The formal properties of non paradigmatic SE

Following Oca (1914), this article argues that passive and impersonal se constructions in Spanish are regular transitive constructions where the pronominal clitic se is the argumental subject. Several arguments (secondary predication, non-argumental predicates, control and obviation, anaphora binding, active morphology or its alignment with overt nominative pronouns, among others) show that (i) both constructions are active structures, (ii) despite what agreement facts might suggest, in both the internal argument of the verb is not the subject but the direct object throughout the derivation, (iii) se is the active nominative pronominal subject of the construction. We argue that the alleged ‘special’ properties of passive-se are not construction-specific but follow from the lexical specifications of se agreeing with Tense as a quirky subject.


Introduction
The goal of this paper is to analyze the role of SE in non-paradigmatic SE constructions in Spanish.Specifically, we deal with impersonal (1a) and passive (1b) SE constructions as described in Mendikoetxea (1999) and other works.
Se dejó a las mujeres maltratadas sin protección efectiva SE left.sgA the women abused without protection effective 'Abused women were left without effective protection' b.
Se dejaron las reivindicaciones de los trabajadores en un cajón SE left.pl the demands of the workers in a drawer 'The workers' demands were left aside' In Standard Spanish, the key difference between (1a) and (1b) lies in verbal agreement.Thus while in the so-called SE impersonal construction (1a) number agreement with the complement a las mujeres maltratadas is not possible (2a), in so-called SE passive constructions agreement between the verb and its complement is mandatory (1b)-(2b).
(2) a. *Se dejaron a las mujeres maltratadas sin protección efectiva SE left.plA the women abused without protection effective 'Abused women were left without effective protection' b.
*Se dejó las reivindicaciones de los trabajadores en un cajón SE left.sg the demands of the workers in a drawer 'The workers' demands were left aside' It is precisely because of this contrast that it is generally assumed that the complement in (1b) raises to subject position, while the complement in (1a) stays in object position, where it receives Differential Object Marking (DOM) (Rivero 2002).Analyses differ regarding the structural details of these relations, but there is a broad consensus in the literature that (1a) and (1b) are instances of different constructions.
In this paper, we update a proposal already anticipated by Oca (1914) who suggested that the clitic SE is the subject of the sentence.We propose a unified derivation of SE-passives and SEimpersonals as regular active constructions where SE is the nominative subject.The "nominative SE" hypothesis has been argued for "impersonal" SE in (1a) by many authors before us (e.g.Cinque 1988;see D'Alessandro (2007), ch. 2 for a comprehensive revision of the literature in this regard), but extending it to "passive" SE, as we do, has been very uncommon.We also argue that the characteristic properties of the constructions and their differences derive from the interaction of Case theory and the "quirky" nature of SE as the subject of the construction; specifically, its lack of person and number ϕ features.The paper is structured as follows.In section 2 we go through some of the more characteristic properties attributed in the literature to non paradigmatic SE constructions in Spanish, and we check whether these properties are compatible with the idea that SE is actually a nominative subject.In section 3, on the other hand, we test the predictions made by this hypothesis, and we show that evidence strongly supports it.

General properties of passive and impersonal SE constructions
In this section we analyze the main distinctive properties of SE constructions discussed in the literature, most explicitly organized in Mendikoetxea (1999), Sánchez (2002) and NGLE, and we complete the standard discussion of control with obviation tests and add new observations concerning the behavior of SE constructions in pseudo relatives.We show that these properties are not only compatible with our proposal, but that they actually support it in a remarkably straightforward way.

Agreement patterns
The general agreement situation in SE constructions in the case of non transitive verbs is for Tense to appear in third person singular, arguably, a default form (3).
En esta empresa se trabaja sin descanso In this company SE work.3sgwithout rest 'In this company people work restlessly' b.
Se anduvo hasta el cruce SE walked.3sguntil the crossing 'People walked until the junction c.
Se On the other hand, when the object cannot be marked for DOM (5a)-(5b), the verb triggers number agreement with it (5c)-(5d) (observe the contrast between (4c) and (5c)).
( ).In such contexts the number-agreeing DP does not need to be the subject of the clause.In fact, as we will show next, this element continues to be the VP internal argument in SE-passive constructions in Spanish.

