Cornell University

Anyone who is asked to write an essay on the state of the art of Spanish syntax and semantics will face the problem of what limits to place on such a project²¹. Without wanting to diminish the relevance of other valuable theoretical approaches, I have chosen arbitrarily to focus in some detail on the theoretical approach that in my judgment has generated the greatest interest among linguists working in modern Spanish syntax. I am referring to the framework that has come to be known as the principles-and-parameters approach, first introduced by Chomsky in his Lectures on Government and Binding (1981; see also Chomsky 1982, 1986a, and 1986b). Further narrowing the scope, I have concentrated primarily on the work of a very active group of linguists here and abroad who seem to be preoccupied with the same questions within this paradigm²². Although a natural outgrowth of earlier generative transformational approaches, the G(overnment and) B(inding) model, as it is also known, represents a radical shift in the nature of linguistic inquiry: it moves away from the study of specific syntactic rules (such as passive, question formation, relativization, etc.) and concentrates on the general principles, constraints, and parameters which might indicate how language acquisition takes place, that is, on what constitutes the species specific «knowledge of language».

One puzzle which confronts us, however, is how to explain the relatively short time (roughly 5-9 years) a child needs to acquire and reach the ability level of an adult speaker, while being exposed to only fragmentary data. This bespeaks a very efficient and uniform language acquisition system, which at the same time must allow for creativity and the great variety of human languages which are known to exist. The hypothesis is that language (understood in a general sense, not merely as Spanish, English or Chinese) is one of the systems of knowledge, a cognitive system, which is in large part innate and as such, part of our biological endowment. This conception of language motivates the quest for the principles of universal grammar (UG) which form part of our language faculty.

One of the principles of UG must be that rules of language do not rely solely on linearity but rather on hierarchy of structure, that is, rules must be structure dependent. As an example, consider how Spanish forms interrogative sentences: from the declarative (1) we can ask question (2) by using the appropriate question-phrase and by fronting the verb.

(1) Los libros están sobre la mesa.

(2) ¿Dónde están los libros?

Next, examine what happens in a more complex situation such as in (3).

(3) Los libros que trajo tu sobrina están sobre la mesa.

The way to ask a question parallel to (2) is obviously not (4) but (5)²³.

(4) *¿Dónde trajo los libros que tu sobrina están?

(5) ¿Dónde están los libros que trajo tu sobrina?

From these examples we surmise that we cannot merely front the first verb we find but

––––––––   833   ––––––––

rather than we must pay attention to structure, and that we (unconsciously) «know» that there is something deeper and more complex than word order which plays a role in question formation. Since the child is never overtly taught how to ask questions, and furthermore, she never makes mistakes of the type in (4), the conclusion can only be that the structure-dependency of rules is part of the innate language faculty²⁴.

On the other hand, the child is anchored to a speech community which partly determines her linguistic experiences, and produces dialectal and language variation. To illustrate, a child learning Spanish as her mother tongue comes to know that this language has object clitics (6a) whose positioning is not the same as that of noun phrases with the same function (6b). Contrast this distribution to the English counterparts.

(6)a Susana lo visitó

Susana visited him.

b Susana visitó a su padre.

Susana visited her father.

This idiosyncratic property of Spanish is learned through experience, though perhaps constrained by a parameter (e. g. the clitization parameter). Parameters are thus language particular options which interact with the principles of UG to yield the grammars of specific languages.

The mentalistic principles-and-parameters model is a highly modular system in which different theories (X-bar theory, Binding theory, Theta theory Case theory, Government theory, Bounding theory, etc.; see below for discussion of some of them) interact so that only permissible sequences arise. Ill-formed sequences are discarded in a principled way, i. e., as a consequence of violating a principle. In essence, the aim is to go beyond mere description in search of explanations, to show how different phenomena can be derived through the interaction of the different theories in conjunction with language specific parameters set in the appropriate way. Hence, the uppermost question that must be answered in every instance is WHY -a challenging task whose ultimate reward is the better understanding of this unique property essential to the nature of human beings: language.

My aim in what follows is twofold: first, I plan to discuss two examples of how the principles-and-parameters model works (in as non-technical a way as possible) in order to give the reader an understanding of how it can contribute to a better knowledge of Spanish. This necessarily implies some simplification, but my objective is to make this essay as accessible and informative to non-linguists as I can. Second, I shall highlight some of the research on Spanish syntax and semantics carried out in this decade. But before beginning, I must preface the discussion with a caveat. Within the model I am presupposing, it is difficult to separate syntax from semantics because sentence semantics, that is, the interpretation of the sentence based exclusively on grammatical factors, is part of the syntax. Furthermore, what this model leaves aside for the most part is functional/pragmatic considerations that are due either to the discourse situation and/or extralinguistic information (such as knowledge of the world)²⁵. The reasoning for this is the assumption that sentence grammar and its structure are prior to the utilitarian use of language since principles of grammar make reference to configuration but not to extra-sentential meaning.

From the above introduction, it should be obvious that it would be quite difficult to discover principles of UG by concentrating solely on Spanish. Not even parameters could be speculated on, since for them too, one needs to look at other languages to see if they behave similarly or differently from Spanish.

If the only language one knows or works on is Spanish, one can (wrongly) deduce that all languages permit sentences of the type in (7), with what traditional grammarians call «sujeto tácito o sobreentendido»

(7)a Compramos un montón de cosas.

b Salgo en 5 minutos.

Nevertheless, we all know that the corresponding English sentences require an overt subject in order to be grammatical.

(8)a *(We) bought a lot of things.

b *(I)'m leaving in 5 minutes.

What distinguishes Spanish (and other major Romance languages with the exception of French) from English in this instance is what has been labeled the Null Subject Parameter (NSP); Spanish is positively specified for it while English adheres to the negative setting. The NSP is (imperfectly) correlated with richness of verbal agreement. The difference between

––––––––   834   ––––––––

(7) and (8) is not the only one captured by this parameter; there are others. For example, Spanish, as opposed to English, does not have expletive subjects; thus, no overt subject ever materializes in the examples in (9).

(9)a Hace sol.

It is sunny.

b Hay un ciervo en el jardín.

