1The Syntactic Operator se in Spanish
Eva Juarros Daussà
University of Massachusetts, Amherst
1 I wish to thank the people who have had the most direct impact in the creation of this paper. Thank you to Peggy Speas and Ellen Woolford; they were invaluable, both intellectually and with their wonderful personal support. Thanks to John Kingston for being a great chairman of my committee, for generous and wise advice and encouragement. Thanks to Barbara Partee and Roger Higgins for punctual remarks and guidance. Thanks to my year mates Paul DeLacy, Nancy Hall, Meredith Landman, Marcin Morzycki and Anita Nowak for patiently listening to my parental explanations at different times of the development of this paper, in the third year seminar.
2. The Data
3.1. About the external argument
3.1.1. Se with transitive verbs
3.1.2. Se with intransitive verbs
3.1.3. Operator se and passive morphology
3.2. About feature blocking
3.2.1. Feature geometry of AgrS
3.2.2. At the interface (syntax and morphology limiting each other)
3.2.3. Other gaps in the paradigm: as predicted
4. Previous accounts
5. Conclusion: the present account
2 has been a persistent problem in Linguistics. The The analysis of the Romance clitic se
challenge consists in explaining the fact that se appears in a variety of constructions that at
first sight do not seem to be related. Those constructions include reflexive, reciprocal, impersonal, passive, and middle constructions, and an array of aspectually nuanced sentences, examples of all of which will be given below. Moreover, se appears also as an
allophone of the third person dative clitic, le(s), when it co-occurs with an accusative
I here present an account of all these constructions in which I suggest that some new ideas in the recent literature can prove useful in describing the se phenomena. Focusing on both
the effect that se has over the syntactic realization of the verbal arguments, on the one hand, and on the agreement and case patterns of se constructions, I propose that, syntactically,
there is one single lexical item se, and that is an operator that has two very defined
functions. The first one is to block the realization of the external argument of the verb, and the second one is to block certain features of the Agreement nodes of the sentence. I show that operator se is uniformly generated in head position of the functional node that would otherwise obligatorily introduce the external argument of the predicate. As a clitic, it adjoins to the verb and undergoes head movement through all the functional nodes of the sentence. As it moves, it interacts with the Agreement nodes of the sentence; more specifically, it selectively inhibits the activation of the features [case], [person] and [number] for further checking purposes. It is also claimed that se acts within the limits set
by morphology; this means that se itself operates randomly, subject to no restriction, but only those sentences that conform to the restrictions of morphology will survive the
12 Cognates of this clitic, inherited from the Latin reflexive pronoun, exist in all Romance languages (si in
Italian, se in Spanish, Portuguese and French, es or se in Catalan, etc). While there are some differences in the
distribution of se within those languages, a satisfactory analysis does not exist for any of them. In this paper I will be focused on the syntactic properties of Spanish se, with the hope that the analysis presented here can be
exported across (and beyond) Romance. 3 For a diachronic explanation, see, for example, Lapesa (19xx: 209), who, in a nutshell, gives the following evolution from Latin illi illum to Spanish se lo: illi illum > [*eljelo] > [*ezelo] > [zelo] > [se lo]. See also
Schmidely, Jack (1978): De “ge lo” a “se lo”. Cahiers de Linguistique Hispanique Medievale, 4, 63-70. The
syncronic coincidence of a dative clitic with se follows principles that are independent of the rest of cases, as
shown by Grimshaw (1998). I will thus ignore dative se in this paper.
derivation, which explains why we have the constructions that we do and no others. Under the view presented in this paper, the different constructions with se have in common a
property that I will identify as being syntactic unaccusatives, and that consists in not having an external argument; they differ in the composition of their agreement nodes, as determined by se.
