (henceforth NCI) in special domains (most commonly when the clause is embedded inside a PP). This is the case in many languages, including West Flemish, Italian, Portuguese, and indeed, Peninsular Spanish as well (Fernández-Lagunilla, 1987; Haegeman, 1985; Piera, 1987; Raposo, 1987; Rizzi, 1982) . The standard analysis given to this exceptional nominative assignment is similar to Reuland's account of the -ing construction in English. A non-finite form of INFL is considered able to govern its subject position if the clause containing it is itself Case marked. This configuration is given in (13):

(13) Prep [IP NP INFL VP ]

\_____

If this is correct, the only difference between CS and other varieties of Spanish will be lack of verb movement in the NCI structure. Regarding Standard Spanish, this analysis implies that it is not verb movement which licenses nominative, but the Case-marked position of the infinitival clause itself. The consequence is that the dialects are not different with respect to where NCI appears, just with respect to the word order inside the NCI clause.

Work with informants giving judgments over about two hundred sentences confirmed that the observation that referential expressions were only allowed in the preposition + infinitive construction in CS. Subjects of infinitives in other environments such as sentential complements, subject clauses, etc., were judged ungrammatical unless the NP was a pronoun which patterned in reference with controlled PRO. Note that there is nothing about Case theory which can explain why an overt NP must be controlled. Case-marked positions can be in principle be filled by any NP type, and Binding Theory is responsible for regulating the interpretation of those NPs. Rather than postulating unusual mechanisms for control, I argue is that those NPs are not the subject of the clause itself, but rather a pronoun position, bound by the true subject of the clause. The true subject is, as with infinitives everywhere, controlled or arbitrary PRO. These pronouns duplicate a subject, and serve as markers for emphasis (Bickerton, 1989; Zribi-Hertz, 1989) . Piera (1986) had warned precisely against confusing an emphatic pronoun in an infinitive clause with the subject of the infinitive. He suggested that in a sentence such as (14), the pronoun may be occupying the same position that the emphatic pronoun occupies in (15). He argues that the correct structure of (14) is as in (16):

(14) Julia quería telefonear ella

Julia wanted telephone she

'Julia herself wanted to telephone'

(15) Julia telefoneó ella

Julia telephoned she

'Julia herself called'

(16) Julia queria [ PRO [ telefonear ella ]]

Note that (14) would have a different word order in Caribbean Spanish, as in (17):

(17) Julia queria [ ella telefonear ]

This need not imply that ella in (17) is in Spec, IP. It is entirely possible that the structure of the sentence is as in (18). If this is correct, the only difference between the dialects is lack of verb raising.

(18) Julia queria [PRO [ella telefonear ] ]

The crucial fact is that in (17), as in (14), the only possible interpretation of the pronoun is coindexed. No other pronoun or NP are possible, and ella cannot refer to any other person but Julia. This indicates that the pronoun is covarying with PRO, and that PRO is the real subject of the clause.

This analysis explains the differences between NCI clauses in CS and other varieties. It shows that there is nothing exceptional about Case assignment in CS infinitives, nor is the phenomenon exceptionally widespread in the dialect. The fact that pronominal pseudo-subjects of infinitives are more frequent in CS follows from the fact that pronouns are more frequent than in other dialects. The apparent exceptionality of the dialect is a conspiracy of two factors: the misanalysis of the emphatic pronoun as a subject, by virtue of the word order difference, and the frequency of the emphatic pronoun construction by virtue of the change towards a non-null subject grammar in that dialect. In other words, it is not the case that NCI is more frequent than in other dialects or that in can appear in contexts where it is unavailable in other dialects. NCI would have the same scope in CS as in other dialects, with different word order. At the same time, the pronominal construction would appear more frequently because pronouns have lost the emphatic value that they have in other dialects.

The correlation between inflectional paradigms and the null subject parameter has been observed across different languages types (Jaeggli & Safir, 1989) . In examples (2) and (5a), we saw how the Dominican dialect has some of the characteristics of non-pro-drop languages: overt expletives, or non referential subjects, and general absence of null referential subjects (referred as 'overuse of subject pronouns' in traditional grammatical descriptions). In addition, due to phonological changes, the dialect has lost the distinction between second and third person singular agreement. It is reasonable to assume, as has been done, that the dialect is undergoing a change from +pro-drop to -pro-drop. But how far has this parametric change evolved, and what are the specific syntactic consequences of the parameter resetting?