There is a deer in the garden.

Moreover, a sentence such as (10) is ambiguous.

(10) Dijeron que el presidente estaba loco.

They said that the president was crazy.

In one interpretation, it means that at least two identifiable persons said something about the president. The second reading is the one in which somebody (maybe just one non-identifiable person) said the same thing. The latter is what is known as the arb(itrary) interpretation because the subject ranges over the set of possible referents without explicitly identifying one. The curious fact is that when one says (11) instead of (10), the sentence is nonambiguous: the second interpretation is no longer available (Súñer 1983, Jaeggli 1986, Hernanz 1988, Súñer 1988b for more on arb subjects).

(11) Ellos dijeron que el presidente estaba loco.

Note that in English the overt subject pronoun in (10) is ambiguous in the same way that the null subject is in Spanish.

A further consequence of the NSP is that the subject of an embedded clause in Spanish can be questioned as in (12a); compare this to the ungrammaticality of the corresponding English sentence in (12b).

(12)a ¿Quién dijo Briana que t traería el vino?

b *Who did Briana say that t would bring the wine?

c Who did Briana say t would bring the wine?

The t (for trace) in (12) marks the place from which the question word originates (more about this later). To my knowledge, Perlmutter 1971 was the first to establish that NS languages do not obey the «that-trace» effect (cf. the italicized segments in (12a-b), where the sequence that-t produces ungrammaticality in English but not in Spanish).

Another phenomenon which may be correlated with the NSP is what Rizzi (1982) calls «free inversion», that is, postposing of the subject in the absence of any trigger.

(13)a Caminaba Andrea por la avenida cuando...

b Descubrieron unos niños el cuerpo de la víctima.

c Murió tu enemigo.

English does not allow the above²⁶.

The main point is that parameters, if properly formulated, tend to intertwine a group of properties which show correlations. Therefore, learning one of the properties encompassed by the parameter should be enough for the child to deduce the rest, thus greatly facilitating her acquisition task. This is one of the assumptions with which we are working.

Discussion of the NSP provides an introduction to what I consider to be one of the most interesting and stimulating aspects of linguistic research: the investigation of empty categories (ec's). Since ec's are devoid of lexical content, the language learner cannot derive their properties and distribution from overt expressions. Thus the assumption is that ec's provide a «hatch» through which to reach for the principles of grammar, of UG. Jaeggli 1982 constitutes an early attempt to capture the identifying traits of two empty elements (PRO and trace) within a modular and comparative approach to syntax.

2. 2. 1. pro. As already mentioned, traditional Spanish grammarians intuitively knew that the example in (14) has an «implicit» subject, a null category.

(14) ec trabaja como una loca.

One of the problems with ec's is to limit their proliferation. It would be undesirable to posit them at will without a system of checks and balances. The null hypothesis is that «missing» elements parallel the behavior of overt elements (barring interference from other aspects of the system of grammar). One powerful and restrictive principle of UG is the Projection Principle. It essentially states that lexical properties of lexical items must be preserved at every level of representation. This means that when a child learns the vocabulary of language X, she also learns the properties of each of the lexical items. For example, consider the verb invitar; we know that it is transitive, i. e., it mandatorily requires a direct object (15a) because (15b) is ungrammatical.

(15)a Mara invitó a José.

b *Mara invitó.

We also know that since Spanish has object clitics²⁷, it is possible to say (16).

(16) Mara lo invitó.

The question that arises is whether lo «him»

––––––––   835   ––––––––

in (6) fulfills the Projection Principle, i. e., whether the clitic maintains the transitivity of invitar. The consensus nowadays is that it does not. Take into account that all dialects use (17a), and many also accept (17b)²⁸.

(17) a Mara lo invitó a él.

b Mara lo invitó a José.

Therefore, the null hypothesis suggests that sentence (16) must actually have an ec as in (18), where the clitic and the ec form a discontinuous element (Borer 1983), a fact we signal by means of superscripts.

(18) Mara loⁱ invitó ecⁱ

But given the reasoning above, why shouldn't adding an ec make the ungrammatical (15b) grammatical? The answer is straightforward. Since ec's are silent, the language learner needs some indication of their existence. In (14), verbal agreement tells us that the nonlexical subject is 3rd p. sg. Moreover, because the adjective is loca in the feminine sg. form, we also know that the subject reference is a female. In (16), the object clitic signals that the direct. object (DO) is 3rd p. sg. and masc. Hence, object clitics help recover the content of the DO ec in the same way that verb agreement helps identify the content of the empty subject. When one pursues this parallelism further, the conclusion has to be that the clitics are agreement morphemes (cf Súñer 1988a), which promotes the assumption that one way to identify ec's is by means of agreement. There are believed to be four types of ec's; the one in (14) and (18) is represented as (little) pro. It has the features (+pronominal -anaphoric), and it may appear in those (governed) environments in which overt pronouns materialize unless something else in the grammar interferes. To illustrate, pronouns can be the objects of prepositions (19a); however, an ec is not allowed in this position (19b).

(19)a Soñé con ella repetidas veces.

b *Soñé con ec repetidas veces.

The null hypothesis predicts the possibility of having the ec. So why is (19b) ungrammatical? The reason is simple: Spanish cannot recover the content of a null category as object of a preposition because the language has no way to identify that content. However, the correctness of the reasoning is validated when one considers languages which do have a way to identify an ec inside a prepositional phrase (PP). One such language is Irish (and other Celtic languages), which, predictably, permits ec's as objects of prepositions because prepositions can inflect for person and number (20) (from McCloskey and Hale 1984).

In sum, pro is the governed pronominal ec whose content is recovered through agreement. Whenever identification fails, pro is barred (cf [15b] and [19b]). Both the Projection Principle and the identification requirement on pro contribute to barring the proliferation of null categories. Much more can be said about little pro²⁹; nevertheless, for reasons of space, I now explore a different type of ec.

2. 2. 2. Trace. Consider next a passive sentence like (21).