A non-overt argument is always needed
As observed in the literature (Mendikoetxea 1999, Sánchez 2002, and references therein for discussion and examples), non-paradigmatic SE constructions are compatible with almost all types of verbs: transitive, unaccusative, unergative, ditransitive, psych verbs, alternating verbs (locative, dative, causative-inchoative, etc.), propositional verbs, etc.; even, under certain circumstances, raising verbs (as we discuss in section 3.2 below).However, from the point of view of the argument structure, there is a general constraint against SE constructions: 4 They are not available with argumentless verbs, typically atmospheric ones (llover 'to rain', atardecer 'to get dark' etc.), as the example in (8) illustrates.

Huele a rosas Smells like roses 4
There are at least two other general restrictions.One of them has to do with the animacy requirement of the subjects and is discussed in sec.2.4; the second one affects constructions where there is another obligatory source for SE in addition to the impersonal one, including inherently SE-marked verbs (i), the anticausative use of transitive alternations (iib), and reflexive constructions with SE (iii), among others: The impossibility of SE-SE combinations is not a trivial matter and raises interesting questions related to both morphology and syntax, but those issues are beyond the scope of this paper and will be left aside here (see Burzio 1986, Cinque 1995, '(S)he smells like roses' 'There is a rose smell' b.
Se huele a rosas se smells like roses 'Someone/Everybody smells like roses' Thus, the distribution of SE-constructions is exactly what is predicted if SE stands for the higher argument: In those cases where V's theta grid lacks any argument, there is no position where SE can be inserted, and, in consequence, a non paradigmatic SE construction cannot be generated.

Infinitive control and subjunctive obviation
A second property of SE-constructions that supports the presence of a syntactically active subject is that they show the same obligatory control/disjoint reference distribution as in infinitive/subjunctive alternations of any other regular active sentences.The minimal pair in (10) illustrates that in Spanish: (10) a. Los trabajadores intentaron enviar un representante a la reunión The workers tried to.send a representative to the meeting 'The workers tried to send a representative to the meeting' b.
Los trabajadores intentaron que enviara(n) un representante a la reunión The workers tried that send.SUBJ.sg/pl a representative to the meeting 'The workers tried for her/him/them (≠ the workers) to send a representative to the meeting' (10a) is a regular infinitival structure with intentar ('try'), where the subject of the embedded infinitival construction must be obligatorily controlled by the matrix subject los trabajadores ('the workers').As is well known, if the embedded infinitival construction is substituted by a subjunctive complement (10b), the subject of that subjunctive clause must be referentially disjoint from the main subject los trabajadores ('the workers').This disjoint reference effect between the two subjects is characteristic of Romance subjunctive constructions with volitional verbs and some other types of predicates like decidir 'decide', intentar 'try', proponer 'propose', etc.Given that, consider now the minimal pair of se constructions in (11).
Se intentó enviar unos representantes a la reunión SE tried.sgto.send a representative to the meeting 'They (generic, indefinite) tried to send a representative to the meeting' b.
Se intentó que enviaran unos representantes a la reunión SE decided.sg that send.SUBJ.plsome representatives to the meeting 'They (generic, indefinite) tried for x (x ≠ them) to send some representatives to the meeting' In (11a), the infinitival subject must be controlled by the matrix "implicit" argument of the SE constructions, whatever its generic/indefinite interpretation is.In contrast, the subject of the subjunctive complement in (11b) must be disjoint in reference from the "implicit" argument of the matrix clause.
Notice also that, in contrast, control/obviation context are highly degraded in analytical passives in minimal pair structures: (12) a * (Allí) fue decidido enviar un representante a la reunión (There) was decided to.send a representative to the meeting b * (Allí) fue decidido que enviara un representante a la reunión There was decided that send.SUBJ a representative to the meeting These facts are, again, completely in line with our proposal.As far as there is a subject in the matrix clause, as we propose SE is, the subject of the embedded clause can have conjoint or disjoint reference with it, and, in consequence, it is expected to have a conjoint subject with infinitives (11a), and a disjoint one with subjunctive mood (11b).5