Previous accounts of se fall into two groups: (a) those that distinguish the different
functions of se on the basis of inherent properties of several homonymous morphemes se
(Kayne (1975), Grimshaw (1982), Belletti (1982), Zubizarreta (1982, 1987), Burzio (1986), Cinque (1988), or Dobrovie-Sorin (1994, 1998); and (b) those that argue that the surface similarities between these constructions are not accidental, and hence give a unified account of all instances of se (Manzini (1986), Wehrli (1986), Postma (1993), Everett
(1996), Bruhn de Garavito (1999). In section ?4 we will make a brief review of work done in both groups. There has been work in both groups trying to explain the properties of the different constructions using existent mechanisms, such as theta and case theory (for example, differentiating between instances in which se receives nominative or accusative
case, or a theta role, and instances in which it doesn‟t). The account presented in this paper follows the unified aim of the second group. It however looks at the problem from a slightly different point of view. By claiming that se is uniformly generated as the head in
complementary distribution with the element that would otherwise introduce the external argument of the verb, it separated theta theory from syntactic realization of arguments. The case effects are related to the rest of agreement effects, and explained by the compositional possibilities of the syntactic features of the Agr nodes of the sentence, as dictated by morphology.
2. The Data
The clitic se in Spanish is used in a variety of constructions. Despite the huge literature on the topic, there is not total agreement in the number and classification of the different constructions. Here I am mainly following the traditional classification found, for example, in Alcina and Blecua (1980); I differ from this source when I present what I call middle
construction, which does not receive a separate treatment in traditional grammars, but is normally resumed under the reflexive passive. The reason for considering the middle a separate construction corresponds to a desire to relate it construction with the similar one found in other languages, such as English. However, the traditional grammars are right in grouping both constructions together, since they are syntactically very similar, as will be apparent in our treatment of them below.
Here I present examples of each construction, with the corresponding English glosses. I then include a short description of the properties that are going to be crucial in my analysis, mainly, those concerning the status of the external argument and the composition of the agreement nodes. As for this latter factor, I adopt the minimalist convention (see, for example, Chomsky 1993, 1995) that each of the agreement nodes of the Spanish sentence (AgrS and AgrO) contains a [case] feature (nominative or accusative, respectively); I will further assume that in Spanish AgrS has the phi-features [person] and [number], since subjects agree in those features with the verb. There is no evidence to conclude that in Spanish AgrO contains such phi-features, since objects never show any kind of agreement
4 with the verb.
Now, it appears from the data that the features of the Agr nodes of the sentence that would otherwise be active can under certain circumstances appear to be inactive; for example, in a typical sentence, the feature [person] would be active in AgrS, forcing the subject DP and the verb to agree in grammatical person when these two elements check their [person] feature against the [person] feature of AgrS. In the case just described, I say that [person] in AgrS is active, and fro the present purposes signal this fact by claiming that AgrS is [+ person]. Now, when agreement in grammatical person between the verb and the DP does not occur, I will assume that the features [person] is somehow inactive for checking purposes, and signal such status by claiming that AgrS is [- person]. Following this convention, the data shows the paradigm summarized in the table below:
4 Chomsky (1993) claims that the composition of the agreement nodes of the sentence should be the same, to facilitate acquisition of the language. However, it seems to me that this theory-driven assumption is not sufficiently founded, and hence I will adopt a more conservative position and refrain from adopting such idea.
External AgrS AgrS AgrS AgrO
argument nom case Person Number acc case
Impersonal - - - - +
Reflex Pass - + - + -
Middle - + - + -
Refl/Recipr - + + + -
The last clarification in the description of the data that follows concerns the so-called “strength” of the Agreement nodes. Here I again follow minimalist assumptions in considering that a functional node is “strong” when its features have to be checked overtly (that is, by the time the sentence is pronounced), since they are not interpretable and would cause the derivations to crash if present in the LF representation. The way to eliminate features is to move a proper “checker” to the specifier (for XPs) or by head-adjunction (for
X) of the functional node in question. On the other hand, “weak” nodes have features that, since they are interpretable, can survive through LF, and therefore the movement of the checker element can be procrastinated until after the spell-out point. The goal of the system is mainly to provide a formal code for the different positions that the elements occupy in the tree. If an element is moved from the position which it is claimed to have been originated in, the relevant functional node that shows an agreement relation with it is said to be “strong”; otherwise, it is “weak”. We are now in a position to examine in detail each construction and motivate the systematic description summarized in the table above.