(21) El libreto fue adaptado (por Julián)

Knowledge of the properties of adaptar tells us that it is a transitive verb, which means that the Projection Principle requires a DO at all levels of representation. But where is the DO in (21)? There does not seem to be one. However, we understand that what was adapted was el libreto, so that at one level of representation this phrase should be the DO of the past participle adaptado. Furthermore, even when the Agent who carries out the action is not overtly expressed, it is still implicitly under stood. Compare (21) with the PP removed to (22).

(22) La pared se desmoronó.

There is no implied agent in (22), so much so that contrary to what happens with passive sentences, (22) is incompatible with agential por-phrases (23a), with «agent-oriented» adverbial expressions (23b), and with purposive clauses (23c)³⁰.

(23) La pared se desmoronó

a) *por los obreros.

b) *voluntariamente / a propósito.

c) *para hacer más espacio.

Let us return to (21). Its (much simplified) underlying structure is (24).

(24) ec[_vpfue[_Pass-do[_VPadapta-[_NPel libreto (por J.)...

The principles of two subtheories, Theta theory and Case theory, conspire in the derivation of passives. Case theory requires that (abstract) Case be assigned to nominal categories, otherwise the sentence is ill-formed³¹. Theta theory mandates that each argument

––––––––   836   ––––––––

receive only one theta-role (=semantic role), and that each theta-role be uniquely assigned to an argument³².

The two crucial properties of passives can now be stated (Chomsky 1981, Jaeggli 1986a, among others). First, the object argument of the participle is unable to receive objective Case from it because this Case is assigned instead to the passive morpheme (-do in Spanish, -en in English) which is nominal in character (Baker 1988)³³. Second, the subject position (the ec in [24]) does not get a theta-role because the passive morpheme is the one which receives the agent theta-role³⁴. These two traits cause the movement of the object argument to subject position; hence, it is assigned nominative Case and it agrees in person and number with the V(erb). Note that the theta-role of the derived subject is that of Theme, i. e., the one given to this argument in its original position by the V. Although in English the subject must appear in preverbal position in order to receive nominative, in Spanish, because of its status as a NS language, the subject is not required to front; see (25).

(25) Fue adaptado el libreto (por Julián).

What remains is to examine the characteristics of the fronting rule that yields the word order in (21), and of the ec left behind in object position, cf. (26).

(26) el libreto_i fue adaptado ec_i

This fronting is called NP-movement, one instance of the general rule of Move-a. This rule establishes a relation between an antecedent and the ec (indicated by the subscripts in [26]). NP-movement can only operate when the target position has no theta-role; otherwise the moved element would end up with two theta-roles in violation of Theta theory.

Recall that the ec is necessary because of the Projection Principle. What characteristics does this ec have? (It cannot be pro because no identification or Case obtains in [26]). We know that the ec in (26) has no independent theta-role; hence, it cannot be referential; furthermore, its antecedent is in argument position. These are the characteristics of a trace, t, the [-pronominal+anaphoric] ec. The theory of Government further requires that traces be «properly governed», a requirement which is fulfilled in (26) by the past participle³⁵.

This discussion of the passive construction serves to illustrate why there is no passive rule as such. Passives are derived through the interaction of different modules of the theory (the Projection Principle, Case theory Theta theory) together with properties of lexical items, and the general rule of Move-a.

For the sake of completeness, I will briefly mention the other two types of ec's.

2. 2. 3. PRO. One of them is the pronominal anaphor PRO which arose from the study of embedded infinitivals and other non-finite clauses; see (27).

(27) a Pilar quiere [PRO irt a la China]

b Pilar loⁱ mandó proⁱ [PRO ir a la China]

c Es peligroso [PRO it a la China]

The theory of Control concerns itself with the choice of antecedent for PRO. In (27a), PRO is understood as coreference with Pilar, so PRO is controlled by the matrix subject. In (27b), PRO is controlled by the DO. In (27c), PRO does not have an antecedent; thus it is said to be arbitrary in reference. The hypothesis is that PRO is assumed to be restricted to subject position of non-finite clauses, and therefore, it is not Case-marked. It carries the features [+pronominal +anaphoric].

2. 2. 4. Variable. The final type of ec is [-pronominal -anaphoric]. Variables act as place holders for referential expressions when quantifier-like expressions move by Move-a. For example, in (28a) the question-phrase fronts to the left-periphery of the sentence leaving behind this ec. The interpretation of sentence (28a), represented in (28b), makes clear that (28a) asks for the precise referent of x.

(28)a ¿A quién_i contrató la jefa ec_i?

b for which x_i, x_i a person, the boss hired x_i

Variables, like NP-traces, need to be properly governed, but unlike NP-traces, they must be Case-marked for their licensing³⁶. The first requisite is violated in (29) because the antecedent of the variable is «too far away», which causes the relationship between the two relevant elements to be interrupted³⁷.

(29) *¿Cuántas páginas_i dijo Mara que el libro que había escrito Juan tenía ec_i?

The Case-marking requirement is evidenced from the ungrammaticality that results when one attempts to extract from a non Case-marked position such as the subject of an infinitival (30a), as opposed to the well-formedness that obtains when extraction is from a Case position, such as that of object (30b).

(30)a *¿quién_i explicó Mara cómo ec_i llegar?³⁸

b ¿qué_i explicó Mara cómo PRO arreglar ec_i?

To summarize, the properties of four types of ec's have been explored. Different mechanisms (agreement, proper government, Case, Theta theory, etc.) interact to ensure that ec's are appropriately licensed, so that the language learner can deduce their presence from the interaction of general principles of the grammar. Assuming the correctness of the typology discussed, for every type of ec there should be a corresponding overt category. With one predictable exception, this is true, see (31).

(31)	features	ec	overt categories
a	[+pron -anaph]	pro	ella, nosotros, etc.
b	[-pron +anaph]	NP-t	uno al otro, reflexives39
c	[+pron +anaph]	PRO	-
d	[-pron -anaph]	vble	Mara, la gente, i. e., any referring expression

The exception is the absence of an overt counterpart for PRO. This gap is expected however, because the features in (31c) demand an ungoverned category; but an overt category requires Case, which is assigned under government (cf. ftn 8), a fact which leads to a contradiction. Therefore, independent factors anticipate the gap in the system.