The "missing" argument is always animate
There is an important property that has been interpreted in different ways in the literature but, considered in a broader context, supports the analysis of SE as the pronominal argument occupying the subject position in the constructions under discussion.We are referring to the well observed fact that the missing argument always has an animate interpretation (Mendikoetxea 1999, Sánchez 2002, and references therein).Consider, for instance, change of state verbs.This kind of verbs typically allow animate agents as well as non-animate causes as subjects in their transitive version (13a).When they enter into a non paradigmatic SE-construction, the cause looses the ability to refer to an inanimate entity.
El fuego/pirómano quemó muchos libros/a muchos inocentes aquella noche The fire/pyromaniac burned many books/ A many innocents that night 'The fire/the pyromaniac burned many books/many innocent people that night' b.
Se quemaron muchos libros aquella noche SE burned-3pl many books that night 'Many books were burned that night" (animate generic agent only) c.
Se quemó a muchos inocentes aquella noche SE burned-3sg many innocents that night 'Many innocent people were burned that night' (animate generic/arbitrary agent only) This restriction equally affects passive (13b) and impersonal (13c) constructions.This property has, at least, two important consequences: On the one hand, it sharply distinguishes "passive" SE from analytic passives, where inanimate causers are perfectly fine ( 14). ( 14) Muchos libros fueron quemados/destruídos por el fuego aquella noche Many books were burnt/destroyed by the fire that night On the other hand, the animacy restriction makes SE constructions incompatible with verbs that do not select animate arguments (e.g., ocurrir 'to happen', transcurrir 'to pass', rielar 'to shimmer (the moon)' ( 15)), making the distribution of SE closer to the one of strong pronouns.
*se/*él transcurrió (un día) SE/he passed (one day) Many impersonal constructions share this property with SE-constructions (Fernández-Soriano & Taboas 1999).As a matter of fact, it has been argued that this requirement is a general property of arbitrary/generic subjects (e.g.Cinque 1988).However, this statement is not completely accurate.It is true that arbitrary/generic subjects tend to be animate in many such contexts,6 but when verbs that generally take inanimate subjects are considered, the generic interpretation is also available for many inanimate implicit arguments, as illustrated in the contrast between (16a) and (16b).
Al PRO ser expresamente recopilados para la enciclopedia, los datos adquieren una importancia especial 'Being specifically collected for the encyclopedia, data acquire a special importance'.b.
#Cuando se es expresamente recopilado para la enciclopedia, se adquiere una impotancia especial When SE is collected specifically for the encyclopedia... Intended: 'When something is collected for the encyclopedia, it acquires a special importance.'It does not matter how easy to obtain an inanimate lecture is, the result is completely ungrammatical with non paradigmatic SE constructions (17). 7(17) a.
#Al rotarse de manera regular, sabemos que el peso está equilibrado When rotating SE in a regular fashion, we know the weight is balanced b.
#Que se tenga demasiada resolución conlleva gastar demasiada energía That SE having too much resolution implies to spend too much energy 'Having too much resolution implies heavy energy consumption' c.
#En esta pajarería se come mucho alpiste In this pet.shopSE eat.3sg a.lot canary.grass"In this pet shop a lot of canary grass is eaten" (only possible if it is eaten by people) d.
# Aqui, si se es rugoso, se sobra (=si la tela es rugosa, sobra) Here, if SE is rough, SE is left over (if the fabric is rough, it is left over As suggested above, the distribution of the facts makes SE close to strong pronouns in Spanish.In fact, the animacy restriction systematically puts together impersonal constructions with 2nd or 3rd person agreement (Fernández-Soriano & Taboas 1999), and se in SE constructions, differentiating it from impersonal silent/implicit arguments in infinitival constructions (16a) or middle sentences (see sec.3.4).That strongly suggests that the animate/non-animate distinction in impersonal sentences falls squarely within Montalbetti's generalization, which establishes that overt pronouns entering into agreement relations obligatorily have animate interpretation.Thus, for instance, in subject position (18a,b), the pronoun ella can only refer to her, but as a P complement it can both refer to her and to (feminine) it (18c).( 18 Similarly, what the distribution of facts in this sections indicates is that impersonals with overt first and second or third person pronouns and, crucially for us, se in SE constructions also fall within the same group inducing Montalbetti animacy effects; as opposed to silent pronominal arguments, which do not.In section 3.1 we will come back to a possible syntactic explanation for Montalbetti's generalization; for the time being, the important point is that if SE is an overt subject pronoun, as we propose, the fact that its interpretation must be animate can be assimilated to the parallel behavior of other pronominal elements.En ese país las instituciones censuran las iniciativas políticas In this country the SE institutions censor.pl the iniciative political 'In this country, the institutions censor the political initiatives' (cfr.las iniciativas SON censurADAS, 'initiatives are censored') Furthermore, this construction is compatible with any kind of verbal complement, crucially including those that reject analytic passive constructions.For instance, complements of the verb tener ('to have') in ( 21) and measurement complements such as ( 22) and ( 23) cannot passivize, but they may appear in "passive" se constructions with no restriction (but see Ordóñez & Treviño 2011) En este país, el opositor es asesinado In this country, the opposing is assassinated 'In this country, dissidents are assassinated'