Referred to as such in virtually all the works on se, the impersonal construction is the most widely known. In Spanish, it appears both with intransitive (1a) and transitive (1b) verbs:
1a. Aquí se vive bien
Here se lives well
"One lives well here"
b. En esta granja se mata a seis mil pollos cada día
in this farm se kill.3s to six thousand chickens every day
5 "In this farm they kill six thousand chicken a day"
In both (1a) and (1b), the external argument of the verb, which would canonically occupy the subject position of the sentence, is not realized. This is apparent in (1a), where there is no DP that would qualify for the subject position. That the post-verbal DP in (1b) is not the subject of the sentence becomes clear when we substitute a clitic for it, which has to be in accusative case:
1c. Se los mata
se cl.acc kill.3s
“They kill them”
6Since the subject is absent, the verb, which would normally agree with the subject, is here
showing a default third person singular morphology. This situation is only possible if we assume that the features in AgrS are not active, and therefore, do not need to be checked. The
7composition of AgrS is then the following: [- (nominative) case], [- person], [- number].
8Moreover, from the data we can conclude that AgrO is [+ (accusative) case].
5 Note that the equivalent translations of the transitive and intransitive variant into English make use necessary use of different impersonal pronouns: „one‟ in the case of intransitives, and „they‟ in the case of
transitives. It is not clear to me the extent to which this fact might be relevant. One possibility is that the difference is a point of view phenomenon, i.e., a matter of whether the speaker/hearer is to be included. For a distinction between two kinds of impersonal constructions with Italian si (the Italian counterpart of the
Spanish se dealt with in this paper), see Cinque (1988). This author makes the case for argumental si (in
tensed and nontensed sentences of transitive and unergative verbs), and non-argumental si (only in
non-tensed sentences, with unaccusative verbs and other verbs that lack an external theta-role). Since the arguments for this distinction presented in Cinque‟s analysis do not transpose to Spanish counterparts of the
Italian sentences (see Zubizarreta 1982, 1987 for a published proof of it), I will not adopt Cinque‟s distinction in my treatment of Spanish se. 6 This claim is at odds with proposals such as the one by Mendikoetxea (1996), in which PRO is claimed to be the subject of the sentence. 7 Notice that nothing is said about the relative strength of the functional nodes of the sentence; since the verb moves through all functional positions in the sentence, there is no reliable way to know whether the object has moved, and hence whether AgrO is strong or weak. We can assume that the strength of AgrO is irrelevant for the analysis, as will become obvious later on. 8 Notice that this state of affaires is a counter-example to Burzio‟s generalization (BG). Recall that BG states
that a verb that does not have an external theta role cannot assign (in Burzio‟s terminology) accusative case. Here, the sentence does not have an external theta role, but it has an accusative DP.