There are some topics that, because of their intrinsic characteristics, continue to attract the attention of linguists. Such topics serve to test assumptions and principles of the grammar, cause the re-evaluation of postulates, and in general, push for theoretical developments. Some of these are highlighted below.

These ubiquitous morphemes show the interrelation of the syntax and the morphology. They raise questions as to their placement, their binding properties, their relationship to ec's (cf 2. 2), and their status.

The hypothesis that inflectional morphology should be part of the syntax proper (Pollock 1988, Chomsky 1988b) together with the assumption that clitics are agreement affixes (Súñer 1988a), implies that clitics should be treated as heads of their own phrases.

Their referential properties merit scrutiny because they are non-homogenous. Compare (32) to (33).

(32) Briana laⁱ acusó ecⁱ

(33) José seⁱ afeitó ecⁱ

In (32) the chain formed by la and the ec has to be obligatorily disjointed in reference to the Subject Briana, while in (33) the exact opposite is true; se and its coindexed ec must corefer with José. (This is the traditional difference between non-reflexive and reflexive pronouns). This distribution is captured by the following principles of Binding Theory (BT)⁴⁰.

(34) Principle A: An anaphor must be bound in the domain of a subject.

Principle B: A (non-anaphoric) pronominal must be free in the domain of a subject.

Bound here must be understood as coindexed with an antecedent, free means non-coindexed.

The implication of BT for (32) and (33) is that in (32) the ec is pronominal(=pro as discussed in 2. 2. 1), but the one in (33) is anaphoric since it seeks the subject for antecedenthood; hence, it is akin to (the nonreferential) NP-t [cf. 2. 2. 2]. Examination of more complex examples, instances of what is descriptively known as clitic climbing or clitic promotion [Luján 1980, Súñer 1980] (because the clitic which «belongs» with the lower verb appears with the upper one) uncovers further differences in behavior.

(35)a El barbero lo quería afeitar ec

b El barbero se quería afeitar ec

(36) a El barbero lo hizo afeitar ec a Juan.

b *El barbero se hizo afeitar ec a Juan.⁴¹

In (35) both lo and se may appear with the matrix verb, but in (36) only lo can. The question is why. The answer is given by the principles of BT in (34). Lo can «climb» in both (35a) and (36a) because its chain is free within the domain of the embedded subject: big PRO in (35a), and Juan in (36a), that is, the pronominal ec has a referent distinct from these subjects. On the other hand, the anaphoric chain in (35b) and (36b) must be bound in the same domain in which the pronominal chain is free. This obtains in (35b) where PRO has the same referent as the matrix subject; thus, se can attach to the matrix verb because of the transitivity of referents. However, in (36b) the embedded and the matrix subjects are distinct, so the link between the se and its ec is disrupted, with the result that the sentence is ungrammatical with the intended reading. What emanates from these disparate patterns and from the speakers' judgments is that ec's, though non-overt, are real for mental computation; that is, native speakers «know» of their presence and their features⁴².

The choice as to whether

––––––––   838   ––––––––

to use indicative or subjunctive in a given environment is of interest in and of itself (cf. Bell 1980 for a general overview, and Blake 1985 for research with pedagogical implications). Moreover, embedded subjunctives display certain idiosyncratic traits⁴³ such as [+/-strict] sequencing of tenses depending on semantic classes of verbs (see Súñer and Padilla 1987, and Súñer 1988c for detailed analysis), transparency concerning matrix negation [37] (Bosque 1980, Súñer and Padilla 1984), and unusual conditions on the referential properties of embedded finite clause subjects under certain verbs [40] (cf. Padilla-Rivera 1985, Súñer 1986, Kempchinsky 1986 and 1987).

(37)a No creo que llame nadie.

b Dudo que llamara nadie.

In (37) the matrix negation no, and the negative verb dudar can extend their scope down into the embedded clause; this allows the negative nadie to stand postverbally without being preceded by no, something that is not possible in matrix clauses (38), or in embedded clauses with indicative mood (39).

(38) *llamó nadie

(39) *dicen que llamó nadie

With respect to the referent of the embedded null subject, it seems to require obligatory disjoint reference from the matrix subject, cf. (40).

(40)a Lola_i quería que pro_*i/k comprara pan.

b José_i lo^k animó/persuadió a que pro_k apagara la luz.

This fact goes against Principle B of BT which predicts that the embedded subject should be free to corefer with the matrix one. Explanations for (40) which have resorted to extending the domain of Principle B of BT, have failed to provide a uniform account which also encompasses examples like those in (41) where subject to subject coreference is possible (though not obligatory), as predicted by Principle B.

(41) a Lola_i quería que Mara y ella_i/k fueran al cine juntas.

b José_i animó a Lía_m a que ella_m/k, y su_m/k/i mamá se hablaran por teléfono.

On the basis of the above data, Súñer 1986b concludes that lexical/semantic factors play the crucial role in the way that the networks of coreference of influence verbs operate. Confirmation for this is found in Padilla-Rivera's 1985 carefully mounted experimental study. His objective was to test the Binding Principles in the language development of a group of eighty 3-9 year old Puerto Rican children. He showed that children initially consider the embedded clause to be the binding domain for the subject regardless of mood. At latter stages of development, they may modify their initial hypothesis because of lexical/semantic properties of verbs. Therefore, data from language acquisition not only confirms the adequacy of BT (part of UG) but also tells us that language specific conditions may interact to over ride UG principles in certain (lexically determined) environments⁴⁴.

The movement of quantifier-like elements to non-argument positions (to the left-periphery of the sentence in Spanish) is known as Wh-movement.

Ever since the pre-GB article «On WH-movement» by Chomsky (1977), this has been a hot topic for linguistic research. The hypothesis is that processes which at first sight appear to be unrelated in fact have Wh-movement (a subtype of the general rule of Move-a) in common. Among the diagnostics for this rule are the gap (=ec) left behind by the preposing of the Wh-element to a position outside the sentence proper, to Comp, and the fact that this movement obeys Bounding theory (i. e., Subjacency). Some of the constructions thought to be affected by this rule are direct and indirect questions, restrictive and nonrestrictive relative clauses, comparatives, topicalizations, cleft sentences, and some focalization constructions. In what follows, I mention some of the major works on these subjects.