SE constructions have active morphology and active syntax
One could argue that passive SE constructions are "a different kind of passivization", but a transformation not subject to any constraint would be almost a unique situation, specially when we consider cases such as ( 22)-( 23), where it is not even clear that accusative Case can be assigned,9 and, in consequence, it is hard to see how a voice transformation may proceed, how it may be triggered.Moreover, as shown in (19a), and in contrast to (24a), in se constructions DOM is assigned as in regular active transitive sentences (see also section 2.1, and Rivero 2002).Altogether, this makes the "passive" hypothesis of se constructions with transitive inanimate objects implausible.Quite to the contrary, SE constructions are regular active sentences with SE in subject position, as proposed.Nothing exceptional happens at the VP-internal level except for the wellknown agreement patterns to which we will return in section 3.
2.6.SE must be in a Nominative Case position: the case of infinitives It has also been observed that SE is incompatible with those infinitives that do not license nominative subjects, but allowed in infinitival constructions where overt subjects are licensed (Cinque 1988;Mendikoetxea 1999).The contrast between ( 25) and ( 26) present minimal pairs with infinitival and tensed subjunctive complements, and ( 27) is a case of prepositional infinitives allowing overt pronominal subjects.In all cases, SE and personal pronouns behave the same way: (25) a.
* This contrast fits particularly well in our proposal, provided that the overt pronoun se is a regular pronominal subject that shows the properties and distribution of any other overt subject in the language, including nominative Case (Oca 1914).

Pseudo-relatives
The behavior of SE-construction with pseudo-relatives also supports our analysis.As the baseline structure for our argument notice first that pseudo-relatives are not allowed in subject position in Spanish, neither in active (28b) nor in passive (28c) constructions (Aldama 2016: eexx. ( 58)-( 59)): (28) a.
Los reporteros fotografiaron a Superman que aterrizaba sobre un rascacielos The journalists took.pictures.ofa Superman that landed on a skyscraper 'The journalists took pictures of Superman landing on a skyscraper' b. ?? Superman que aterriza sobre un rascacielos es una estampa para no perderse Superman that lands on a skyscraper is a sight to not miss 'Superman landing on a skyscraper is a sight not to miss' c. ?? Superman que aterrizaba sobre un rascacielos fue fotografiado por la multitud Superman that was.landing on a skyscraper was photographed by the crowd 'Superman landing on a skyscraper was photographed by the crowd' As Aldama observes, the sharp marginality of the passive construction is directly related to the presence of the pseudo-relative: if it is eliminated the sentence is perfectly acceptable.Moreover, even the minimally differing (29b), where the pseudo-relative is substituted by a non-restrictive relative clause is acceptable: (29) a. Superman fue fotografiado por la multitud Superman was photographed by the crowd b.
Superman, que aterrizaba sobre un rascacielos, fue fotografiado por la multitud Superman that was.landing on a skyscraper was photographed by the crowd 'Superman, who was landing on a skyscraper, was photographed by the crowd' In clear contrast, pseudo-relative phrases can appear with the overt argument of passive SE constructions: (30) a.
Se veían los trenes que llegaban a cocheras SE saw.pl the trains that arrived to sheds 'The trains could be seen arriving to sheds' b.
Se fotografiaron los pájaros que pasaban hacia el norte SE took.pictures.of.pl the birds that passed.3pltoward the North 'Pictures of the birds were taken heading towards the North' The contrast between (28b-c) and ( 30) indicates that the VP-complement --los trenes in (30)--is not the subject and that passive SE constructions are not passive at all.Contrary to what agreement facts might suggest, the complement of the verb is not in subject, but in object position.Again, this is what we expect if SE is the actual subject.