Toribio (1993) has argued that CS is undergoing a parametric change in the nature of INFL, and that this change results in differences in word order and in the availability of referential and non-referential null NPs. She proposes that two parameters of INFL are involved in: ± Lexical INFL, and ±Pronominal INFL. non-lexical INFL is able to assign nominative in IP by Spec-head agreement. Non-pronominal INFL has no phi-features of its own, and thus fails to identify a subject, resulting in the impossibility of null expletives and null referential subjects. She suggests that INFL in this dialect has a dual nature, so that constructions can be generated by either setting of the parameter.

Toribio shows that the overt impersonal subjects (first introduced above in example (2)), appear with weather verbs, raising verbs, existential constructions and unaccusatives, listed respectively in (19)

(19) a. Ello ta lloviendo.

it is raining

'It is raining'

b. Ello parece que no hay azúcar.

it seems that no is sugar

'There seems to be no sugar left'. (From Toribio (1993))

c. Ello no hay naranja.

it no is orange

'There is no orange left.' (From Jiménez Sabater (1984))

d. ello llegan guaguas hasta allá.

it arrive buses up-to there

'Buses arrive all the way to that place' (From Toribio (1993))

These examples shows that ello, impersonal, surfaces in places where there is no thematic role assigned to the subject position.

Although the evidence to this point suggests that Caribbean Spanish has become a non-null subject language, the fact remains that null subject pronouns and null expletives coexist with the overt ones. Toribio suggests, following Roberts (1993), that both parametric options may coexist in a particular transitional stage in the evolution of the language, so that what in fact we consider one dialect, Dominican Spanish, is in fact two: one which retains a pro-drop grammar, and one which is a non-pro-drop dialect. In her own words,

"…this dialect, in the process of change, may be said to demonstrate a dual nature, in accordance with the proposals of Roberts (1993), who suggests that in the change from a property X to a property Y, a language may demonstrate both X and Y." (Toribio, 1993: 5)

The dual parametric nature of the dialect can be explored further considering recent approaches to the structure of null subject languages. It has been suggested that in null subject languages with rich agreement, the specifier position of AgrS is always filled by an empty category (pro) and preverbal subjects are the result of left dislocation (Barbosa, 1995) . This analysis is represented in (20):

(20) AGRSP

5

Overt DP AGRSP

5

pro AGRS'

: 5

1 AGRS TP

1 3 @

1 AGRS V …t…

z-------------------m

Barbosa (1995) shows that the Northern Italian dialects Trentino and Fiorentino share properties of null subject languages like Italian and non null subject languages like French are in fact pro-drop languages which show expletive agreement morphologically. I will show below some parallels between Dominican Spanish, and these 'semi-pro-drop' dialects.

Like French, and unlike Italian, these dialects require an obligatory subject clitic. The examples in (22) shows that Dominican Spanish patterns with Trentino and Fiorentino in not allowing null subjects in the absence of agreement,

(21) a. Parli (Italian)

b. *Parle (French)

c. *Parli (Fiorentino, Trentino)

d. Tu parli (Fiorentino)

e. Te parli (Trentino)

'(you) speak'

(22) a. Tu habla

b. *Habla (Dominican Spanish)

'(you) speak'

Some constructions with inverted subjects resemble French expletives. Fiorentino has a preverbal expletive clitic, and Trentino lacks such clitic, but the construction shows neutral agreement:

(23) a. Gl' è venuto qualche putela (Fiorentino)

b. E' vegnú qualche putela (Trentino)

c. Il est venu des filles (French)

'(it) has arrived some girls'

In Dominican Spanish, the attested examples with unaccusatives have third person plural agreement, which is default in the so-called 'impersonal' construction. This is shown in (19d), repeated here as (24).