b) Reflexive Passive
This construction receives other names throughout the literature: passive (Belletti (1982), Zubizarreta (1987), Everett (1995) etc.), middle (Manzini 1983), impersonal reflexive (Bruhn de Garavito (1999)). Here I name it after its traditional terminology (Alcina and Blecua‟s (1980) pasiva refleja). In the reflexive passive, as in the impersonal, the external argument of the verb is not realized. It differs from the impersonal construction in that the verb shows some agreement with the post-verbal DP, which is allegedly in nominative case (as its ungrammatical substitution by an accusative clitic shows in (2b):
2a. Se necesitan buenas ideas
se need.3pl good ideas
"Good ideas are needed"
2b. *Se las necesitan
se acc.cl need.3pl
“They are needed”
The fact that an accusative clitic is not possible indicates that AgrO is [- (accusative) case]. The post-verbal DP is therefore marked with nominative case, and hence AgrS should be [+ (nominative) case]; since the DP remains in post-verbal position, we should assume that
AgrS is weak. As for verbal agreement, it is clear that the verb agrees in number with the post-verbal DP, as a change in the number of the DP obligatorily triggers a change in the verb (compare (2a) with (2c) below):
9 2c. Se necesita/*n una buena idea
se need.3sg/*pl a good idea
"A good idea is needed"
9 Notice that sentence (2c) could equally be analyzed as an impersonal construction; this is so because the morphology of the verb coincides with the default third person singular of impersonals; to avoid this ambiguity, in the future I‟ll use the marked plural when I am referring to passive reflexives.
Now, to test for person agreement, we can change the person (but not number) of the nominative, post-verbal DP, and observe whether such operation triggers a change in verbal person. As (2d) shows, changing the person of the verb along with that of the DP (in this case a second person pronoun) results in ungrammaticality:
2d. *Se necesitáis vosotros
se need.2pl you.nom.2pl need.2pl
10 Intended: “you are needed”
From (2d), we can safely conclude that AgrS is [- person], [+ number].
Zubizarreta (1987) and Burzio (1986) call this construction ergative, and Everett (1996) follows them in this terminology. Postma (1995) gives the alternative terms inchoative, ergative and anticausative to this group of constructions. Frequently claimed to be idiosyncratic in its distribution, it is widely ignored in most of the treatments of se. I here
deny that this construction be lexically determined; in part, my choice of the term “middle” is
intended to relate this construction to the tradition of studies about the middle construction in English (see, for example Hale and Keyser (1987, 1993)).
The middle construction is identical to the reflexive passive, except that AgrS is strong, since the DP is preverbal. Hence, the external argument is not realized, AgrO is [-
10 The grammatical way to express the intended meaning is with the impersonal in (i), in which the verb presents no agreement with the accusative clitic
i. Se os necesita
Se you.pl need.3s
“You are needed”
It is not clear to me why one cannot construct a Reflexive Passive with a non-third person plural DP (in nominative) and allow the verb to agree in number, with the default third person morphology:
ii. * Se necesitan a vosotros
se need.3pl to you.2pl
Further research must show whether the ungrammaticality of (ii) is due to factors of the distribution of strong and weak pronouns in Spanish.
(accusative) case] (as shown by (3b)), and AgrS is [+ (nominative) case], [+ number]
11 (shown by (3c)): (shown by (3a)) and [- person]
3a. Las puertas se han cerrado de golpe
The doors se have.3pl closed with sudden
"The doors suddenly closed"
3b. * Se las cerraron de golpe
se acc.cl closed.3pl with sudden
“Them suddenly closed”
3c. *Vosotros se necesitáis
you.nom.2pl se need.2pl
Intended: “you are needed”
d) Reflexive and Reciprocal
Spanish reflexive sentences with a third person are constructed with se and the verb agreeing
in person and number with the DP of the sentence:
4a. Marta se está limpiando
11 There is a construction that is very similar to the middle (actually I am not sure it is different from it). It arises when the DP is not in third person; then, we have a first or second person pronoun, a first or second person oblique clitic, and the verb agreeing in person and number with both:
i. Yo me cerré en banda
I obl.cl closed.1s in line
“I closed up” (*”I closed myself”)
I discard this case as a counterexample to my analysis of inert AgrO for several reasons: (1) all the instances of non-third persons I could find are idiomatic, i.e., they are highly crystallized in the language, and as such might not be analyzable in the same way as their creative counterparts; (2) it is not clear to me which case the clitic in these constructions has, since the forms are homophonous for both dative and accusative in all cases; (3) the translation of these sentences into English requires the use of a particle construction, unlike the true
middle construction, suggesting that we are dealing with two different constructions.