Summary
The properties discussed through this section support a derivation of impersonal and passive SE constructions in which SE itself is the "missing" argument and raises to subject position entering into a regular nominative Case relation with T. That fits nicely with most of the characteristic properties discussed in this section: (i) the obligatory presence of an argument; (ii) its control/obviation properties also supporting the presence of a syntactically active argument; (iii) the animacy requirement as naturally derived from SE, a pronoun, being in subject position, subject to Montalbetti's generalization; (iv) the active morphology and syntax of both impersonal and passive constructions, (v) the regular distribution of the pronoun SE in terms of Case-requirements, and (vi) the possibility of pseudorelatives modifying the verbal complement in passive SE constructions, showing that it is not the subject of the construction despite the number agreement facts.
The fact that when the right conditions are met (transitive verb with inanimate object), the verbs agrees in number with the object, is the most evident divergence with the general picture we have presented.However, given the lack of number features in SE, this fact is still compatible with our proposal if SE is treated as a quirky subject.In the next section we deal with the basic morphosyntactic facts of our proposal: the type of features it encodes, how it raises to subject position, etc.We also provide additional evidence that supports our analysis and deal with some potential problems.

The role of the defective argument
The idea that SE is the missing argument required for SE constructions is not a new one (see Cinque 1988, Mendikoetxea 1999, D'Alessandro 2007, and references therein).The proposal that SE raises to subject position is not new either (see Oca 1914), but it has been discarded by most authors mainly for two reasons.First, the fact that it is a clitic.In G&B approaches, this essentially means that it cannot raise to subject position (but see Fr. il ('he')); therefore, SE was condemned to attach to a head, (e.g.Cinque 1988).And second, the agreement facts -interpreted as indicating that the object receives nominative-do not conform to the "nominative SE" hypothesis straightforwardly.
However, the battery of properties discussed in section 2, including agreement facts, are straightforwardly explained if SE is mapped to an θ-position and moved to (Spec, TP), where it receives nominative case. 10In addition, G&B theoretical objections to clitic movement are no longer tenable.In this section we develop some ideas concerning the nature of SE and explore some of their consequences.The result will be a structure where SE is a regular nominative pronoun encoding animacy but not person.

SE as a nominative pronoun
In section 2.4 it was observed that the missing argument is necessarily interpreted as animate.Cinque assumes that it is so because animacy is a "characteristic" of the class of generic, arbitrary NPs (Cinque, 1988: 536); however, as we have shown animacy restriction with SE must be subsumed under Montalbetti's Generalization (MG).What is interesting about that generalization is that pronouns are not inherently animate; they are so only when they enter into an agreement relation, as the contrast in (18), repeated in (31)  On the other hand, the fact that the complement of the verb agrees in number with the verb does not imply it is the subject; not only for all the reasons provided in section 2, but also because under certain circumstances -again, as in other quirky Case configurations-it fails to trigger agreement.Consider the sentence in (34).
(34) se les envió (a los enfermos) todas las medicinas que necesitaban SE 3pD sent.3s(DOM the sick.people)all the medicines that needed.theySick people were sent all the medicines they needed What is interesting about (34) is the fact that Tense does not agree with any of the internal objects, neither with the goal (sick people), nor with the theme (all the medicines).The verb appears in a default 3rd singular despite the fact that the DO is plural.If the theme were the subject it would obligatorily trigger agreement in (34), as it does in analytic passives (35).
(35) las medicinas les fueron enviadas (a los enfermos) the medicines 3pD were sent (DOM the sick.people)'Sick people were sent medicines' Again, this is a strong indication that the issue in passive SE constructions is not about being transitive, not even about the [-DOM] object to require some feature checked, but about T's ability to check its number agreement, left unchecked by its subject SE, with some proximate element, not necessarily the theme (see also sec.2.5).In (34), the subject position is filled by SE and number agreement is still available.The theme is of the right type to check that agreement, but the goal, being closer to T, blocks that relation.Being a transitive construction, the theme is in its regular position and does not need to agree with T to check its features.Consequently, no agreement is realized, but the sentence is grammatical (see, for instance, D'Alessandro 2003; Sigurðsson & Holmberg 2008; and references therein for a similar situation in other quirky subject configurations, and experiencer intervention in Torrego 1996).
In sum, we have proposed that SE is generated in a theta position and moved to (Spec, TP).As a consequence, it behaves as any other overt pronoun and is subject to MG.On the other hand, since it lacks number features, T number may --but does not need to--be checked by other local arguments, if available, a situation common to other quirky subject configurations.