In Spanish, the syntax of questions has been studied by Plann 1982, Torrego 1983 and 1984, Contreras 1984, and Súñer 1989 (for the selection of interrogative words, see Bosque 1984a). Not all of these works focus exactly on the same areas: Plann and Súñer focus on the peculiarities of indirect questions to show that Spanish (but not English) needs multiple positions on the left-periphery of embedded clauses in the environment of verbs of communication (42).

(42) Sara preguntó/susurró [que] [a quién] habría invitado la niña.

Torrego examines the effects of V-fronting (which causes subject verb inversion) and successive cyclicity in some Spanish dialects (43),

––––––––   839   ––––––––

and finds that these are characteristics of some focalized constructions as well (44)⁴⁵.

(43)a ¿Qué pensaba Juan que le había dicho Pedro que había publicado la revista?

b *¿Qué pensaba Juan que Pedro le había dicho que la revista había publicado?

(44)a Un viaje a las Canarias dice Juan que quería la gente que hiciera Antonio este verano.

b Un viaje a las Canarias dice Juan que la gente quería que Antonio hiciera este verano.

And Contreras explores the possibilities of extraction in relation to the relative order of multiple question phrases (45).

(45)a ¿Qué paquete no sabes cuánto pesa?

b *¿Cuánto no sabes qué paquete pesa?

Relative clauses have also received considerable attention. The one who has done the most research in this area is Rivero (see 1980b, 1984, 1986b); others are Plann 1980, García-Bellido 1981, D'Introno 1985, and Súñer 1984 and 1985. Rivero 1980b discusses a wide variety of relative clause types including those with a plain que (46a) as opposed to those with a true relative pronoun (46b), and infinitival relatives (46c).

(46)a El profesor que/*el cual/*quien habla es poeta.

b El cuchillo con el cual cortamos...

c ¿Has encontrado un libro que leer?

It is also Rivero (1980b) who carefully studies the properties which distinguish left-dislocated structures (47) from topicalized ones (48).

(47) Al partido carlista, dicen que lo legalizaron para las elecciones.

(48) Dinero, dicen que no tiene ec.

Note that only in the former is there need for a pronoun (or a NP) elsewhere in the string; in the latter, there is a gap.

Finally, Spanish comparatives are researched by Rivero 1981, Piera 1983, and Plamn 1984. Pseudo-clefts are analyzed in Goldsmith 1981.

Because of the similarities but also because of the subtle differences, causatives in Romance have a generative tradition of having been dealt with comparatively (cf. Bordelois 1974 for an early treatment of Spanish and French; Burzio 1983 for Italian and French, among many others).

Recent studies which highlight Spanish data include Zubizarreta 1985 and 1987, Goodall 1987, and Bordelois 1988. The many issues surrounding the analysis of causatives are too complex to cover in a few short paragraphs; therefore, what follows should only be taken as a teaser.

One of the problems with sentences like those in (49) is that they seem to give mixed signals.

(49)a Mara hizo/dejó hablar a Paco.

b Hicieron apagar el fuego (por/a los bomberos)

On the one hand, these sentences appear to be made up of two different clauses (the main and the embedded); on the other, they act as a single clause with respect to negation (50) and auxiliary elements (51) (which are restricted to one occurrence per clause), and with respect to the positioning of clitics (see 3. 1).

(50) *Dejó no cortar la leña.

(51) a *Hicieron ser apagado el fuego por los bomberos.

b *Dejó haber cortado la leña.

In addition, causatives are problematic regarding the position and Case-marking (accusative or dative) of the overt subject of the lower verb.

The classic generative analysis for causatives is to postulate a predicate-raising rule which fronts both the lower verb and its objects. Nevertheless, dissatisfaction with the lexically triggered nature of this rule plus questions about the landing site for the fronted verb phrase, together with the peculiarities pointed out above have led researchers to entertain other hypotheses. Zubizarreta 1985 contends within a theory of lexical structure, that causative verbs function morphosyntactically as affixes, even though they are certainly words from the morphophonological perspective. Goodall 1987 derives the idea partly from Zubizarreta that causatives have both a biclausal and a monoclausal structure coexisting at a single level of representation. The implication is that causative verbs subcategorize both for a clause and a verb phrase complement simultaneously. Bordelois 1988 offers the most novel approach. She claims that the only feature specific to causatives is that their complements lock Infl(ection). All their other characteristics fall out from independently postulated processes, i. e., processes needed elsewhere in the grammar. One of them is the double subcategorization that a verb like dejar enters into which gives rise to the controller pattern (52), and the ergative one (53)⁴⁶, where the former is independently justified

––––––––   840   ––––––––

by the behavior of ordenar, the latter by parecer.

(52) hice [PRO abrir la puerta] [a Pedro]

(53) hice [pro venir Juan]

Many other areas have been investigated, and it will not be possible to do justice to this diversity given the space limitations of this essay. I briefly mention but a few in no particular order.

The Empty Category Principle (ECP), the principle of the theory of Government which guarantees that non-pronominal empty categories are properly governed, has been the topic of much linguistic investigation. Rare is the work that does not touch on this core principle of the grammar in one way or another. However, within Spanish, it is Jaeggli 1988 (but see also Piera 1983, Contreras 1986) who especially centers his study on exploring the proper formulation of the ECP by comparing the different/similar behavior of Spanish and English in a variety of environments.

Parasitic gaps, as their name intimates, are ec's which are parasitic on a trace, that is, on another ec. Only infinitival adverbial clauses seem to allow pg's. Two examples are found in [55] ([55a] is from Contreras 1987, [55b] from Bordelois 1986).

(55)a Este es el autor [cuyos artículos]_i archivé t_i sin leer ec_i

b ¿[Qué libro]_i te llevaste t_i sin pagar ec_i?

Bordelois 1986 examines the conditions under which pg's may appear, while Contreras concludes that pg's are licensed by a modified version of Principle A of Binding theory (cf. [34]).

Zagona (1986 and 1989) hypothesizes that tense, just as any other category, needs licensing. She conceives of «times» as «temporal arguments» of the clause. She limits her studies to simple tenses using a comparative approach to capture the differences between Spanish and English.