(24) ello llegan guaguas

it arrive-PL bus

Toribio (1993) indicates that the inverted subject with overt expletives is limited to unaccusatives in Dominican Spanish, which is the case in French, but not in Fiorentino:

(25) a. Gl' ha telefonato delle ragazze

b. *il a telephoné des filles

The parallel between French and the northern Italian dialects breaks down with lexical subjects, because in Trentino the subject clitic is obligatory, as shown in (26a-b), in contrast with French, where it is optional, as (27c-d) shows. Dominican Spanish simply disallows subject doubling, as shown in (27e-f):

(26) a. Mario e parla

b. *Mario parla (Trentino)

c. Mario parle

d. Mario, il parle (French)

e. Mario habla

f. *Mario él habla (Dominican Spanish)

Barbosa concludes that the dialects are pro-drop, with the subject clitic base generated as an agreement head under AGRS, with visible expletive agreement. Her analysis predicts certain differences between pro-drop and non-pro-drop languages in scope interaction with indefinites and relative clause extraposition. In fact, the predictions of her theory may serve as tests for sorting the null subject status of transitional dialects.

In null subject languages, indefinite preverbal subjects obligatorily take scope over a modal, because they are in coreference relation with pro. Thus indefinite preverbal subjects are ungrammatical in null subject languages like Italian and Spanish, even though they are grammatical in non-null subject languages like French or English. This is shown in (27), below. In (28) we see that Dominican Spanish patterns with null subject languages in disallowing the possibility of a preverbal indefinite in these contexts:

(27) a. A letter of recommendation was required.

b. Une lettre de recommendation est requise. (French)

c. Se necesita una carta de recomendación. (Spanish)

d. É necessaria una lettera de recommandazione. (Italian)

f. *Una carta de recomendación es necesaria. (Spanish)

d. *Una lettera di raccomandazione é necessaria. (Italian)

(28) a. *Una carta de recomendación es necesaria

b. Es necesaria una carta de recomendación (Dominican Spanish)

Barbosa notes the following contrast between French and English and Null subject Romance with respect to relative clause extraposition:

(29) a. Un homme est arrivé qui veut te parler. (French)

b. A man has arrived that wants to talk to you.

c. *Un hombre apareció que desea hablar contigo. (Spanish)

d. *Un uomo è arrivato che vuole parlarti. (Italian)

Again, Dominican Spanish patterns with the null subject languages:

(30) *Un hombre apareció que desea hablar contigo. (Dominican Spanish)

'A man appeared that wishes to speak with you'

If the analysis sketched above is correct, we may conclude that the structure of Dominican Spanish is still that of a null subject language, although there is evidence of transitional properties.

5. V-movement

The third aspect of the syntax of CS is the absence of postverbal subjects in environments where other varieties of Spanish would make inversion obligatory. This has been observed of several of the Caribbean dialects (Lipski, 1977) .

(31) Cuándo ella llega?

when arrives she

'When does she arrive?

If we consider the fact that verb position has been changing in the evolution of Spanish, the above observations about Caribbean Spanish would place this aspect of the dialect as part of a long-standing diachronic change in Romance: the transition from a V2 (verb second) language, to a non V2 language. It has been suggested that the distribution of clitics in Old Spanish was evidence of verb second phenomenon in that dialect (Aleman, 1988; Lema & Rivero, to appear; Ramsden, 1963; Runner, 1992; Wanner, 1987) . Old Spanish (OS) clitics appeared postverbally in OS in matrix clause, and preverbally in embedded clauses. The matrix order depended on movement of the verb to Comp, but this movement was blocked by negation or quantification (Aleman, 1988; Ramsden, 1963) . It can easily be shown that such restrictions do not apply in contemporary Spanish, and that Spanish no longer has an embedded/matrix word order difference. Rizzi has proposed, however that Italian (and possibly Spanish) has what he calls 'residual V2' effects with questions. Rizzi's controversial proposal is beyond the scope of this paper, but his argument about the word order difference between English and Italian and Spanish is important. He proposes that the same process that forces movement of the auxiliary in English in (32) is the cause for the obligatoriness of VS order in Italian in (33)

(32) What has Mary said?

(33) a. Che cosa ha detto il direttore?

what thing has said the director?

b. *Che cosa ha il direttore detto?

what thing has the director said

'What did the director say?'

Rizzi (1991) follows Rizzi and Roberts (1989) in suggesting that I-to-C movement destroys the context for nominative assignment in Romance. Inversion blocks nominative assignment in [Spec, AGRP] in French because in French nominative Case cannot be assigned under government. In Italian and Spanish, where postverbal subjects are grammatical, this assignment of nominative under government applies only where Tense has not combined with AGR (Epstein, 1984b) . This stipulation allows Rizzi and Roberts to account for the contrast between the grammaticality of auxiliary raising in infinitives, as in (34a), and the ungrammaticality of (34b).

(34) a. Por haberme Miguel ayudado , …

by have-INF-me Miguel helped

by Miguel having helped me'

b. *¿Qué ha Juan dicho?

what has Juan said

'What did Juan say?'

c. ¿Qué ha dicho Juan?

what Juan has said

'What did Juan say?'

The non-finite clause allows just the auxiliary to move to C, but finite sentences require movement of the whole auxiliary+verb complex. In Rizzi's and Roberts account, this prevents nominative assignment. Given this account, the syntactic difference between Standard Spanish and Caribbean Spanish is one of absence of I-to-C movement. The structure of (35) reflects the difference in word order in infinitives.

(35) CP

2

Spec C'

2

C0 AGRSP

: 2

1 Spec AGRS'

1[+Nom] 2

1 AGRS0 TP

z--m 2

: Spec T'

1 2

1 Tense0 AGRO P

z--m 2

Spec AGRO'

2

AGRO0 VP

2

SPEC V'

2

V NP

In Standard Spanish, infinitives allow the T/Agr elements to raise to C0. This movement is not required in CS. In the analysis of infinitives presented before, movement is not a requirement for NCI, only case marking of the clause. If movement does not happen, and the subject and the verb remain in Agr S and VP, the observed SVO order occurs. The emphatic pseudo subject pronoun in infinitives could be in the pro position in Barbosa's analysis in (20). The lack of verb raising would explain the pronoun+V order in that environment as well.

Toribio (1993) proposes that the connection between V-movement and Nominative assignment is the true source of the difference between CS and SS. The change towards the -pro-drop setting affects the ability of the language to allow nominative assignment under government, and not in Spec IP. Movement of the verb + INFL complex to C0 concurrently becomes unnecessary. This situation resembles the evolution of French. As Roberts (1994) points out, Old French lost clitic climbing, long infinitive movement, aux-to-comp movement and null subjects together. In his view, French went from a null subject, productive V2 language to a non-null subject language without productive V2. In Spanish, the change seems to be taking place on two stages: from a null subject V2 system, to a null subject with residual V2 (at least in infinitives), to a non-null subject language without residual V2. Runner (1992) proposes that the transition from a full V2 system to a residual V2 system in the Spanish Middle Ages depended on a loss of a requirement to have 'full' lexical heads in sentence initial position. He argues that it was not loss of verb movement itself, but of the condition that motivated it. If one accepts Toribio's approach to Caribbean Spanish, the same reasoning can be extended. If the verb moves to ensure licensing of nominative in Standard Spanish and such movement is no longer necessary in CS, it would simply fail to apply.

To conclude, we suggest that Caribbean Spanish represents the third stage on the evolutionary history of Spanish:

Old Spanish Full V2 language

Standard Spanish Residual V2

Dominican Spanish Not V2

6. Conclusion

If the 'exceptional' nature of CS subjects of infinitives is not exceptional, what is the difference between this dialect and others? I have suggested that the difference did not involve exceptional nominative assignment, but two other domains, verb movement, and the null subject parameter, and that these two factors conspired to create a semblance of exceptional nominative subject of infinitives. The results of this analysis of Caribbean Spanish indicate that a unified account of all the subject properties of the language is undesirable. I have shown that the dialect can be considered a 'defective' null subject language, because despite the preference for overt subject pronoun and overt expletives, it preserves some of the other properties of null subject systems. The discussion regarding nominative subjects of infinitives showed that the tendency towards overt referential pronouns resulted in an appearance of an overgeneralization of nominative subjects of infinitives. This proved to be misleading: true nominative in infinitives was shown to be restricted to the prepositional construction, a common case of the nominative cum infinitive phenomena. Last, it was shown that the only evidence of syntactic innovation in the dialect depends on absence of verb movement, and the licensing of overt referential and non-referential subject pronouns. So, while a unified account of the three subject-related phenomena explored here is undesirable, it is plausible that the defective nature of the pro-drop system is eliminating the necessity of verb movement, warranting a unified approach for the changes involving the V2/null subject properties of the language.