Raising to subject
If, as we have argued, SE moves to subject position as any other pronoun, it is expected also to be allowed in raising to subject constructions, moving from an embedded context to the matrix subject.It has been claimed in the literature that this prediction is not borne out (see Sánchez 2002: 45 and references therein).In that respect, it is worth observing that Martins & Nunes (2016) note a dialectal split between Brazilian and European Portuguese with respect to raising, and there are many examples in Spanish accepted as grammatical by Spanish speakers. 12.In sentences from (36) to (39) SE raises, as 12 Very plausibly, the relative marginality of this construction and what looks more as an idiolectal variation rather than a clear dialectal difference, are connected to the relative marginality of 1st and 2nd person argument raising in the same contexts : (i) a.
Juan parece haber sufrido mucho Juan seems.3sghave suffered a lot 'Juan seems to have suffered a lot' b. (??) (Vosotros) parecéis haber sufrido mucho You seem.2pl have suffered a lot 'You guys seem to have suffered a lot' c. (??) (Nosotros) parecemos haber sufrido mucho [cfr."(nosotros) parecemos tontos"] We seem.1pl have suffered a lot 'We seem to have suffered a lot' a regular DP, to matrix subject position.In the b-pairs it is shown that SE comes from the embedded sentence. 13  (36) a.
se parecía haber perdido los sentidos SE seemed.3sShave lost the senses 'Senses seemed to have been lost' b.
se parecía haber ganado en altura SE seemed have gained in altitude 'Altitude seemed to have been gained' b.
*se parecía que (se) había ganado en altura SE seemed that SE have been in altitude (38) a.
Con ello, se parece abogar por dotar a la percepción de una unidad ... With it, SE seem.3sSadvocate for provide to the perception of a unity... Lit.: 'With that, providing perception with a unity seemed to be advocated for' Justo cuando parecía que todo le iba a ir bien, que se parecía haber liberado Although we do not have an explanation for the marginality of either (ib-c) or some of the se-examples with raising, their parallelism is quite suggestive, given our discussion of the animacy effects in sections 2.4 and 3.1. 13Given that we are aware that this point is a quite controversial one, we have gathered quite a few examples from CORPES and from a Google search, a few of which are reproduced in the text and in this note.
Just when seemed.3s that all to.him was to go well, that SE seemed have liberated de su sambenitos psicosomáticos from his stigmata psychosomatic 'Just when everything seemed to go wel for him, that he seemed to have gotten rid of his psychosomatic stigmata... ' (http://www.enclavedecine.com/2013/01/midseason-2013vuelven-californication-y-girls.html 15-3-2016) However, data in (41) seems to be restricted to unaccusative SE constructions, or constructions where SE has an aspectual value, where crucially the clitic does not stand for an argument (see Grant 2013 and references therein).In contrast, clitic climbing is not available for cases of paradigmatic SE with argumental import (e.g.reflexives) (42) nor for other argumental clitics ( 43)-( 44).( 42 This contrast suggests that raising is available for cases in which the clitic has V-funcional import, possibly an aspectual one, but it is not available when it represents an argument relation, either as an agreement head or as an argument.In that case, SE raises to subject position in (36)-(39) a-pairs.Data on this subsection, although admittedly marginal (see fn. 12, and 13), strongly supports the idea that SE is actually a nominative pronoun.

Secondary predication
An additional argument that supports our analysis is found in secondary predication.46)-( 47), notice that other examples may also be constructed where no such factor seems to be playing a decisive role: (48) a. Allí, todos los años se celebran las fiestas vestidos con trajes típicos There, every the years SE celebrate.plthe festivities dressed.masc.pl.with vests typical 'There, they (gen.)celebrate their festivities every year dressed in regional costums' b.
En mi tiempo se abordaban concentrado los problemas In my time SE address.past.plurconcentrated.masc.sg. the problems 'When I was young, you would address problems concentrated' c.
en esta empresa se trabaja siempre, se trabaja dormido y se trabaja despierto in this company SE works always, SE works asleep and SE works awake 'In this company you always work, you work asleep, and you work awake.' Moreover, SE constructions clearly contrast with analytical passives, where secondary predication on the implicit argument is impossible even when the sentence is tempered or embedded in some modal contexts: (49) a. * ¡María no es besada borracho!/ ¡No es hecho el examen borracho!Maria not is kissed drunk.masc/ not is done the exam drunk b. * Allí, todos los años las fiestas son celebradas vestidos con trajes típicos There, every the years the festivities are celebrated dressed.masc.pl.with vests typical In this section we have shown that in both impersonal and passive constructions, it is possible to have a secondary predicate on SE, showing that the argument is syntactically present.