Due to the recent split of lexical (V, N, A, P) vs. functional (Agr, Tn, Comp, Det and others) categories (Fukui and Speas 1986, among others), together with the assumption that the latter are heads of phrases just as the former, there has been renewed interest in exploring what are now considered functional projections. One fruitful area for Spanish syntax has been the investigation of the internal structure of NP's (56), seen as complements of D(eterminer) P.

(56) el retrato del rey de Goya de la benefactora Gómez

The hypothesis is that DPs have a structure parallel to that of sentences and that nouns have arguments just as verbs do. So in (56) el rey is interpreted as direct object/theme, Goya as subject/agent and la benefactora Gómez as the possessor. The possibilities of extraction from inside the DP domain have also been studied at length (see Torrego 1987, Campos 1988, Mallén 1989). Rivero 1986c concerns itself with the binding properties of reflexives, non-reflexives, and possessives within NPs.

The interpretation of overt and null pronouns is the topic of Montalbetti 1984 and 1986. He examines the circumstances under which pronouns can be linked to quantified expressions (such as muchos, todos, etc.), and discovers that contrary to the English (57a), which is ambiguous between the interpretation in (57b) and (57c), neither of the Spanish translations in (58) is. Version (58a) can only be interpreted as the coreferential (57b), and (58b) only as the bound (57c).

(57) a Many students think they are intelligent.

b (many x: x a student) x thinks they are intelligent

c (many x: x a student) x thinks x is intelligent

(58)a Muchos estudiantes piensan que ellos son inteligentes.

b Muchos estudiantes piensan que pro son inteligentes.

This difference, which occurs in other languages as well, gives rise to the Overt Pronoun Constraint (OPC) which essentially states that an overt pronoun cannot be locally bound by a quantified expression whenever the alternation overt/empty obtains. Luján 1987 has also investigated overt/null pronouns and concludes that their distribution corresponds to the stressed/unstressed distinction of non-null subject languages such as English. To illustrate, although Juan and pro can be coreferential in (59), the same does not happen when the corresponding overt pronoun is used (60), where only in (60b) may the two elements be understood as coreferent.

(59) a Cuando pro trabaja, Juan no bebe.

b Juan no bebe cuando pro trabaja.

(60) a *Cuando él trabaja, Juan no bebe.

b Juan no bebe cuando él trabaja.

Reciprocal and distributive anaphors have been the target of studies by Bosque (1985,

––––––––   841   ––––––––

and 1989). In each case the phethora of excellent examples is accompanied by a formal analysis which addresses the question of binding. Of special interest is the uncovering of the properties of sendos, a distributive quantifier, since neither most of the Romance languages nor Germánic have a counterpart for it.

Two dissertations published in Spain deserve mention. Hernanz 1982 investigates the different types of environments and conditions necessary for infinitives to surface, including instances with overt subjects (61), as well as «independent» infinitives (62).

(61) a Sonar el timbre y sobresaltarse la abuela fue todo uno.

b Lo mejor sería ir yo también.

(62)a ¿Darles azotes a los niños?

b ¡Ah, ser joven!

c Salir en seguida.

Brucart 1987 (and 1988) mainly concentrates on the examination of verbal gaps (63), comparatively speaking, a theme which has not received much attention in Spanish (but see discussion of Zagona 1988 below).

(63) a Pedro alquiló un piso y yo__ un estudio.

b Luis enseñaba gramática a Alejandro y María__ a Pedro.

c *Pedro alquiló un piso y yo __ __.

Among the requirements that «gapping» must fulfill is that there be a lexical «remnant» (cf. [63a] vs. [63c]) whose mission is to make the recovery of the null element(s) possible. Hence, verbal gaps must be identifiable much as nominal gaps (cf. 2. 2 ) must be, a welcome and not surprising result given the hypothesis that licensing principles apply across categories.

Another important dissertation, this one from the United States, is that of Zagona 1988. As its title, Verb Phrase Syntax, indicates, it is a detailed study of the structure of verb phrases in English and Spanish, and of the syntactic processes involved. It is shown that although the structure dominating «auxiliary» verbs is alike in both languages, the syntax of verb projections such as VP-deletion (64), Tag-questions (65), VP-preposing (66), and V-movement (67) distinguishes the two languages.

(64) a These books have been read and those have [e] too.

b *Estos libros han sido leídos, y esos han [e] también.

(65)a They haven't left, have they [e]?

b *No han salido, ¿han [e]?

(66) a They said they would leave, and [leaving] they are [e]

b *Dijeron que saldrían, y [saliendo] (ellos) están [e].

(67) a *Left John?

b ¿Salió Juan?

Zagona argues that the way to capture the above differences is by parametrizing surface structure representations as justified by the dissimilarities in inflectional features which affect government relations differently in each language.

In this subsection, I would like to stress additional areas in which theoretical studies on Spanish syntax and semantics can contribute to Spanish studies in general. Although some work has already been carried out, much more could be achieved if more researchers were to see the potential (and rewards) of further research in these subjects.

Spanish linguistics is rich in descriptive/ structural essays concerned with language variation, be they from the historical («vertical») point of view, or from the synchronic («horizontal») side (such as dialectology and language in contact situations). What many researchers have failed to appreciate is that the comparative approach of the principles-and-parameters model is ideally suited for explaining language variation in all its forms. For obvious reasons, in what follows I concentrate on syntactic works.

Almost single-handedly, Rivero has successfully begun to bridge the gap between philological studies and current theories in the area of diachrony (see Rivero 1984, 1986a, and Lema and Rivero 1989). In her 1986a article, Rivero dwells on the characteristics of Old Spanish (ca. 1200-1450) clitics and compares them to their contemporary counterparts. Their different categorial status naturally accounts for differences in their positioning and mobility, and has consequences for the analysis of clitic-doubling. (For the position of Old Spanish clitics, see also Alemán 1985).

Lema and Rivero 1989 offer a contrastive essay on «inverted conjugations» from the Romance and Balkan perspectives. Old Spanish partakes of this type of non-finite verb movement to Comp where the auxiliary stays behind (68).