Aleman, I. (1988). The position of clitics in Modern Spanish. Paper presented at the Symposium on Spanish Linguistics, University of Illinois, Chicago.

Barbosa, P. (1995). A new look at the null subject parameter. Paper presented at the Meeting of the Linguistic Society of America, New Orleans.

Bickerton, D. (1989). He himself: Anaphor, Pronoun or…? Linguistic Inquiry, 18, 345-348.

D'Introno, F. (1989). Empty and full pronouns in Spanish. Hispanic Linguistics, 3, 27-47.

Epstein, S. (1984b). Quantifier-PRO and the LF Representation of PROARB. Linguistic Inquiry, 15, 499-505.

Fernández-Lagunilla, M. (1987). Los infinitivos con sujetos léxicos en español. In V. Demonte & M. Fernández-Lagunilla (Eds.), Sintaxis de las lenguas románicas Madrid: El arquero.

Gonzalez-Escribano, J. L. (1991). Una teoría de la Oración. Oviedo: Servicio de Publicaciones de la Universidad de Oviedo.

Haegeman, L. (1985). INFL, Comp and nominative Case assignment in Flemish infinitivals. In P. Muysken & H. v. Riemskijk (Eds.), Features and Projections Dordrecht: Foris.

Henríquez-Ureña, P. (1940). El Español en Santo Domingo (Second Edition (1975) ed.). Santo Domingo: Taller.

Jaeggli, O., & Safir, K. (Eds.). (1989). The null subject parameter. Dordrecht: Kluwer.

Jiménez-Sabater, M. (1984). Mas datos sobre el español de la republica dominicana. Santo Domingo: Editora de la Universidad Autonoma de Santo Domingo.

Kany, C. (1970). Sintaxis Hispanoamericana. Madrid: Gredos.

Lema, J., & Rivero, M.-L. (to appear). Inverted conjugations and V-second effects in Romance. In C. Laufer & T. A. Morgan (Eds.), Theoretical analyses in contemporary Romance linguistics Amsterdam: John Benjamins.

Lipski, J. (1977). Preposed subjects in questions. Hispania, 60, 61-67.

Lipski, J. (1988). In search of the Spanish personal infinitive. LSRL 18.

Morales, A. (1986). Gramáticas en contacto: análisis sintácticos sobre el español de Puerto Rico. Madrid: Biblioteca de autores de Puerto Rico.

Morales, A. (1988). Infinitivo con sujeto expreso en el español de Puerto Rico. In R. Hammond & M. Resnick (Eds.), Studies in Caribbean Spanish Dialectology Washington: Georgetown University Press.

Pérez-Leroux (1995). Case assignment in infinitives and the structure of Caribbean Spanish. Penn State University.

Piera, C. (1987). Sobre la estructura de las cláusulas de infinitivo. In V. Demonte & M. Fernández-Lagunilla (Eds.), Sintaxis de las lenguas románicas Madrid: El arquero.

Ramsden, H. (1963). Weak pronoun position in the Early Romance languages. Manchester: U. of Manchester Press.

Raposo, E. (1987). Case theory and INFL to Comp: the inflected infinitive in European Portuguese. Linguistic Inquiry, 18, 85-109.

Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris.

Roberts, I. (1993). Verbs and diachronic syntax. Dordrecht: Kluwer.

Runner, J. (1992). On the decline of V2 in Spanish. U. of Massachusetts, Amherst.

Suñer, M. (1986). Lexical Subjects of Infinitivals in Caribbean Spanish. In O. Jaeggli & C. Silva-Cortalán (Eds.), Studies in Romance Linguistics Dordrecht: Foris.

Toribio, A. J. (1993). Parametric variation in the licensing of nominals. Doctoral dissertation, Cornell University.

Wanner, D. (1987). The development of Romance clitic pronouns. Berlin: Mouton de Gruyter.

Zribi-Hertz, A. (1989). Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse. Language, 65, 695-727.

Notes