Middles
Further evidence for our proposal comes from comparing passive SE with middle SE constructions.One of the key properties of middle constructions is word order.In contrast to passive SE construction, where the preferred word order is V -Theme,15 in middle constructions the theme must precede the verb (Mendikoetxea 1999).
Esta madera se corta fácilmente This wood SE cut easily 'This wood cuts easily' b. # Se corta esta madera fácilmente SE cut this wood easily Although a full justification is beyond the scope of this paper, the reasons to assume that the clause initial DP occupies the subject position in the middle construction in (50a) are quite compelling. 16In that case, if we extended our proposal to analyze SE in middle constructions as the subject, both esta madera and SE would be satisfying the same syntactic relation, an unwelcome result.However, there is an important difference between middle and impersonal/passive SE constructions that suggests that the clitic has a very different nature in each of them: sharply contrasting with impersonal and passive SE constructions, there is no animacy requirement in the case of middle constructions.Mendikoetxea (1999) explicitly observes that middle sentences "describe a state that requires the participation of a cause or an agent [... but passive and impersonal SE constructions] describe an action or an activity which requires necessarily and forcefully the intervention of an intentional agent…" (p.1643).Following our analysis, animacy is a property of agreement, and, in absence of person features, it has to be satisfied by SE (see sec.3.1).Given the lack of animacy effects in middle constructions SE cannot be the overt subject pronoun.Our proposal instead is that it should be treated as a voice morpheme.Evidence comes from three different sources.First, although it is not the preferred pattern, in middle constructions animate preverbal DPs may appear without DOM, triggering subject number agreement (examples from M. Zorraquino, apud Sánchez 2002: 66): 17 (51) a.
Los maridos no se encuentran fácilmente The husbands no SE found.3peasily 'Husbands are not found easily' b.
Las mujeres no se conocen nunca bien The women no SE know.3pnever well 'Women are never known well' Examples in (51) show that the diathesis of the verb is somehow affected in middles (optionally) blocking object Case assignment, as in the case of inchoative constructions.This option is not available for transitive impersonal constructions, suggesting that the middle construction is not active (see sec.2.5).
A second piece of evidence comes from the fact that in many cases SE is not required to form a middle construction.For instance, (50a) is not ungrammatical if SE is suppressed (52a).This option is in general available for those verbs that may have a non-agentive external argument. 18  (52) a.
las cuerdas de la guitarra tensan fácilmente 'Guitar strings tauten easily' f. estos palos rompen fácilmente 16 Notice that our concern here is not about the much debated issue of which specific position preverbal subjects occupy in Spanish.Our point is that whatever that position is, it is SE in "passive" SE constructions, but the argument DP in middle constructions that occupies that position. 17Sentences in (51) contrast with those in (i), so called impersonal middle constructions, with the same meaning, but retaining DOM, and lacking number agreement: (i) a los maridos no se les encuentra fácilmente dom the husbands no SE 3pD find.3seasily (ii) a las mujeres no se las concoce nunca bien dom the women no SE 3pO know.3snever well 18 However, it may also appear with verbs as vender 'to sell'.
threw.away.sg the cigarette butts on.thefloor'Thecigarette butts were thrown away on the the floor' This description does not honor dialectal variation, but it corresponds to the standard description of the phenomenon, and it accurately describes the dialects of the authors (Central and Basque Spanish).On the other hand, when an animate object does not receive DOM, it must agree (7) (but seeOrdóñez & Treviño 2016for some observations in Mexican Spanish).
Ormazabal & Romero 2013)te that what is relevant for agreement is DOM, not the properties of the DP .So, in those cases where DOM is assigned to secondary predicates (6a,b) or infinitival subjects (6c) (seeOrmazabal & Romero 2013), the verb cannot express number agreement.