(68) a Si yo vivo, doblar vos he la soldada (Cid 80).

b E mandar vos he dar todo (Zif 128).

This category is split into two: dialectal diversity, and languages-in-contact situations.

4. 2. 1. Dialectal situations

Not many are the studies which focus on dialectal variation from a theoretical perspective. Súñer 1986c explores the conditions for clitic-doubling of direct objects (DOs) in four Spanish dialects (Porteño, leísta and laísta variants of Madrid, and Quiteño). The conclusion reached is that regardless of the shape which DO clitics manifest in each dialect, DO clitic-doubling obeys a specificity constraint in each. This argues in favor of the speakers' unconscious knowledge of grammatical relations.

Súñer in 1986a (and to a lesser degree Súñer 1989) investigates the different possibilities for the positioning of the subject argument in Caribbean and general Spanish. The former allows for examples like those in (69) with a preverbal subject, which general Spanish bans.

(69)a No tiene más que uno parar la oreja.

...está muy caliente para yo tomarlo.

b ¿Cómo tú te llamas?

¿Qué Iván dijo de eso?

The difference is attributed to the weakness of Agr(eement) in Caribbean and its relation to Case theory.

4. 2. 2. Languages in contact: Code-switching

Code-switching (CS) is the label given to the phenomenon particular to bilinguals in which they switch back and forth between two languages [70] (examples cited in Woolford 1983).

(70)a Estaba training para pelear.

b En Puerto Rico he would say que cortaba caña, even though tenía su negocio, you know.

c *I went to the house chiquita.

d *yo lo bought.

e *Arrived the gran jefe.

CS is motivated by social reasons; thus, it has been intensively studied by sociolinguists. However, many of the constraints proposed on CS support the hierarchical structure of linguistic constituents (X-bar theory) and moreover, since switches occur only in those places where the constituent boundaries and the rules of the two relevant grammars coincide, CS is a highly constrained process which demonstrates the linguistic competence of the speakers in two distinct formal systems; i. e., there is no CS grammar per se.

Woolford 1983 develops a model of bilingual CS which illustrates how the two grammars cooperate to end up with only the allowable sequences in (70). Klavans 1985 argues that CS is asymmetrical and that the notion «matrix language» (as defined by verbal inflection) helps in explaining certain constraints on CS.

This decade has witnessed the emergence of Romance and to a lesser extent Spanish syntax as a contributor to general theory, a situation which did not obtain during the structuralist period or even during early transformational grammar. Contreras's 1981 essay about Romance linguistics in the seventies states «Romance linguists have definitely not been at the forefront of theoretical development. There are indications, however, that this situation might change». And indeed it has.

In particular, Spanish syntax has greatly contributed to the development and understanding of the traits and conditions for little pro, especially as subject of null subject languages both in its definite and its generalized arbitrary interpretations. To my knowledge, it was Súñer 1982b who first pointed out that the properties of null subjects in finite sentences had very different characteristics from the ungoverned Caseless PRO. Its arb reading is studied in Súñer 1983, Jaeggli 1986b , and Hernanz 1988.

In addition, other significant theoretical contributions must minimally include the following: Rivero 1988, and Lema and Rivero 1989, who broaden the class of possible motivating factors for verb movement from the perspective of the Balkan and Romance languages; Zagona 1988 and 1989, who works to establish the parameters of tense identification and licensing; and Jaeggli 1988, who pushes for a split characterization of the ECP with important theoretical repercussions concerning the levels at which licensing of traces should be effected.

The role of linguists working on Spanish can also be appreciated by noting that their works are increasingly cited in the bibliographies of more general texts; see Radford 1988, and Harris and Vincent (eds.) 1988, for two examples.

The eighties has been an exciting period for syntax and semantics. The comparative and modular approach to the study of the properties of language is evident throughout the works mentioned in this essay. And the results are clear: we have a better understanding of how languages are structured. But if much has been accomplished, the consensus has to be that a lot remains to be done. In the first place, even in the areas that have been researched, it would be premature to claim that we have exhausted the subject. It must be kept in mind that other crucial data may be uncovered, that new theoretical developments may show how to integrate the modules of the grammar more tightly so that redundancies are avoided, or that considerations from other languages may lead us to rethink a particular analysis. All of this is part of the scientific method where one work builds on another, thereby making progress possible. Secondly, it must be pointed out that some studies have created controversy; unanimity of opinion is hard to come by in linguistics probably because we have access only to performance data while what we are really interested in explaining and understanding is the speaker's competence. Therefore, discussion of some of these controversial issues is bound to continue⁴⁷. Thirdly, areas of research exist where work has just begun: besides those examined in Section 4, we can suggest those aiming at integrating inflectional morphology with the syntax, or at deriving syntactic properties from the lexicon⁴⁸, or at exploring in depth the different word order possibilities and their derivation, among many others.

Two related informational issues that might be of interest to the reader are where to start reading about the principles-and parameters model, and how to keep informed about Spanish syntax and semantics. As answer to the first question, I would recommend two of the more accessible books by Chomsky: The Managua Lectures (1988a) which uses mostly Spanish examples, and Knowledge of Language (1986a), read in that order. For the second question, in addition to articles published in major specialized journals and books, it must be remembered that in April 1990, the University of Ottawa is sponsoring the Twentieth Linguistic Symposium on Romance Languages, a long standing and successful forum for Romance linguists where one can gather considerable information about the field. Selected papers from the proceedings of these meetings are always published. I have included the ones from LSRL 12-17 among the references. Moreover, Spanish topics are generally also represented in other national conferences as well.

Before closing I would like to stress that this essay is both non-exhaustive and subjective. Non-exhaustive, because not all works of the eighties have been mentioned, which means that omissions, maybe even important and non-intentional ones, have occurred. Subjective, because I am writing from my own perspective, highlighting the issues that I consider most interesting, and using my own research for illustration in many instances. For all possible shortcomings and misinterpretations, I apologize⁴⁹.

Alemán, I. 1985. «The Position of Clitics in Old Spanish». Cornell Working Papers in Linguistics 7: 1-17.

Baker, M. 1988. Incorporation. Chicago: Univ. of Chicago Press.