(Rivero 2002)view of voice morphology, both SE impersonal and SE passive constructions are active sentences(Rivero 2002).As shown in (19a)-(20a), the verb has exactly the same form as in their regular active pairs (19b)-(20b): PRO Tener demasiada resolución conlleva gastar demasiada energía To.have too.many resolution implies to.consume too.manyenergyd.La teoría ontológica nos dice que si ocurre, PRO debe existir.The theory ontological 1pD says that if happens, PRO must exist These examples are construed with verbs selecting inanimate or mostly inanimate arguments.In (17a)-(ia), the verb rotar, 'rotate', takes usually a very specific set of arguments: those characterized for having a regular geometry, what makes them capable of doing a regular spin on an axis.Since the argument is so restricted, it is easy to make generic statements with inanimate interpretation.This fact is even more clearly shown by ocurrir in (id), a predicate that cannot take animate subjects.From .
It thus seems reasonable to assume that the locus of MG is agreement.In other words, we propose that the agreement properties of T in Spanish specify not only person and number, but also animacy.The nominative pronominal paradigm is thus morphologically realized according to (32).That SE lacks number agreement features altogether is a common place in the literature.This leaves open the possibility for number features in T to be checked by other arguments, as it is regularly the case in other quirky configurations.
Mendikoetxea 1999;02;eement in Tense specifies the following features: person (and maybe also the features required to distinguish 1/2/3), animacy, specificity/EPP, and number (seeHarley & Ritter 2002; D'Alessandro & Alexiadou 2003). Thre, when T expresses animacy, but not any specific person, it is represented by SE.Note that the subject pronoun SE has no inherently specified person features, hence it is not semantically confined to the third person.As observed byOca (1914)(see also, for instance,Mendikoetxea 1999; D'Alessandro & Alexiadou 2003), SE may be interpreted as a first person (33a), a second person (33b), an indefinite (33c), or a generic (33d).
SE seemed.3sSthatSEadvocate.pastforprovideto the perception of a unity... (39) Las dos adolescentes desencajaban...The two adolescents un-fitted.3rd.pl.a.... entre la angustia y la impaciencia que se parecía respirar en la casa among the anguish and the impatience that SE seemed.singto.breathe in the house 'The two teenagers didn't fit into the anguish and the impatience that one seemed to breathe in that house'[adapted from CORPES, 17/11/2016] b. * ... entre la angustia y la impaciencia que se parecía que (se) respiraba(n)among the anguish and the impatience that SE seemed.singthatSE breathe.sg/pl[Arce,Alicia: «En las buenas y en las malas».Núñez, Agustín (coordinador) Radioteatro.Asunción: Arandurã, 2008.From CORPES, consulta 17/11/2016] Moreover, if we consider relative judgments, SE-raising is considerably better than analytical passives, which are absolutely impossible in these contexts (40).... entre la angustia y la impaciencia que eran parecidas respirar among the anguish and the impatience that was seemed.singto.breathe.sg/plItcould be argued that in these contexts SE is not actually moving to subject position, but that it is just clitic-climbing to parecer by head to head movement at PF.That seems to be the case of unaccusative SE in examples like (41): * Saab (2014)e ungrammaticality of examples like (45), the general claim in the literature is that secondary predication on the "implicit" argument is not possible in SE constructions (example slightly modified fromSaab (2014): ex.(29); also see Pujalte & Saab (2012); although see Mendikoetxea 2008): María drunk.masc'You cannot kiss Mary drunk!(whoever does it, he will appear in her black list)' b.Se puede ir a clase/conducir/besar a María borracho SE may.sg go to class/drive/kiss A Mary drunk.masc'You may go to class/drive/kiss Mary drunk' c.A la política se viene llorado de casa To the politics SE come.3sgcried from home Lit.'To politics one comes cried from home' [intended: 'In politics, you do not complain when deals do not fit your wishes'; A.Rivera to Rajoy (Spanish president)] d.Si se nace pobre, lo más probable es que se muera miserable 14 If SE is.born pour, the most probable is that SE die.subjmiserable 'If you are born poor, it is most likely that you will die miserable' As observed to us by Andrés Saab (p.c.), modality is playing an important role in the grammaticality of these examples in a direction that is not totally clear to us (see Pujalte & Saab 2012 and D'Alessandro 2007 for interesting observations on the different semantic interpretation of SE constructions and their syntactic correlations).Without denying the role modal operators and other elements may play in ( ¡No se besa a María borracho!(Quien lo hace entra en su lista negra) not SE kiss.sgA