Baldi, P. (ed.). 1984. Papers from the XIIth Linguistic Symposium on Romance Languages. Amsterdam: Benjamins.

Bell, A. 1980. «Mood in Spanish: A Discussion of Some Recent Proposals». Hispania 63: 377-90.

Birdsong, D. and J. P. Montreuil (eds.). 1988. Advances in Romance Linguistics (LSRL 16). Dordrecht: Foris.

Blake, R. 1985. «From Research to the Classroom: Notes on the Subjunctive». Hispania 68: 166-73.

Bordelois, I. 1974. «The Grammar of Spanish Causative Constructions». Dis. MIT

——. 1982. «Transparency». Linguistic Analysis 9: 161-203.

——. 1986. «Parasitic Gaps: Extensions of Restructuring». In I. Bordelois et al. (eds.).

——. 1988. «Causatives: From the Lexicon to the Syntax». Natural Language and Linguistic Theory 6: 57-93.

Bordelois, I., H. Contreras, and K. Zagona (eds.). 1986. Generative Studies in Spanish Syntax. Dordrecht: Foris.

Borer, H. 1983. Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris.

Bosque, I. 1980. Sobre la negación. Madrid: Cátedra.

——. 1984a. «La selección de las palabras interrogativas». Verba 11: 245-73.

——. 1984b. «Sobre la sintaxis de las oraciones exclamativas» Hispanic Linguistics 1: 283-304.

——. 1985. «Sobre las oraciones recíprocas en español» Revista española de lingüística 15: 59-96.

——. 1989. «Anáforas distributivas: La gramática de sendos». Univ. Complutense de Madrid unpublished ms.

——(ed.). Forthcoming. Tiempo y aspecto en español. Madrid: Cátedra.

Brucart, J. M. 1987. La elisión sintáctica en español. Bellaterra: Univ. Autónoma de Barcelona.

——. 1988. «La elipsis parcial». In V. Demonte and M. Fdez. Lagunilla (eds.). 291-328.

Burzio, L. 1986. Italian Syntax. Dordrecht: Reidel.

Campos, H. 1986. «Indefinite Object Drop». Linguistic Inquiry 17: 354-59.

——. 1988. «Extraction from NP in Spanish». Romance Linguistics and Literature Review 1: 1-32.

Campos, H. and M. Zampini. 1989. «Focalization Strategies in Spanish». Georgetown Univ. unpublished ms.

Casagrande, J. 1972. «Syntactic Studies in Romance». In J. Casagrande and B. Saciuk (eds.), Generative Studies in the Romance Languages. Rowley, MA: Newbery House, 1-22.

Chomsky, N. 1977. «On WH-Movement». In P. Culicover et al. (eds.), Formal Syntax. NY: Academic Press, 71-132.

——. 1981. Lectures on Government and Binding. Dordrecht: Foris.

——. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, MA: MIT Press.

——. 1986a. Knowledge of Language. NY: Praeger.

——. 1986b. Barriers. Cambridge, MA: MIT Press.

——. 1988a. Language and Problems of Knowledge: The Managua Lectures. Cambridge, MA: MIT Press.

——. 1988b. «Some Notes on Economy of Derivation and Representation», MIT unpublished ms.

Contreras, H. 1981. «Romance Syntax in the seventies». Proceedings of the Tenth Anniversary Symposium on Romance Linguistics. H. Contreras and J. Klausenburger (eds.). Papers in Romance, Supplement II, vol. 3: 26-40. Seattle, WA: Univ. of Washington.

——. 1984. «Multiple Questions in English and Spanish». In P. Baldi (ed.), 121-33.

——. 1985. «Clausal Case-marking and the CRP». In L. King and C. Maley (eds.), 19-30.

——. 1986. «Spanish Bare NPs and the ECP». In I. Bordelois et al. (eds.), 25-49.

——. 1987a. «Small Clauses in English and Spanish». Natural Language and Linguistic Theory 5: 225-43.

——. 1987b. «Parasitic Chains and Binding». In C. Neidle and R. Núñez-Cedeño (ed.), 61-78.

——. 1989. «Closed Domains». To appear in Probus.

Demonte, V. 1987. «C-Command, Prepositions and Predication». Linguistic Inquiry 18: 147-57.

——. 1988. «Rección y minimalidad en el sintagma nominal». In V. Demonte and M. Fdez. Lagunilla (eds.), 252-90.

——. 1989. «Linking and Case: The Case of Prepositional Verbs», Univ. Autónoma de Madrid unpublished ms., read at the LSRL 19.

Demonte, V., and M. Fernández Lagunilla (eds.).1988. Sintaxis de las lenguas románicas. Madrid: Ed. Arquero.

D'Introno, F. 1985a. Sintaxis y semántica en la lingüística generativa: El modelo gramatical GB. Caracas: Instituto Pedagógico de Caracas.

——. 1985b. «Relativization». In P. Baldi (ed.), 135-52.

Fukui, N. 1986. «A Theory of Category Projection and Its Applications». Dis. MIT.

Fukui, N. and M. Speas. 1986. «Specifiers and Projections», MIT Working Papers, vol. 8.

García-Bellido, P. 1981. «Switching Category in the Syntax: The Nominalization of Relative Clauses». In H. Contreras and J. Klausenberger (eds.), 101-14.

Goldsmith, J. 1981. «On Spanish Pseudoclefts». Glossa 15: 3-15.

Goodall, G. 1987. «Case, Clitics, and Lexical NPs in Romance Causatives». In C. Neidle and R. Núñez Cedeño (eds.), 93-105.

Gruber, J. 1965. «Studies in Lexical Relations». Dis. MIT.

Harris, M. and V. Nigel (eds.). 1988. The Romance Languages. NY: Oxford Univ. Press.

Hernanz, M. L. 1982. El infinitivo en español. Bellaterra: Univ. Autónoma de Barcelona.

——. 1988. «En torno a los sujetos arbitrarios: La 2ª persona del singular». Univ. Autónoma de Barcelona unpublished ms.

Hernanz, M. L. and J. M. Brucart. 1987. La Sintaxis. Barcelona: