SE IN SPANISH: PROPERTIES, STRUCTURES, ANALYSES*

Antonio Fábregas
University of Tromsø-Norway's Arctic University

ABSTRACT. This article provides an overview of Spanish SE, covering the main empirical facts,
analyses and theoretical issues that it raises, and exploring the prospects to unify all uses –which
are over a dozen in the literature– under one same object that keeps its surface properties
invariable. We will show that it is almost inescapable to propose that SE is associated to two
types of objects, defective arguments and defective verbal heads; both objects share the property
of not introducing referentially independent DPs, which can be argued to be the result of the
grammaticalisation of a reflexive element in contemporary Spanish. The chapter proposes that a
treatment of SE as a projection introduced high in the clausal structure and acting as an agreement
locus can be a fruitful way to unify all uses of SE.

Keywords. SE; reflexive SE; impersonal SE; passive SE; aspectual SE; inherent SE;
anticausative SE; reciprocal; middle SE; spurious SE; agentive SE; transitive SE; antipassive
SE; factitive SE

RESUMEN. Este artículo proporciona una visión general del SE en español, cubriendo los
principales hechos empíricos, análisis y cuestiones téoricas que produce, y explorando la
posibilidad de unificar todos los usos de SE –que son más de una docena en la bibliografía– bajo
un solo elemento que mantenga propiedades superficiales invariables. Mostraremos que es casi
inevitable proponer que SE se asocia a dos objetos de distinto tipo: argumentos defectivos ny
núcleos verbales defectivos, donde la defectividad se refiere a que ninguno de esos objetos
introduce SD referenciales, algo que puede entenderse como el resultado de la gramaticalización
del reflexivo latino en español actual. El capítulo propone que un análisis del SE como una
proyección introducida en una posición alta de la estructura oracional y que funciona como un
núcleo de concordancia puede ser un modo de unificar todos los usos de SE.

Palabras clave. SE; SE reflexivo; SE impersonal; SE pasivo; SE aspectual; SE inherente; SE

anticausativo; SE recíproco; SE medio; SE espurio; SE agentivo; SE transitivo; SE antipasivo;
SE factitivo

1. Introduction: one clitic, many roles

Within the grammar of clitic pronouns, SE is the one that has a more complex distribution
and the one whose grammatical properties are, by far, most disputed (see, among many others,
Contreras 1966, 1974; Schroten 1979, Martín Zorraquino 1979, Martínez 1981, Cano Aguilar
1981, Fernández Ramírez 1986, Iglesias Bango 1991, Gómez Torrego 1992, Mendikoetxea
1992, 1999; Rodríguez Díez 1996, Otero 1999, 2002, Sánchez López 2002, RAE & ASALE
2009: §41, to name just a small number of basic references). The complexity of this element is
reflected, to begin with, on the many issues that it has in the grammar of Spanish. (1) illustrates
a few of the basic constructions that have been proposed in the literature.

*
I am grateful to two anonymous reviewers, linguistic twitter, Margot Vivanco and in particular Luis García
Fernández for comments, observations and criticisms that have been crucial in building up the version that is
being published. All disclaimers apply.

ã Antonio Fábregas. Borealis: An International Journal of Hispanic Linguistics, 2021, 10 / 2. pp. 1-235.
http://dx.doi.org/10.7557/1.10.2.5934

This is an Open Access Article distributed under the terms of the Creative Commons Attribution License
(https://creativecommons.org/licenses/by/4.0/legalcode) which permits unrestricted use, distribution, and reproduction in any
medium, provided the original work is properly cited.
 ANTONIO FÁBREGAS

(1) a. Pedro se mira en el espejo.

Pedro SE looks in the mirror
'Pedro sees himself in the mirror'
b. Pedro y María se miran en el espejo.
Pedro and María SE look in the mirror
'Pedro and María see each other in the mirror'
c. Luis se enfadó.
Luis SE got.angry
'Luis got angry'
d. Luis se cayó.
Luis SE fell
'Luis fell down'
e. Luis se arrepintió de sus actos.
Luis SE repented of his acts
'Luis repented of this actions'
f. Se venden casas.
SE sell houses
'Houses are sold'
g. Se recibió a los invitados.
SE received DOM the guests
'Someone received the guests'
h. Se lo dimos a mis padres.
SE it.acc gave to my parents
'We gave it to my parents'

In (1) above, only a subset of the uses of SE documented in the literature are presented. The
uses illustrated in (1a-e) are traditionally described as instances of 'paradigmatic' SE (RAE &
ASALE 2009: §41.10a), where 'paradigmatic' is defined as (2).

(2) In a paradigmatic use of SE it is possible to substitute SE with person-marked pronouns.

As such, (1a-e) exhibit SE, in principle, because the subject is a third person singular or
plural DP. If the subject is replaced by a first person plural, for instance, SE is replaced.

(3) a. Nos miramos en el espejo.

us look in the mirror
'We see ourselves in the mirror'
'We see each other in the mirror'
b. Nos enfadamos.
us got.angry
'We got angry'
c. Nos caímos.
us fell
'We fell down'
d. Nos arrepintimos de nuestros actos.
us repented of our acts
'We repented of our actions'

The remaining cases in (1) are called 'non-paradigmatic SE cases', where in principle it is
not possible to make SE alternate with personal pronouns. However, note that also in these

2
 SE IN SPANISH

cases a third person is involved and 1st and 2nd persons are excluded: these are simply cases
where bona fide independent reasons force the third person in the context.
Among the paradigmatic SE cases, (1a) is considered a reflexive structure, as witnessed by
the fact that the pronominal double sí mismo 'SE self' can be added to it without change in
meaning –although we will see below that this is not always the case–. The general description
of such cases is that the subject and the object are coreferential.

(4) Pedro se miró a sí mismo en el espejo.

Pedro SE lokked DOM SE self in the mirror
'Pedro looked himself in the mirror'

(1b) is considered a reciprocal sentence in the interpretation where each one of the two
individuals performs the event on the other one, that is, Pedro sees Mary in the mirror and vice
versa, or in other words they see each other. This reading, intuitively related somehow to the
reflexive one, generally allows the addition of el uno P el otro.

(5) Pedro y María se miran el uno al otro.

Pedro and María SE see the one DOM-the other
'Pedro and María see each other'

As for (1c), this one is considered an anticausative SE structure, where the same verb must
allow two versions, one causative and one non-causative, and the SE one marks the latter. This
structure has some 'passive' flavour that has prompted traditional Spanish grammar to use the
term 'middle' (medio), as we will see. However, unlike passives exclude external causers that
set the event in motion; in some sense, the idea is that the subject has the change of state
initiated internally, without the external intervention of an agent or causer.

(6) a. Luis enfadó a Pedro.

Luis angered DOM Pedro
'Luis made Pedro angry'
b. Pedro se enfadó.
Pedro SE got.angry
'Pedro got angry'

The SE case in (1d) is much more controversial to classify. In the last 20 years or so, it has
received the name of 'aspectual SE', although we will see that this label has been very disputed,
in particular for intransitive verbs like the one in (1d). The crucial property is that the verbal
base has a version without SE that, unlike (1c), does not contrast in its causative or non-
causative component, but on the lexical aspect of the predicate, with the SE version related to
a compulsorily telic reading where there is some emphasis on the result that is acquired.

(7) a. Los precios cayeron durante meses.

the prices fell for months
b. Luis se cayó.
Luis SE fell
'Luis fell down'

(1e) is considered an inherent SE, a 'default' term that captures the fact that the verb cannot
be built without this pronoun that on the surface seems to be a reflexive one.

3
 ANTONIO FÁBREGAS

(8) *Pedro arrepintió de sus actos.

Pedro repented of his acts

In the non-paradigmatic SE cases, (1f) represents a passive SE structure. In contrast to (1c),

(1f) involves a reading where there is an external trigger of the event –someone that sells the
houses– and is paraphrasable –although with relevant semantic and syntactic differences that
we will see below, §8.3– with a periphrastic passive.

(9) #Las casas son vendidas.

the houses are sold

This SE is non-paradigmatic in the sense that it appears in a context where the reflexive
pronoun must be third person and there is no alternation with other persons, in essence because
this type of passive is –unlike the periphrastic passive– restricted to noun phrase subjects, for
reasons that are controversial but relatively robust across varieties and speakers.

(10) a. Se contrataron {profesores / *ellos}.

SE hired teachers / they
b. #Me contraté yo.
me hired I
Intended: 'I was hired'
c. Fui contratado yo.
was hired I
'I was hired'

(1g) is an instance of impersonal SE, a construction where SE marks that the subject is
existential or generic but the object can be present with its case marking assigned as in any
other active sentence. Again, this SE is non-paradigmatic because the subject is forcefully third
person in such cases, and Spanish in fact blocks an interpretation that includes the speaker
(11a). This structure allows pronominal arguments, in contrast with the passive one (11b).

(11) a. *Se contrató a los profesores entre el decano y yo.

SE hired DOM the teachers between the dean and I
Intended: 'The dean and I hired the teachers'
b. Se me contrató a mí.
SE me hired DOM me
'I was hired'

Finally, (1h) is an instance of so-called spurious SE, where the form replaces a bona fide
third person dative pronoun when the accusative clitic is present. Again, this SE is non-
paradigmatic exclusively in the sense that this replacement only happens when the pronoun is
third person, singular or plural.

(12) a. Se lo dimos a mis padres.

SE it.acc gave to my parents
b. Les dimos el regalo a mis padres.
them.dat gave the present to my parents
(13) a. Me dieron el regalo.
me.dat gave the present
'They gave me the present'

4
 SE IN SPANISH

b. Me lo dieron.
me.dat it.acc gave
'They gave it to me'
c. *Se lo dieron.
SE it.acc gave
Intended: 'They gave it to me'

The study of SE is not just complicated by the existence of these constructions, which
already tells us that we have an element which should almost behave as a wild card that can
appear in many different syntactic contexts and with different functions that, as we will see,
are not always clearly displayed even in the clearest cases. The problems in the study of SE are
complicated, beyond this, by three basic facts.

a) The literature does not agree on all judgements, even at the most basic empirical level.
Sánchez López (2002: 26) consideres (14) grammatical, while Ordóñez (2021) considers such
impersonal episodic sentences built over periphrastic passives ungrammatical.

(14) Se fue asesinado sin piedad.

SE was murdered without mercy
'Someone murdered some specific person without mercy'

As we will see, there are equally judgement problems with the impersonal SE constructions
in non-finite contexts (§7), with the interpretation of anticausative SE structures and which
elements force the interpretation (§4) and many other cases. In some instances, there are well-
known dialectal distinctions that simply tell us that the grammaticality of one sentence depends
on the variety spoken by the informant –such as the case of impersonal vs. passive SE structures
with inanimate internal arguments, (15)–, but in cases like the one in (14) there are no known
variation facts that explain why some speakers consider it grammatical and others feel it as
radically ungrammatical.

(15) a. Se vende pisos.

SE sells houses
b. Se venden pisos
SE sell.pl houses

b) There is no agreement about how many distinct classes of structures should be

distinguished with SE. The examples in (1) perhaps reflect the most usual classes discussed in
the grammars, but there is no agreement with respect to (i) how many other classes might be
necessary to posit, (ii) whether some of the classes in (1) can be conflated and (iii) how
individual sentences should be classified in one of the groups. From this perspective, aspectual
SE structures have been viewed as particularly conflictive (§5). Some authors treat as aspectual
sentences like (16a), (16b) and (16c).

(16) a. Juan se cayó.

Juan SE fell
'Juan fell down'
b. Juan se leyó todas las lecciones.
Juan SE read all the lessons
'Juan read up all the lessons'

5
 ANTONIO FÁBREGAS

c. Juan se fue.
Juan SE went
'Juan left'

Other authors, to be discussed in due course, only consider (16a) aspectual, while (16b)
should be viewed as an agent-oriented SE and (16c) should be rather considered a SE that
modifies the argument structure of the verb and whose aspectual role is less defined. To
complicate matters more, some authors have argued that the anticausative SE in fact also has
an aspectual role, possibly leading to conflating aspectual SE structures with sentences like
(1c), where (1c) should be viewed just as the subcase of aspectual SE where the predicate could
be causative.
As we will see in §5 and other sections, aspectual SE is very controversial, and part of the
controversy refers to the problem of whether, in sentences like (16b) above, one should give
more relevance to the aspectual information that SE seems to make compulsory –roughly,
telicity– or to the involvement of the subject as an affected entity in the process, that in some
way gets positively or negatively affected by the event. This affectedness interpretation is
related in some instances with notions of intensity in performing the action, as pointed out by
Fernández Ramírez (1986). In general, this affectedness interpretation that some authors use as
a label to what others call 'aspectual SE' with transitive predicates has traditionally been treated
as a dative reflexive with a benefactive meaning, an ethical dative or a superfluous dative that
happens to be coreferential with the agent. Its use has also been identified in structures like
(17) below, that some authors have treated as transitive anticausatives (see §4.4) below.

(17) Juan se cortó el dedo.

Juan SE cut the finger
'Juan accidentally cut his finger'

Similar problems and complications are triggered by sentences like (18), which have a
passive flavour but, additionally, express generic claims that are not episodically anchored to
specific time periods and worlds. These are, by no means, the only conflictive cases, as we will
see.

(18) Esta clase de libro se vende bien.

this class of book SE sells well
'This type of book sells well'

The confusion is increased by several proposals that posit additional classes of SE structures
that are not so broadly accepted. Here we will highlight five of such additional classes.

i) SE in middle structures, where middle is defined as a generic statement that expresses the
disposition of a type of entity to participate in an eventuality by virtue of its internal properties
(Lekakou 2005).

(19) Estas camisas se arrugan demasiado.

these shirts SE wrinkle too.much
'This type of shirt has a tendency to get wrinkles'

In (19) above we are not predicating from a specific shirt that it participates or has
participated in the event of getting wrinkles. Rather, we state that, because of their internal
properties, shirts like those have a predisposition to get wrinkles, even if they never got

6
 SE IN SPANISH

wrinkles (yet) and we are deducing that this should be their tendency because we know the
properties of the fabric they are made of. (19) has the syntactic structure of a passive SE, but
is different from standard passive SE structures in its genericity and in the preferred word order,
with a preverbal subject (Mendikoetxea 1999). Other languages would express this type of
dispositional statement with active verbs, like English.

(20) This type of shirt wrinkle too much.

ii) Figure reflexive SE (Wood 2013, 2014) is restricted to movement verbs, and in them the
reflexive pronoun forces the presence of a prepositional structure that expresses a relation
between the subject and a result location; the subject must end in the location defined by the
PP, and the subject must act agentively, trying consciously to get into that location.

(21) Juan se metió *(en la habitación).

Juan SE got into the room

Figure reflexive SE is similar to aspectual SE in that it emphasises some kind of result state,
this time a location reached by the subject, but differs from it in that aspectual SE –at least in
its standard definition– does not force the presence of a syntactic structure that expresses the
location. The term comes from Talmy's (2000) terminology, where in a locative prepositional
structure the object taken as reference to define the location is the ground and the entity that is
located by reference to it is called the figure: SE in such cases, by hypothesis, refers to the
figure, the subject of motion that ends up located in a space.
This type of SE in Spanish is difficult to identify, and beyond the aspectual SE connection
it also holds a connection with naturally reflexive verbs expressing motion (see §3), which have
been considered endo-reflexives (Haspelmath 1987) or autocausatives (Creissels 2006), such
as mover-se 'to move SE': with animate subjects, they express events that are triggered
volitionally by the subject, like figure reflexive verbs, and they involve events where the
movement applies to the same subject. Its status, as we will see, is not clear in Spanish.

iii) Antipassive SE refers to structures where the SE-less predicate introduces an internal
argument marked in accusative case and the presence of SE makes the internal argument
project as a prepositional complement –'complemento de régimen'–, typically with meaning
differences (Masullo 1992, Clements 2006). Typically, antipassive SE has effects on the lexical
meaning of the predicate, sometimes radically changing the theta role assigned to the internal
arguments, as in (23a,b) and (23c,d).

(22) a. confesar algo

to.confess something
b. confesarse de algo
to.confess-SE of something
(23) a. cuidar algo
take.care.of something
b. cuidar-se de algo
take.care.SE from something
'to protect oneself from something'
c. enfadar a alguien
to.anger DOM someone

7
 ANTONIO FÁBREGAS

d. enfadar-se con alguien

to.anger-SE with someone
'to get angry with someone'

In contrast to passive SE, this SE seemingly demotes the internal argument from a direct
object to an object that needs to be introduced by a preposition. Similarly to reflexive and
passive SE, one might argue that SE in (21b) has the role of removing the accusative marking
and therefore the remaining overt argument needs to be marked with a preposition. Similarly
to anticausative SE, in some cases (23b) the presence of SE has also the effect of turning the
verb from externally caused to internally caused: in (23a) the subject causes the feeling in the
object, while in (23b) the subject experiences the feeling, which is directed to the con-
argument.

iv) Agent reflexive clitic SE (Armstrong 2013) involves cases of SE which force the subject
to be interpreted as an agent, which performs the action wilfully, and excludes any type of non-
volitionality.

(24) El submarino se hundió tres barcos.

the submarine SE sank three boats
'The [people operating the] submarine wilfully sank three boats'

Like aspectual SE, agent reflexive SE requires the event to be telic, specifically an
accomplishment, but unlike it it imposes an agent-requisite to the subject that must have the
conscious intention of achieving the goal expressed by the telic event. In this sense, agent
reflexive SE contrasts with 'pure' aspectual SE structures like (25), where there is no entailment
of wilful intent.

(25) Juan se tomó un veneno sin querer.

Juan SE took some poison without wanting
'By accident, Juan took some poison'

v) Causative or factitive SE structures are structures where SE marks that the event was not
performed by the subject, but rather on the subject by an external agent, where the subject is in
any case the causer that makes the implicit agent perform the action on her or him.

(26) Luis se cortó el pelo.

Luis SE cut his hair
'Luis had a haircut'

In (26), in contrast to a reflexive structure, Luis is not the agent that gives himself a haircut,
but rather a causer that makes someone else give him the haircut.

c) Finally, beyond the disagreement on some judgements and the disagreement on how
many different classes of SE structures should be distinguished –with at least 15 different
constructions–, the third problem that complicates the study of SE is that there is some degree
of terminological dispersion that does not help navigate the vast quantity of literature that has
treated SE structures in Spanish. Unfortunately, different authors might use the same term to
refer to different SE structures, as it is the case for instance with 'middle SE', which some
authors use for sentences like (1c) and others use for sentences like (18), among many other
cases of terminological disagreement.

8
 SE IN SPANISH

The next table presents some of the terms used in the literature for the different SE classes,
concentrating on the most frequent cases. Note that 'reflexive' / 'reciprocal' are called in this
way in the literature, without variation.

Table 1. Labels for the main types of SE structures in the literature

Example Some of its labels Label used in this article
La tele se rompió Middle Anticausative
'The TV broke' Inchoative
Intrinsically reflexive
Internally reflexive
Anticausative
Pseudo-reflexive
Juan se come un bocadillo. Dative reflexive Aspectual
'Juan eats up a sandwich' Dative of interest (although divided with
Expletive SE agentive and transitive SE)
Optional reflexive
Ethical dative
Emphatic reflexive
Pseudo-reflexive
Juan se arrepintió Pure pronominal Inherent
'Juan repented' Inherently reflexive
Obligatory reflexive
Se vive bien aquí Reflexive impersonal Impersonal
'One lives well here' Impersonal
Indefinitising SE
SE-impersonal
Non lexical impersonal
Syntactic impersonal
Se venden casas. Reflexive passive Passive
'Houses are sold' (pasiva refleja)
Non-periphrastic passive
Passive
Episodic passive
Este libro se entiende bien. Medio-passive Middle
'This book is easy to Modalised passive
understand' Generic passive

All these factors pile up on top of the range of constructions where SE appear to make its
discussion a titanic enterprise within an overview article. However, we will do our best to cover
all the major facts, analyses and theoretical options. We want to highlight that, given the
existence of two major overviews of SE in Spanish (Mendikoetxea 1999, Sánchez López 2002),
here we will favour the overview of the analyses after 2002, with some classical exceptions
that we will also include.
The rest of this article is structured as follows. The article divides into three parts: first, we
will introduce the main properties of SE within the clitic system in Spanish and specify the
other classes of structures where SE has been proposed within Spanish (§2). The second part
of this article contains sections §3-§10, which concentrate on the empirical aspects of each one
of the uses of SE in Spanish and provide an overview of the main analyses that have been
proposed for each one of these constructions. The discussion of the theoretical and analytical
implications in these eight sections will be limited to the relation between at most three SE

9
 ANTONIO FÁBREGAS

uses. The third part of the article is devoted to the general theoretical and analytical status of
SE as an element and its role in the structures where it is used in Spanish, from the perspective
of its nature (§11), its position (§12) and the prospects of reducing all types of SE to the same
kind of operations.
The conclusion (§13) we will reach is that it is virtually inescapable to accept that SE is
linked to two types of objects, defective elements in argument positions and defective verbal
heads. What the two types of object have in common is that they lack referential capacity of its
own: the arguments are defective because they cannot introduce a referent of their own and
have to link to another participant, and the verbal heads are defective because they introduce
argument structures where one of the expected positions is not filled or is restricted in its
referential capacity to indefinite arguments, or arguments with a prototypical denotation. An
approach where SE spells out an element that is high in the clausal structure (Kayne 2000) and
acts as a probe that introduces agreement to license defective elements (Reuland 2011) is a
promising way to explore in order to unify all uses of SE.
This is going to be a long and complex discussion, so let us start immediately.

2. SE types: chaos and common properties

The structure of this section is as follows. We will start in §2.1 by providing an overview of
the main roles that have been differentiated in the study of SE. Despite the high number of
constructions where SE appears, there are reasons –beyond traditional accounts– to treat the
whole set of structures as involving the same element, and to accept a division that depends on
whether SE appears within the verbal complex or outside it. We will then focus on what the
types of SE have in common, considering three aspects of its behaviour: that SE behaves as a
clitic (§2.2), that its morphological properties are always the same (§2.3) and that its ordering
within a clitic sequence is always the same as well (§2.4).

2.1. The many facets of SE

A cursory overview of the different classes of SE that were introduced in the previous
section show that a distinction between paradigmatic SE and non-paradigmatic SE, albeit
traditional, might not be very useful. The reason is that in the allegedly non-paradigmatic SE
we still have signs that a third person pronoun is involved, and it simply happens that the
properties of the constituent associated to what becomes SE presuppose that the person has to
be 3rd, with 2nd and 1st participants excluded.
For this reason, in this overview we will adopt a moderately different –but not
groundbreaking– division: one that refers to whether the role of SE affects the lexical properties
of the predicate or not, where the lexical properties can be of the following types:

i) addition or supression of arguments

ii) imposing specific additional entailments on existing arguments
iii) manipulating the lexical aspect of the predicate
iv) altering the conceptual meaning of the predicate

Most SE structures –reflexive, reciprocal, anticausative, aspectual, inherent, figure

reflexive, antipassive, agent reflexive, and causative SE belong to this class, which can be
further subdivided with respect to which one of the four dimensions is manipulated. The second
class will be the types of SE that affect the broader clausal structure –grammatical aspect,
modal and temporal notions, clausal syntactic functions, etc.–, but without altering the lexical
information of the predicate. Spurious, passive, impersonal and middle SE fall into this class.

10
 SE IN SPANISH

Table 2. Types of SE divided by their role

Traditional Predicate vs. Name of SE type Example Comment
type Clause
Paradigmatic Predicate- 1. Reflexive Juan se lavó. Affects the
affecting (§3.1) 'Juan washed' coreference
between
arguments.
Paradigmatic Predicate- 2. Reciprocal Juan y María se Affects the
affecting (§3.2) miraron. coreference and
'Juan and María the theta role
looked at each reading of
other' arguments.
Paradigmatic Predicate- 3. Anticausative La televisión se Affects the
affecting (§4) rompió. number of
'The TV broke' arguments
Paradigmatic Predicate- 4. Dative Juan se cortó el Affects the
affecting benefactive dedo. relation between
(§4.4, §5) 'Juan hurt his subject and
finger' indirect object
Paradigmatic Predicate- 5. Aspectual Juan se salió. Affects the
affecting (§5) 'Juan got out' lexical aspect of
the predicate
Paradigmatic Predicate- 6. Agent Juan se leyó tres Affects the
affecting reflexive novelas. entailments of
(§5.2) 'Juan managed to the arguments
read three
novels'
Paradigmatic Predicate- 7. Transitive La mesa se come Creates a new
affecting (5.2) la habitación predicate with
'The table potentially new
occupies the argument
whole room' structure
Paradigmatic Predicate- 8. Factitive Juan se cortó el Affects the
affecting (§6.1) pelo. entailments of
'Juan got a the arguments
haircut'
Paradigmatic Predicate- 9. Antipassive Juan se encontró Affects the
affecting (§6.2) con alguien. entailments of
'Juan the arguments
accidentally met and case
someone' assignment
Paradigmatic Predicate- 10. Inherent Juan se jactó de Affects the
affecting (§6.3) su éxito. lexical meaning
'Juan boasted of the predicate
about his
success'
Paradigmatic Predicate- 11. Figure Juan se metió en Affects number
affecting reflexive casa of arguments
(§6.4) 'Juan got home'
Non- Clause-affecting 12. Impersonal Se vive bien Affects the
paradigmatic (§7) aquí. interpretation of
'One lives well the subject
here'

11
 ANTONIO FÁBREGAS

Non- Clause-affecting 13. Passive (§8) Se hablan Affects diathesis

paradigmatic idiomas.
'Languages are
spoken'
Non- Clause-affecting 14. Middle (§9) Estas novelas se Affects diathesis
paradigmatic venden bien. and the
'Such novels sell interpretation of
well' the subject
Non- Clause-affecting 15. Spurious Se lo di. Affected by
paradigmatic (§10) 'I gave it to her' other clitics

As can be easily be seen above, the paradigmatic/non-paradigmatic distinction corresponds

to a division that is perhaps more informative, one where the divide is whether SE affects the
predicate –and is therefore likely to be introduced somewhere within the verbal structure– or
not. From this perspective, the non-paradigmatic cases are just cases where the third person is
somehow forced by the clausal structure, and the paradigmatic cases are instances where SE is
introduced somehow within the verbal structure, and when that complex inflects SE inflects
with it.
However, why are these structures all considered instances of the same element? When we
think about it, this conclusion is not granted. As Otero (2002) notes, SE could be viewed as a
phonological default manifestation: a single CV syllable, with the unmarked /e/ vowel that is
anyways used as epenthetic in other contexts (cf. standard > estándar) and an /s/ consonant
that is equally unmarked and in fact is extrametrical in many processes (Harris 1983). Why
don't we pose homophony relations between two or more distinct SE elements, then? The
reason is that from other perspectives, SE always displays the same set of properties, which we
will review in the remainder of this section. Given that these properties are identical, the
conclusion that there is only one SE is forced.

2.2. Properties of SE as a clitic

The first set of facts that force the conclusion that the 13 structures mentioned in the table
above involve the same element is that SE always behaves as a clitic, with the same set of
properties. As a clitic, in proclitic position SE always precedes prefixes of the verb, such as for
instance medio 'half-way':

(27) a. Se medio miraron. Reflexive / Reciprocal

SE half-way looked
'They almost looked at themselves' / 'They almost looked at each other'
b. Se medio rompió. Anticausative
SE half-way broke
'It almost broke'
c. Se medio acusó a los testigos. Impersonal
SE half-way accused DOM the witnesses
'Someone sort of accused the witnesses'
d. Se medio escribieron informes. Passive
SE half-way wrote reports
'The reports were partially written'
e. Esta clase de persona se medio enamora fácilmente. Middle
this type of person SE half-way falls.in.love easily
'This type of person has a tendency to sort of fall in love easily'
f. Se lo medio regalaron. Spurious
SE it half-way gave

12
 SE IN SPANISH

'They sort of gave it to him as a present'

Negation systematically precedes SE in all these cases too.

(28) a. No se miraron. Reflexive / Reciprocal

not SE looked
'They didn't look at themselves' / 'They didn't look at each other'
b. No se rompió. Anticausative
not SE broke
'It didn't break'
c. No se metió en la cama. Figure
not SE got into the bed
'He didn't get in bed'
d. No se cortó el pelo. Factitive
not SE cut the hair
'He didn't get a haircut'
e. No se leyó tres libros. Aspectual
not SE read three books
'He didn't manage to read three books'
f. No se olvidó de ello. Antipassive
not SE forgot of that
'He didn't forget about it'
g. No se arrepintió. Inherent
not SE repented
'He didn't repent'
h. No se durmió. Aspectual
not SE fell.asleep
'He didn't fall asleep'
i. No se lo dimos. Spurious
not SE it gave
'We didn't give it to him'
j. No se venden casas. Passive
not SE sell houses
'Hiuses are not sold'
k. Estos libros no se venden bien. Middle
these books not SE sell well
'These books don't sell well'
l. No se vive bien aquí. Impersonal
not SE live well here
'One doesn't live well here'

Similarly –and with only some restrictions that refer to the presence of impersonal SE with
non finite categories, see §7– all types of SE allow rising within periphrases.

(29) a. Se pueden mirar. Reflexive / Reciprocal

SE can look
'They can look at themselves' / 'They can look at each other'
b. Se puede romper. Impersonal
SE can break
'It can break'

13
 ANTONIO FÁBREGAS

c. Se puede lanzar al agua. Figure

SE can throw into the water
'He can throw himself into the water'
d. Se puede cortar el pelo. Factitive
SE can cut the hair
'He can get a haircut'
e. Se puede leer tres libros. Aspectual
SE can read three books
'He can manage to read three books'
f. Se puede enfrentar a ello. Antipassive
SE can confront to that
'He can face it'
g. Se puede arrepentir. Inherent
SE can repent
'He can repent'
h. Se puede dormir. Aspectual
SE can fall.asleep
'He can fall asleep'
i. Se lo puedo dar. Spurious
SE it can give
'I can give it to him'
j. Se pueden vender casas. Passive
SE can sell houses
'Houses can be sold'
k. Estos libros se pueden vender bien. Middle
These books SE can sell well
'Such books can be easily sold'
l. Se puede vivir bien aquí. Impersonal
SE can live well here
'One can live well here'

Finally, let us not forget that SE structures share three other properties: (i) they involve third
person entities, (ii) they have some type of effect in the participants in the eventuality, as
arguments or as syntactic functions, depending on whether they operate on the lexical
properties of the predicate or on the clause structure –with the obvious complication that in
inherent SE structures we cannot compare with a SE-less version of many verbs to see what its
effect is– and (iii) they presuppose some kind of verbal structure, and never combine with verb-
less predicates, noun phrases or adjectival phrases.

2.3. Morphological properties

Moreover, the morphological properties of SE are also homogeneous in all cases, most
visibly in the instances where SE is paradigmatic and one would expect it to alternate with
other pronouns. Consider in this respect the place of SE within the clitic system of Spanish (cf.
Bonet 1991).

(30) a. lo accusative masculine singular third person

b. la accusative feminine singular third person
c. los accusative masculine plural third person
d. las accusative feminine plural third person

14
 SE IN SPANISH

(31) a. le dative singular third person

b. les dative plural third person
(32) a. me singular first person
b. nos plural first person
c. te singular second person
d. os plural second person
(33) se third person

The pronouns in (30) and (31) are third person pronouns, like SE. They differentiate case
marking –dative vs. accusative–, as witnessed by the following pairs –which abstract away
from possible leísmos, laísmos and so on, cf. Fernández Ordóñez (1999)–.

(34) a. Le dieron una información.

her.dat gave some information
b. La informaron.
her.acc informed
'She was informed'

The accusative clitics in (30) differentiate gender and number. The dative clitics in (31) do
not differentiate gender in Spanish, but they do differentiate number.
The clitics in (32) mark person, as first or second person; following Benveniste (1960) we
could assume that 1st and 2nd are the only true values of person, and treat third person as the
absence of person. In any case, the empirical fact is that (32) contrast with both (31) and (30)
in the person information, but unlike (30)-(31) they do not differentiate between accusative and
dative. The same forms in (32) are used in both syntactic contexts, without distinctions.

(35) a. Me dieron una información.

me gave some information
'They gave me some information'
b. Me informaron.
me informed
'I was informed'

The clitics in (32) do not differentiate between gender, but like the ones in (31), differentiate
number.
How about SE in (33)? SE does not differentiate between case, as witnessed by (36), where
the subject can be coreferential with the direct object or with the indirect object:

(36) a. Juan se vio a sí mismo.

Juan SE saw DOM SE self
'Juan saw himself'
b. Juan se dio un regalo a sí mismo.
Juan SE gave a present to SE self
'Juan gave a present to himself'

SE does not differentiate, furthermore, between masculine and feminine or between singular
and plural, as (37) shows:

15
 ANTONIO FÁBREGAS

(37) a. Juan se mira.

Juan SE looks
'Juan looks at himself'
b. María se mira.
María SE looks
'María looks at herself'
c. Ellos se miran.
they.m SE look
'They look at themselves'
d. Ellas se miran.
they.f SE look
'They look at themselves'

In consequence, SE is the pronoun that, within the clitic system, makes the smallest number
of morphological contrasts: only person, on the assumption that third person should be viewed
as 'person' and not as absence of any person value.

(38) Person Number Case Gender

lo-la x x x x
le x x x
me-te... x x
se x

If Person is removed from third person pronouns, then SE would rather correspond to a
pronoun without any phi features, making no case distinctions.

(39) Person Number Case Gender

lo-la x x x
le x x
me-te... x x
se

What this comparison emphasises is yet another property that most analyses of SE highlight:
SE is defective in features, that is, contains the smallest number of features within the
pronominal system of Spanish. The analyses, as we will have a chance to see in the sections
dedicated to this topic, share the intuition that the broad range of structures where SE appears
is made possible by the almost complete absence of features that this element carries, making
it able to be introduced in different syntactic context without those features restricting its
distribution. In some approaches, like Reuland (2011), in fact the reflexive pronoun SE is
simply a set of unvalued nominal features that take its value from other elements, or can be
used as a head that agrees with other elements in order to license them (see in particular §11.4
below and the discussion in §12 and §13). SE, from this perspective, would be a category which
might only have its category, [SE], as an interpretable feature, and therefore it would not really
be a member of the same natural class as Spanish clitics. However, let us for the moment
present the rest of its properties within the clitic system.

16
 SE IN SPANISH

2.4. Clitic ordering

Spanish shows a rigid ordering when it comes to clitic clusters or clitic combinations.
Whenever two or more clitics combine in the same clause, it is impossible to alter a fixed order,
irrespective of the information structure associated to each constituent. (40) presents the rigid
ordering of clitics in the standard varieties.

(40) IV III II I
se > me > le > lo
te les la
nos los
os las

This ordering deserves some comments. The fact that SE is the initial clitic in the cluster
can be tested without problems, with sentences such as (41).

(41) a. Se me cayó el libro.

SE me.dat fell the book
'I dropped the book accidentally'
b. Se le entregó una carta.
SE him.dat delivered a letter
'A letter was delivered to him'
c. Se las atacó.
SE them.acc attacked
'Someone attacked them'

It is also possible to test that personal clitics in III are ordered before accusative clitics in I
through sentences like (42).

(42) a. Me lo dijeron.
me.dat it.acc said
'Someone said it to me'
b. Nos las dieron.
us.dat them.acc gave
'They gave them to us'

When it comes to the ordering between III and II, the evidence is less direct. Sequences
where III and II are combined together are, in principle, instances of clitic combinations that
should be ruled out by the Person Case Constraint (Perlmutter 1971, Bonet 1991,
Anagnostopoulou 2003, Béjar & Rezac 2003, Nevins 2007): situations where a dative third
person clitic is incompatible with an accusative personal clitic.

(43) *Me le presentaron.

me.acc him.dat introduced
Intended: 'I was introduced to him'

However, there are situations where this effect gets at least amiliorated for some speakers:
cases with double dative, where the personal pronoun is the so-called 'ethical dative' –also
known since Bello (1847) as 'superfluous dative' or 'dative of interest'–, introducing the
personal affective involvement of the speaker in the eventuality reported. In such cases, such
as (44) it can be checked that the personal pronouns in III precede the datives in II.

17
 ANTONIO FÁBREGAS

(44) No me le deis caramelos al niño.

Not me.dat him.dat give candy to.the child
'Don't give candy to the child, I beg you'

The ordering between II and I is not so easy to test directly, because systematically
combinations of a dative clitic with an accusative clitic involve substitution of the dative clitic
by SE –the 'spurious SE' cases that were mentioned in §1–.

(45) *Les lo dimos.

them.dat it.acc gave
Intended: 'We gave it to them'

However, Kayne (2010), building on results from Harris & Halle (2005) on mesoclisis of
the pronoun in some Spanish varieties, finds an indirect argument that justifies the ordering
between II and I despite absence of overt transparent clitic combinations. These authors discuss
cases like (46), involving a 3pl subjunctive form in the imperative, where the clitics follow the
verb in Spanish.

(46) a. Diga-n-me
say-3pl-me.dat
'Tell it to me'
b. Diga-me-n
say-me.dat-3pl
c. Diga-n-me-n
say-3pl-me.dat-3pl

The relevant cases are (46b) and (46c), where the inflection appears after the pronoun. Harris
& Halle (2005) note that there are 4 relevant varieties that contrast in the types of pronouns
that can appear before the inflection. Interestingly, the pronouns that allow (46b)-(46c) in each
variety are in a containment relation:

(47) a. Variety I: se
b. Variety II: se, me (and by hypothesis the other personal pronouns)
c. Variety III: se, me, le
d. Variety IV: se, me, le, lo

Every variety that allows mesoclisis allows it with SE. There are varieties that in addition
to SE allow the pronouns in III; there are further varieties that add also the clitics in II, and
finally a variety that also allows mesoclisis with the accusative clitics in I. Kayne (2010)
proposes that this is the expected result if two conditions are granted: (i) the clitic ordering
reflects a hierarchical syntactic relation like the one in (48), where the clitics to the left are
clitics that c-command the others (as expected in Kayne 1994) and (ii) if the mesoclisis happens
when subject agreement can appear in a lower clausal position instead than the usual T head
(46b) or in addition to it (46c).

18
 SE IN SPANISH

(48) ...seP

se XP

X meP

me XP

X leP

le XP

X loP

lo XP

X ...vP

The tendency would be, then, to have this subject agreement morphology as high as possible.
Varieties which allow mesoclisis with SE but not with the other clitic classes (49) would be
varieties like (51), where the additional agreement position is immediately below seP. Varieties
that, in contrast, allow all pronouns (50) would be varieties where the agreement position is
active in the lowest XP, so that this inflection follows any clitic, no matter how low in the
structure (52).

(49) a. Siente-se-n
sit.down-SE-3pl
b. *Diga-me-n
tell-me-3pl
c. *Diga-le-n
tell-her-3pl
d. *Diga-lo-n
tell-it-3pl
(50) a. Siente-se-n
sit.down-SE-3pl
b. Diga-me-n
tell-me-3pl
c. Diga-le-n
tell-her-3pl
d. Diga-lo-n
tell-it-3pl

19
 ANTONIO FÁBREGAS

(51) ...seP

se XP

X meP
-n
me XP

X leP

le XP

X loP

lo XP

X ...vP
(52) ...seP

se XP

X meP

me XP

X leP

le XP

X loP

lo XP

X ...vP
-n

The absence of a variety that allows (50d) without allowing (50c) and the existence of
varieties that allow (50c) but not (50d) imply that the clitic pronouns in II are higher than the
clitic pronouns of I, and excludes the possibility that they are located in the same position –
something that in principle could have explained their incompatibility–.
Note that Kayne (2010) differs with respect to Kayne (2000) in whether SE and the person
clitics should be distinguished or not. It is true that SE (at least in the cases traditionally called
'paradigmatic') alternate with person marked clitics (Se caer 'SE falls, He falls', Me caigo 'ME
fall, I fall'), but there are other reasons to propose a distinction. In the representation above,
and given the facts that involve imperatives and clitic combinations, the existence of varieties
where SE licenses the post-clitic inflection but personal pronouns don't, require that the two
clitics occupy two different positions. Also, the possibility that SE and a personal clitic co-
occur suggest that analysis (Se me cae 'SE ME falls, It falls from me).
An interesting fact if one assumes some kind of syntactic hierarchy like (48) is that the two
known clitic incompatibilities that one identifies in Spanish refer to clitics that are in adjacent

20
 SE IN SPANISH

areas. The Person Case Constraint involves a clash between clitics in the III region and those
in the II region; the spurious SE cases replace a clitic in the II region when there is a clitic in
the I region.

(53) ...seP

se XP

X meP
PCC clashes
me XP

X leP
*le-lo clashes
le XP

X loP

lo XP

X ...vP

Interestingly, the solution to the 3rd-3rd clash in (53) involves using a SE element that is in
a higher region, not adjacent to the accusative one. In the case of PCC one could also argue
that a solution is SE, although in a less obvious sense: when the personal pronoun comes from
a reflexive SE that agrees in person features with the subject the PCC effect is less serious
(Quiere-te-me 'Love-yourself-me, Love yourself for me') than when the personal pronoun is
not related to a reflexive construal (*Te me quieren, 'You me love.3pl, They love you for me')
A second interesting fact is that the morphological relation that we identified in (38)-(39) is
replicated in the syntactic hierarchy. Clitics in the I area are those that make the biggest number
of distinctions (gender, case, number, person); clitics in the II area lose the gender distinction;
clitics in the III area furthermore lose the case distinction, and SE in the IV area only keeps a
person distinction.

21
 ANTONIO FÁBREGAS

(54) ...seP

se XP
person
X meP

me XP
person
number X leP

le XP
person
number X loP
case
lo XP
person
number X ...vP
case
gender

From this perspective, the lower clitics are the ones that keep more pronominal features, and
as we go up in this syntactic hierarchy pronominal features are missing.
Beyond this, SE is always the first clitic in the sequence in any structure where it can be
combined with other clitics. This shows that, at least for the linearisation of the forms, we have
only one clitic SE for all these cases, even if the role of that clitic seems to be different
depending on the context:

(55) a. Se lo pueden quitar.

SE it can remove
'They can remove it to themselves' / 'They can remove it to each other'
b. Se le puede romper.
SE him.dat can break
'It can break on me'
c. Se le puede lanzar encima.
SE can throw into the water
'He can throw himself on top of him'
d. Se lo puede cortar.
SE it can cut
'He can get (a haircut)'
e. Se los puede leer.
SE them can read
'He can manage to read them'
f. Se le puede enfrentar.
SE him.dat can confront
'He can face him'
g. Se le puede arrepentir.
SE him.dat can repent
'He can repent for him'
h. Se les puede dormir.
SE them.dat can fall.asleep

22
 SE IN SPANISH

'He can fall asleep for them'

i. Se lo puedo dar.
SE it can give
'I can give it to him'
j. Se le pueden vender casas.
SE her.dat can sell houses
'Houses can be sold to her'
k. Estos libros se les pueden vender bien.
These books SE them.dat can sell well
'Such books can be easily sold to them'
l. Se los puede recibir.
SE them.acc can receive
'One can receive them'

These generalisations apply to all types of SE, and as such should be taken into account for
the analysis of all types of SE: the reason is that –as noted in §2.1 and §2.2– all types of SE
constructions contain the same clitic, characterised by the same feature defectiveness, the same
placement with respect to an inflected verb or negation, and the same position within the
ordering of clitics. These facts should be taken as a background for any analysis that tries to
unify the role of SE across constructions.
In the following sections, and before we start discussing the analytical options, however, we
will review the main empirical properties of each one of these constructions, starting with the
predicate-affecting SE uses.

3. Predicate-affecting SE uses (1): reflexives and reciprocals

The first use that we will discuss is the reflexive, where we will also treat reciprocals as
possibly an additional set of interpretations that emerge from the reflexive under certain
conditions. As we will see, analyses of other types of SE as involving a reflexive have been
proposed in anticausative cases, aspectual cases, agentive cases, figure cases, and even passive
and impersonal cases.
One reason to pick the reflexive SE constructions as the first is that many theories treat the
reflexive SE as the basic manifestation of this pronoun, from where the other structures should
be derived (see for instance Armstrong & MacDonald's recent 2021 overview). Undoubtedly,
if one looks at the historical evolution it is clear that the reflexive has to be taken as the basic
SE form, given that the Spanish pronoun emerges from the Latin reflexive pronoun se, which,
under certain conditions, behaved as a logophoric pronoun. From a historical perspective, a
well-documented fact of Romance is that the reflexive pronoun SE develops early on, already
in Latin, a use that traditional grammars call 'middle' and which corresponds to what in this
article we call 'anticausative' (Cennamo 1999; Cennamo, Eythórsson & Bardal 2015; Gianollo
2014). The standard grammaticalisation account of this evolution is generally explained as
follows. There is an initial meaning (reflexive) where there is an agent and an internal
argument, accusative or dative, which happen to be coreferential. From there, SE takes a
meaning where the patient or benefactive of the event is taken to be the sole responsible entity
in initiating the event, without any help from an external agent; this has been interpreted often
as a grammaticalisation process that moves from a bi-argumental construal to a construal with
one single participant where the notion of patient and agent are collapsed in the same
participant –hence the term 'middle', that was intended as an intermediate position between the
active and the passive voice–. Sánchez López & Vivanco (in press) note that in Spanish this
process takes place earlier with psychological predicates, where presumably the notion of
experiencer behaves as an intermediate category between (volitional) agent and patient. From

23
 ANTONIO FÁBREGAS

there, the next class of SE that appears in Romance is the passive (Maddox 2021, Wolfsgruber
2021), presumably building from the association between SE and a non-agentive subject in the
anticausative; from there, the last step, which is not performed in all Romance languages, is
the impersonal, which generalises the relation between SE and structures where the agent is
not syntactically projected (Monge 1955, Wolfsgruber 2021). Thus, the historical evidence
supports an analysis of SE that takes the reflexive form as basic.
This privileged position of reflexive SE within the uses of SE is confirmed by acquisitional
studies, which have generally reached the conclusion for Spanish that the reflexive and the
anticausative uses are acquired at earlier stages than the passive and the impersonal SE (Bruhn
de Garavito 1999, Montrul 2001, Tremblay 2006, Gómez-Soler 2015, Escobar & Teomiro
2016, García Tejada, Cuza & Lustres Alonso 2021, among many others).
It is also plausible to adopt this type of view from the perspective that many languages use
the reflexive pronoun to –typically– express also passive and anticausative meanings (see
Koontz-Garboden 2009). However, this claim is by no means granted when one looks at the
internal synchronic situation in Spanish and other Romance languages (see Otero 2002, Labelle
2008, among others), because SE is neither sufficient nor necessary to express reflexive or
reciprocal meanings. Let us examine the facts.

3.1. Reflexives
SE is sometimes identified with the spell out of a reflexive function which could be
contained within a projection of the verb, or acts as a reflexive marker of the verb –perhaps not
corresponding to a distinct syntactic head, but as some kind of morphological increment of the
verbal stem–. In this view, the reflexive marker SE acts as a verbal operator that reduces the
number of arguments that the predicate contains, by forcing that two theta roles are identified
with each other (Quine 1961, Grimshaw 1982, Reinhart 1996, Chierchia 2004). In fact, Koontz-
Garboden (2009: 83) proposes that reflexivisation, defined as in (56) below, corresponds to the
denotation of Spanish SE.

(56) lℜlx[ℜ(x,x)]

This analysis takes ℜ to be a variable that ranges over transitive verbs; in other words, as
Reinhart & Reuland (1993) proposed, reflexivisation involves marking a predicate as reflexive
by imposing an additional restriction on its argument structure. This denotation explains, in
principle, the semantic interpretation of sentences like (57), where the same individual is
interpreted as receiving the theta roles corresponding to two argument positions.

(57) a. Pedro se criticó a sí mismo.

Pedro SE criticised DOM himself
b. María se contrató a sí misma.
María SE hired DOM herself
c. Luis se peinó.
Luis SE combed
'Luis combed his hair'
d. Ernesto se hizo un regalo a sí mismo.
Ernesto SE made a present to himself

The analysis of reflexive SE as a function that operates on the verbal theta structure also
captures the generalisation that reflexivisation cannot apply to verbs that have only one
argument, or potentially lack any arguments.

24
 SE IN SPANISH

Beyond this, there is a significant controversy with respect to whether reflexive structures,
even within one single language, should be considered as containing transitive or intransitive
configurations. Moreover, a relevant question is whether the overt argument that is not
expressed as SE occupies an external or an internal argument position.
The second option, that the overt argument is in an internal argument position, would make
reflexive predicates virtually unaccusatives (Marantz 1984). In these works, reflexive verbs are
taken to be unaccusatives, meaning that the role that actually is suppressed by reflexivisation
was the one related to the external argument –as evidenced, for instance, by the use of the same
auxiliary in the case of reflexive verbs and in the case of unaccusative verbs like 'arrive' in
French or Italian–.

(58) a. Jean est arrivé. French

Jean is arrived
'Jean arrived'
b. Jean s'est regardé dans le miroir.
Jean SE-is watched in the mirror
'Jean watched himself in the mirror'

In this analysis, syntactically, SE can be generated either in an argument position (Pesetsky

1995) where it appears instead of the suppressed argument, or it can be taken to be the way in
which the relevant verbal projection gets spelled out (Marantz 1984). In both types of analyses,
SE blocks the introduction of an external argument.
However, the characterisation of reflexives as unaccusatives in Spanish is far from obvious.
Here and in the rest of the article we will assume the following tests as properties that should
be associated to unaccusatives:

i) The absolute participle structure, which is possible with internal arguments, and therefore
can only appear with subjects in the case of unaccusative verbs (Muerto el perro... 'Dead the
dog...')
ii) The possibility of having postverbal subjects projected as bare nominals, on the
assumption that bare nominal arguments are only licensed in internal argument positions
(Llegan niños 'Arrive.3pl children').
iii) The impossibility of adding additional accusative / direct object arguments, on the
assumption that there is one single position for direct objects and in unaccusative predicates
that position is occupied by the subject.

As it is well-known, all these tests can be questioned, but this is not a state of the art about
unaccusativity –that will be some other time–, so we will take them at face value. Given these
tests, the nature of reflexive predicates as unaccusatives can be questioned in Spanish. The
absolute participle construction with reflexives is impossible (Visto Juan, se fue de casa 'Seen
Juan, he left home' cannot be interpreted as meaning 'Once Juan saw himself, he left home';
one could argue that the structure is out due to the referential restrictions between pronouns
and proper names, but still the test is negative).
Second, reflexive predicates can be built with direct objects, and the reflexive pronoun
seems to appear in the position of an indirect object (Juan se lavó las manos 'Juan SE washed
his hands').
Moreover, reflexives do not allow bare nominal subjects postposed to the verb (#Se vieron
niños 'SE saw.3pl children' cannot be interpreted as 'Children saw themselves', but as a passive
SE, 'Children were seen', cf. §8.1). For this reason, even in the analysis where the overt
argument of a reflexive is merged in an object position, one probably wants to force movement

25
 ANTONIO FÁBREGAS

of the argument to an external position. One option is to force movement of the argument to an
external argument position, so that the verb is 'unaccusative' from the perspective of its base
position, but the argument moves to an external position.
For instance, Embick (2004) –see also Folli (2001)– treats reflexive SE as merged in the
external argument position (for him, spec, vP); from there, it cliticises to the head v. The
argument that is visible in the surface corresponds to the internal argument, that rises to spec,
vP to c-command the SE anaphor.

(59) ...vP

DP v
Juan
se v

v VP

V Juan

Note that if one assumes a ditransitive structure, this type of analysis could derive the
unaccusative nature of the predicate while keeping a direct object: the displaced argument
could be the dative.
From this perspective, SE has been viewed as an impoverisher of the verbal structure,
something that will be a common trend in the analyses presented in this section, and then in
§3-10, with spurious SE being a rare exception that does not seem to affect the assignment of
arguments, aspect or case in the verb. SE is either the spell out of a verbal head that, with
respect to the SE-less version, is less structurally robust or an element that, adjoined to a head,
impoverishes it.
However, the impoverishing function of SE varies from one analysis to the other, making
the intuition much less easier to express than one would have expected. For some authors, SE
would have the role of absorbing the case, generally accusative, that otherwise the verb could
have assigned to the internal argument (see Manzini 1986, who originally propose the idea,
although focusing on anticausatives; otherwise, the internal argument would not move up, and
empirically could be substituted by accusative clitics). The problem is, of course, that reflexive
predicates can carry direct objects. An alternative view, but related to it, is that SE absorbs the
theta role that otherwise will be assigned to the external argument (Cinque 1988).
A minimal variant of the analysis in (59) is to locate SE in the position that otherwise would
have been occupied by the external argument (McGinnis 1998), the specifier of vP or
whichever projection introduces agents, and displace the overt subject from an internal
argument position. This illustrates another analytical dichotomy that we will see once and again
in this overview: SE as an argument introduced in a specifier or complement position vs. SE
as a head (or attached to a head) that defines a verbal head.

26
 SE IN SPANISH

(60) ...XP

DP vP
Juan
se v

v VP

V Juan

Be it as it may, and with all these analytical options on the table, the problem for analyses
like (59) is that the reflexive predicate is treated as an unaccusative, which does not seem to
correspond to the grammatical behaviour of these predicates in Spanish. Later on in this section
we will see analyses that share the core intuition but base-generate the overt argument of the
reflexive directly on an external argument position, thus removing any chance to treat the
reflexive verbs as unaccusative.

Table 3. Main analytical options in the location of SE

Argument Verbal head
Head SE is an X nominal element SE is the spell out
that occupies an argument (alternatively, marks) a
position verbal head
XP SE is an XP that occupies an SE is a projection within the
argument position set of functional projections
of the verb

3.1.1. SE is not necessary for reflexivization

However, there are several problems with the analysis that treats SE as an object that forces
movement of an argument to satisfy a reflexive relation. Importantly, Koontz-Gardboden
(2009) restricts the variable ℜ to transitive verbs –in a broad sense of transitive where
transitives and ditransitives are included–. This is empirically crucial, but arbitrary from the
perspective of reflexivisation.
The restriction about the type of verbs that ℜ can correspod to in (56) is due to the fact that
SE appears in Spanish reflexives only when the two arguments have specific syntactic
functions. The first argument must necessarily be a subject (61a), which explains the
grammaticality of sentences like (57d) above and the ungrammaticality of (61b) below.

(61) a. Carlos se miró en el espejo.

Carlos SE looked in the mirror
b. Pedro (*se) entregó a los prisioneros a sí mismos.
Pedro SE surrendered DOM the prisoners to themselves
'Pedro surrendered the prisioners to the prisioners themselves'

This subject is agentive, or at least can be interpreted broadly as an initiator (Ramchand

2008), in contrast to anticausative SE structures, as witnessed by the compatibility with some
volitional adverbs and final clauses.

27
 ANTONIO FÁBREGAS

(62) a. Luis se lavó voluntariamente.

Luis SE washed voluntarily
b. Luis se lavó para ir bien a la fiesta.
Luis SE washed in.order.to go well to the party
'Luis washed so that he could attend the party with a good image'

The second argument can be a direct or an indirect object of a verbal predicate for SE to
appear in the reflexive meaning. Other semantically transitive verbs which introduce the
internal argument with a preposition, or instances of any other syntactic function, allow
reflexivisation, but again do not mark it with SE.

(63) a. Luis se hizo un tatuaje a sí mismo.

Luis SE made a tattoo to himself
b. Luis se asustó a sí mismo.
Luis SE scared DOM himself
(64) a. Luis (*se) piensa en sí mismo.
Luis SE thinks on himself
b. Luis (*se) depende de sí mismo.
Luis SE depends on himself
c. Luis (*se) cuenta consigo mismo.
Luis SE counts on himself
(65) a. Luis (*se) acudió a la fiesta consigo mismo.
Luis SE attended to the party with himself
b. Luis (*se) le compró un filete al carnicero para sí mismo.
Luis SE him.dat bought a steak to.the butcher for himself

Trivially, SE does not appear in any type of reflexive predicate. If the reflexive predicate is
a noun or an adjective, SE is impossible even if one can identify the coreferential argument as
notionally a patient or goal. One partial exception is a few cases of adjectives that take datives
and that, in combination with a copulative verb, might mark the dative with SE (66c); in such
cases one might think that the copulative verb and the adjective behave syntactically and
semantically as one single predicate, thus producing in combination one single verbal predicate
with two arguments that happen to be coreferential.

(66) a. el retrato de Goya de sí mismo

the portrait of Goya of himself
b. un hombre orgulloso de sí mismo
a man proud of himself
c. Juan (se) es fiel a sí mismo.
Juan SE is faithful to himself

This, in essence, means that SE is not a necessary element of reflexive structures.

(67) SE is only compulsory in reflexive structures in Spanish when the coreferential

arguments are subject and direct object or subject and indirect object.

The analysis in (59) does not completely fail in explaining this pattern. The crucial problem
in the cases where SE is not present could be that the contexts above are all cases where the
lower argument that must c-command the SE anaphor cannot move out of its immediate

28
 SE IN SPANISH

constituent to license the anaphor –that is, the problem would be that in these examples the
movement step whereby the internal argument c-commands SE cannot take place–.

a) Examples where the two reflexive arguments are the direct object and the indirect object
could be instances where the direct object cannot move up to spec, vP to license the anaphor
that would be located in the indirect object, or cases where the anaphor would not be licensed
because it is embedded under a lot of structure. This presupposes, however, that SE can only
cliticise to the verb and not to other heads, and it establishes an asymmetry with structures like
(63a) that we will come back to.
b) Examples where the internal argument is introduced by a preposition distinct from the
dative one could be instances where the preposition already assigns case to the argument and
therefore this argument cannot move up to license the anaphor.
c) Similarly, examples where the coreferential constituent is not an argument of the verb
could be excluded if these are adjuncts that cannot extract or become subjects.

The analysis, however, makes it more difficult to explain cases involving a subject and an
indirect object. For such cases one would have to assume that the indirect object is not
introduced from a preposition, that is, one has to avoid a structure like (68).

(68) ... vP

se v

v VP

V ...PP

P DP
a Luis

The reason is that in order to generate from here 'Luis SE made a tattoo for himself', we
would have to extract Luis from the PP. If this movement is possible, then we must provide an
explanation of why it is not possible in the cases of (64), where the absence of SE would be
explained precisely by this. The alternative is to propose that in cases like (63a) we always
have an applicative structure (Pylkkännen 2002, Cuervo 2003).

(69) vP

se v

v VP

V ApplP

(a) Luis Appl

Appl a tattoo

In principle, applicatives assign case to the goal argument; assume however that the
presence of SE has an effect on case that makes Appl have to assign case to the direct object in

29
 ANTONIO FÁBREGAS

this case, and Luis is caseless, therefore being able to move up. This might explain sentences
like (63a), but then the problem is why we cannot get SE when the two arguments involved are
the goal and the theme. Imagine that now we introduce SE in the goal position.

(70) a. Pedro (*se) entregó a los prisioneros a sí mismos.

Pedro SE surrendered DOM the prisoners to themselves
'Pedro surrendered the prisioners to the prisioners themselves'

b. vP

Pedro v

v VP

V ApplP

se Appl

Appl the prisoners

Assume that SE has the effect of absorbing case in Appl, like in other cases. If 'the prisoners'
can move up to spec, VP, the anaphor would be licensed. Note that we can assume that 'the
prisoners' does not move to spec, ApplP because it already establishes a syntactic dependency
with that head, and that movement to spec, vP would not solve anything because Pedro would
still be closer to the anaphor than 'the prisoners', which precludes 'the prisoners' from licensing
the anaphor.

(71) vP

los prisioneros v

Pedro v

v ...ApplP

se Appl

Appl los prisioneros

The only way in which movement of the theme would license the anaphor is if the DP can
move to spec, VP.

30
 SE IN SPANISH

(72) vP

Pedro v

v VP

the prisoners V

V ApplP

se Appl

Appl the prisoners

This analysis, then, has to presuppose that some independent principle blocks movement of
the theme argument to the specifier of VP.
Alternatively, the analysis could presuppose that the ungrammaticality is not due to the
impossibility of moving an argument up to license the anaphor. The alternative explanation
could be that SE cannot just be introduced in any argument position, but rather that it cannot
be introduced in argument positions created by non-verbal projections, such as Appl. This
would block the insertion of SE in spec, ApplP –for that matter, in the specifier of this category
and any other category that is not verbal, explaining also why SE cannot appear as a marker of
reflexivity in nouns or adjectives–. Maybe the reason could be that SE needs to cliticise to a
verbal head, and if introduced in Appl SE could not cliticise to the verbal head.

3.1.2. SE is not sufficient for reflexivisation: the nature of the double mismo
However, this does not solve the problem of what SE does as an element. If SE was really
the spell out of a reflexivising function, one would expect that it would be compulsory in any
kind of reflexive predicate. If the syntactic restrictions on movement or on what type of head
it can combine with explain the ungrammatical cases, under this proposal the result would be
that such predicates could not be reflexivised, not just that SE cannot appear with them.
Moreover, if SE was a reflexivising function, then one would not expect it to co-occur with
other markers of reflexivity, or at least to appear only when one wants to put focus or emphasis.
There is at least one exponent, auto- 'self-', that has been argued in Spanish to correspond to
this type of function (Feliu 2003). Its presence does not preclude SE, which has the same
distribution as in cases without auto-. In fact, SE is compulsory in such cases, and it is not
possible to remove it, nor its presence is perceived as emphatic, contrastive or any of the other
semantic effects that are generally associated to sequences where 'redundant' material is
present.

(73) a. auto-destruir-*(se)
self-destroy-SE
b. auto-criticar-*(se)
self-criticise-SE

The English translation 'self-', in fact, matches with other analyses of reflexivity, such as
Labelle (2008), who proposes that if anything should be viewed as a reflexiviser, it is not SE,
but the double that can appear with it in reflexive contexts –in French, language that she
analyses, PRONOUN-même–.

31
 ANTONIO FÁBREGAS

(74) Le ministre se parle à lui-même.

the minister SE talks to him-self

Let us then talk a bit more about this double, first starting with its empirical description and
then moving to the problems that it presents for the analysis that treats SE as introduced in an
argument position. The empirical distribution of the double in Spanish (75), involving mismo
'same' in combination with a reflexive pronoun– makes it correlate, like any other pronominal
doubling with clitic in Spanish (76), with a notion of focus, contrast and emphasis.

(75) a. Juan se vio en el espejo.

Juan SE saw in the mirror
b. Juan se vio a sí mismo en el espejo.
Juan SE saw DOM himself in the mirror
(76) a. Juan la vio en el espejo.
Juan her saw in the mirror.
b. Juan la vio a ella en el espejo.
Juan her saw DOM her in the mirror

However, the question is whether there is a second effect between (75a) and (75b) that is
not found in the pair (76a)-(76b). Several authors have argued that, if reflexive SE structures
reduce the valency of the predicate and make it act as an unaccusative (at least partially), the
presence of the double involving self-même-mismo preserves the agentivity of the predicate.
This is part and parcel of the analysis of reflexives by Reinhart & Reuland (1993), Reinhart
(1996), Piñón (2001) or Härtl (2003).
A big part of the discussion of whether the double increases the agentivity of the predicate
with respect to the non-doubled reflexive involves so-called grooming verbs, such as those
illustrated in the following pairs.

(77) a. Juan se lava.

Juan SE washes
'Juan washes'
b. Juan se lava a sí mismo.
Juan SE washes DOM himself.
(78) a. Juan se peina.
Juan SE combs
b. Juan se peina a sí mismo.
Juan SE combs DOM himself
(79) a. Juan se maquilla.
Juan SE puts.make-up
b. Juan se maquilla a sí mismo.
Juan SE puts.make-up
(80) a. Juan se afeita.
Juan SE shaves
b. Juan se afeita a sí mismo.
Juan SE shaves DOM himself

Grooming verbs are the best defined class of so-called 'naturally reflexive verbs', which are
those predicates which lack the expectation that the two arguments –the one corresponding to
agent and the one corresponding to patient– should be referentially disjoint (Schäfer 2012).
Other classes of verbs that tend to be inherently reflexive are posture verbs –a class less

32
 SE IN SPANISH

productive in Spanish but also found (levantarse 'to stand up', sentarse 'to sit down', acostarse
'to lie down', tumbarse 'to lie down'...), which in principle define also actions that one
necessarily performs by moving body parts of one 's own, that is, that is performed on oneself–
, and some verbs of displacement, such as mudarse 'to move', moverse 'to displace', desplazarse
'to displace', or internal bodily motion, such as agitarse 'to shake', removerse 'to fidget',
revolverse 'toss and turn'. The terms endo-reflexive (Haspelmath 1987) and autocausative
(Creissels 2006) have been used to refer to this last class of predicates, which denote events
that –with animate subjects– are triggered volitionally but whose action falls on the subject.
In languages that have a two-form reflexive system, like Dutch, where there is no
pronominal doubling, these verbs reject the self form, unless contrasted.

(81) a. Jan waste zich. (cf. 77a).

Jan washed SE
'Jan washed'
b. ??Jan waste zichzelf (cf. 77b).
Jan washed himself

In other words: the self-form, as diagnosed in these languages, is the one that is used with
verbs where one does not expect a reflexive construal. If the verb, given its meaning, makes us
expect reflexivisation, the self-form cannot be used –unless one uses it for contrast–. The
conclusion would then be that SE is not a reflexivisator, because if it was (81a) would be as
marked as (81b), and the real reflexivisator is SE, zich in Dutch.
The question is whether Spanish shows the same effect as Dutch not in terms of the contrast
that can be inferred from the presence of the double, but in terms of the mismo form being used
precisely in the cases where the predicate would otherwise be naturally interpreted as involving
arguments with disjoint reference, and related to a higher degree of agentivity than the form
without the double. What decision should be taken here is far from obvious; in fact, De Benito
Moreno (2015), in her detailed study of SE marking across varieties in Spain, proposes that
many of the reflexive predicates of this type should be considered as instances of
deobjectivation where the direct object is in fact removed, rather than cases of real reflexivity.
Several empirical properties suggest that the mismo form is the one that reflexivises in
Spanish, not SE. Beyond the obvious fact that mismo means 'same' and thus is a natural
semantic candidate to express co-reference, we have a number of empirical facts that point
towards this conclusion.

a) There are plenty of cases, reviewed above, where the predicate is semantically reflexive
without SE provided a pronoun with mismo appear.
b) The addition of mismo allows a reflexive interpretation of non-anaphoric third person
pronouns, as in the following examples. Note that the reflexive pronoun in Spanish is generally
characterised as having two forms: the form SE, which appears in object contexts –in theory,
never as a nominative pronoun, but see §7.5 below– and the oblique form sí. The interesting
property is that mismo is compatible both with the reflexive form and the non-reflexive oblique
form (82a), (82b). In contrast, absence of mismo would make the pronominal be interpreted
with disjoint reference (82c), but addition of this adjective forces the reflexive reading.

(82) a. Pedro siempre piensa en él mismo.

Pedro always thinks on him self
b. Pedro siempre piensa en sí mismo.
Pedro always thinks on him self
c. *Pedroi siempre piensa en éli.

33
 ANTONIO FÁBREGAS

Pedro always thinks of him

c) The most relevant two classes of monotransitive predicates that allow naturally a reflexive
interpretation without the double are grooming verbs and some verbs of perception that
presumably do not introduce the expectation that the person perceiving something is
necessarily perceiving something other than himself.

(83) a. Juan se mira.

Juan SE looks
b. Juan se ve en la foto.
Juan SE sees in the photo
c. Juan se escucha cuando habla.
Juan se listens when he.talks

In the realm of ditransitive predicates, the reflexive interpretation without the double is
natural in verbs that can be naturally interpreted as benefactive, also involving in some form
the personal care of the agent to satisfy its own needs.

(84) a. Luis se puso un vaso de agua.

Luis SE poured a glass of water
b. Juan se sirvió una copa.
Juan SE served a drink
c. Pedro se hizo la cena.
Pedro SE made the dinner
d. Juan se quitó la ropa.
Juan SE removed the clothes

Beyond these classes of predicates, SE constructions without the double are not naturally
interpreted as reflexives, with some lexical variation difficult to pin down (eg., 85a is
marginally better than the rest for some speakers).

(85) a. ?Juan se critica.

Juan SE criticises
b. ??Juan se contrata.
Juan SE hires
c. ??Juan se entregó una carta.
Juan SE gave a letter.
d. ??Juan se escribió una carta.
Juan SE wrote a letter
e. ??Juan se besó.
Juan SE kissed

d) In the presence of auto-, the double sounds redundant even in cases of strong contrast.

(86) Juan se autocritica (??a sí mismo).

Juan SE self-criticises DOM himself
(87) La carta se autodestryó (*a sí misma).
the letter SE self-destroyed DOM itself

34
 SE IN SPANISH

With a lot of variation, the double can also feel marked in some of the naturally reflexive
predicates we have mentioned above, but not in all of them.

(88) a. ??Luis se quitó la ropa a sí mismo.

Luis SE removed the clothes to himself
b. ??Luis se hizo la cena a sí mismo.
Luis SE made the dinner to himself

This takes us to the problem of whether the double is doing something different from what
one would associate to reflexive SE constructions without the double, beyond the contrast that
is visible in the cases where the double can be present. There are three conceivable possibilities
to address the contrast between (89a) and (89b), and as a result explain (82-88).

(89) a. Luis se afeitó.

Luis SE shaved
b. Luis se afeitó a sí mismo.
Luis SE shaved DOM himself

a) The double only acts as a reflexiviser. This is the conclusion that one has to draw from
(82-88). The pair in (89), for the case of Spanish, is a fake minimal pair where only the second
(89b) is really reflexive. This means that the sentence in (89a) is in fact close or completely
assimilated to the anticausative verbs (90), where the external argument is not really interpreted
as an external agent and the crucial interpretation is that the subject initiates the event by its
own capacities and properties without the intervention of an external causer, which as we will
see is the meaning that anticausatives generally associate to their sentences. The sentence in
(89b), in contrast, treats the external argument as an external agent that sets in motion an event
that coincidentally has himself as a patient.

(90) La televisión se rompió.

The TV SE broke
'The TV broke'

This involves treating grooming verbs without the double on a par with another class of
naturally reflexive verbs, verbs of translational motion (Schäfer 2012), where the version
without the double would probably be anticausative despite the interpretation where the subject
initiates the movement volitionally.

(91) a. Juan se desplazó.

Juan SE displaced
b. ?Juan se desplazó a sí mismo.
Juan SE displaced DOM himself
(92) a. Juan se cambió de sitio.
Juan SE changed of place
b. ?Juan se cambió de sitio a sí mismo.
Juan SE changed of place DOM himself

Extending things even more, this would allow treating inherently reflexive verbs as verbs
which impose the entailment that the theta roles are assigned to the same entity. They always
reject the double because in their theta-grid there are no two distinct positions to assign both
theta-roles. We will get back to these verbs in §6.3.

35
 ANTONIO FÁBREGAS

(93) Juan se arrepintió.

Juan SE repented

This means, then, that the two examples, with and without the double would potentially
have different structures.

b) The second interpretation is that both structures, with and without the double, are
reflexive, sharing the same structure, but the double imposes additional requisites which
delimit the conditions under which it can appear.
In many approaches, such as Chierchia (2004), what happens with the double in examples
like (82-88) is that the double imposes an agentivity condition that is not compatible with all
verbs. Structures without the double are compatible with interpretations where the subject
initiates a process that affects himself, and where therefore the subject is not strictly an agent,
but rather an effector which is not external to the event (see §4.3 below for more details about
this). In the case of structures with the double, the external argument has to be interpreted as
keeping the interpretation that it is an external agent of the eventuality, and the eventuality is
not initiated by its internal properties. In other words: the double forces an agentive reading,
where being an agent means that one is in principle external to the change that is triggered into
the patient. From here it follows that predicates that generally happen on the same entity that
initiates them will differ from predicates that one normally assumes to happen on an entity
distinct from the agent.
This makes predicates that are naturally reflexive less compatible with the double because
in them the internal argument ends up being the subject. The double is acceptable then only
when, despite this syntactic movement or the semantic interpretation of reflexivity still allows
that subject to be interpretable as an agent. Koontz-Garboden (2009), for instance, adopts this
view, and argues that the incompatibilities have more to do with the nature of the pronoun that
the self-form combines with, that tends to impose an animacy requirement, than with the nature
of reflexivisation itself.
In this sense, the double is preferred with predicates which provide the expectation that the
external argument is, from the perspective of the entity acting as a patient, external to the event,
and volitionally initiates a process that normally would fall on a different entity. The cases
where the double is not possible or very marked are those where the predicate supposes that
the agent is not external to the effect that the event has on the patient, because generally the
event is initiated by and acted on the same entity. The contrast in (89) is simply one where the
presence of the double cancels the expectation that the eventuality should have fallen on the
same entity that initiates it, but both are reflexive.

c) The double introduces a notion of contrast, not of agentivity. Like (b), reflexivity is not
related to the double, and the asymmetries depend on an additional piece of information
introduced by the double, but that piece is not the expectation that there is an external agent.
The problem is that these doubling forms are bigger than clitics, phonologically tonic, and as
such play a role on information structure in the form of introducing contrast. Verbs that expect
that the two arguments are distinct prefer having the double because in that case the double
contrasts the reflexive argument with other disjoint arguments that could have appeared. Verbs
that do not like the double are those where contrast would be redundant in the case of a
reflexive, because the verbs expects reflexivity already. (89) is simply two informationally-
different structures of a reflexive, the first used in an unmarked context where one expects joint
reference and the second used in a context where one could have expected a disjoint construal.

36
 SE IN SPANISH

d) A final, perhaps less popular theory, treats the double as completely dissociated from
agentivity and even contrast. In this proposal (cf. for instance Charnavel 2019, Alcaraz 2021),
mismo marks logophoric pronouns that must be coreferential with the entity to which the claim
made in the clause is attributed –Sells' (1987) logophoric center or pivot–, which can be distinct
from the agent and the speaker of the clause. Like that, Alcaraz (2021) explains cases such as
(94), where the mismo element is coreferential with an element introduced in a previous clause.

(94) Zarcoi ha reconocido que ese rendimiento le ha sorprendido a sí mismoi.

Zarco has admitted that that performance him has surprised to himself
'Zarco admits that his performance has surprised himself'

In this theory, the double would be a general reflexive marker whose condition of
appearance would not be conditioned by the thematic properties of the predicate or the
likelihood that the predicate would involve distincts subject and object or not, but rather by the
possibility of establishing the relevant logophoric relation between the center of the clause and
the relevant argument; the dissociation from SE would be maximal, as seen in (94), where SE
is not even present despite the anaphor involving in principle a dative.

Beyond these options, the presence of the double proposes in fact a reassessment of the
analysis that was presented in (59) above. From the perspective of Labelle (2008), that analysis
has two problems: (i) the double is the one that seems to correlate better with the semantic
reading of reflexivisation; (ii) if the SE element is introduced in an argument position and the
internal argument moves, where is the double introduced? Note that the double would carry
the same case or prepositional marking as the argument that hypothetically has moved, if there
had been no movement. This is problematic for two reasons. First of all, the verb should be at
least partly unaccusative according to the analysis in (59). Second, the position where the
double mismo gets case in (95b) should be the same where the object is in (95a), before it
moves up to become the subject in (95b).

(95) a. Juan atacó a su hermana.

Juan attacked DOM his sister
b. Juan se atacó a sí mismo.
Juan SE attacked DOM himself

Labelle's (2008) proposal is that SE –in reflexives, like in anticausatives and passives– is
the spell out of a voice head, where Voice is close to what Embick (2004) means by vP, because
Voice is the head where the external argument is introduced –following Kratzer (1996)–. The
crucial property of the Voice head spelled out as SE is that the verbal predicate it combines
with must be open, that is, it should have an unsaturated argument position, so that the SE-
voice can allow the interpretation that the unsaturated argument position, a variable, covaries
with the argument introduced as the external argument.
The presence of the double, in fact, does not semantically saturate the predicate, according
to Labelle (2008): she is neutral to why this is the case, but suggests that the structure of the
self-form might lack D projections that make it saturate the variable, or that the semantic
reading composes the double with the predicate through restriction (Chung & Ladusaw 2003).
Be it as it may, the double is located in the syntactic position of the argument but does not have
semantic effects on the theta-grid of the predicate or, alternatively, that position is open. (96a)
presents the analysis with the double, and (96b) without it. The crucial difference with (59) is
that SE is not in an argument position and there is no movement of a second argument, which
means that this analysis does not treat reflexives as unaccusatives and has to explain the choice

37
 ANTONIO FÁBREGAS

of auxiliaries that motivated unaccusative analyses as deriving from some other property, for
instance a reduced valency in the predicate.

(95) a. VoiceP

Juan Voice

Voice VP
SE
V Double (= variable)
a sí mismo

b. VoiceP

Juan Voice

Voice VP
SE
V

Labelle (2008), then, would have to adopt positions (b) or (c) for the pairs with and without
the double, including crucially the idea that SE is not a reflexiviser. When it comes to why the
SE form only appears when the open argument position is one that would have resulted in an
accusative or a dative case marked argument, Labelle suggests that SE can check those cases
–and only those cases– from the verb. This aspect of her analysis is not very clear, given that
the double would still syntactically receive that case, and therefore it is unclear why SE would
need to absorb that case if the double is present; remember also from §2 above that SE does
not make case distinctions, a property that does not grant the conclusion that SE receives case.
This case absorbing property, in fact, would not fare well with Spanish, because SE is used in
Spanish in impersonal sentences where accusative or dative case can be still assigned normally
to other arguments. In Spanish, SE should not be an element that checks the case of the verb,
although this is not a problem for Labelle (2008), as she is only talking about SE in French, a
language where SE does not have an impersonal use.
The conclusion of this section is the following: despite the historical origin of SE in Spanish,
there is no evidence that SE should be considered mainly a reflexivising function. SE does not
appear in many reflexive predicates, including any case where the co-referential argument is
not a direct object or an indirect object and any type of non-verbal structure. SE is not sufficient
to reflexivise a predicate in all cases. It is unclear under which conditions the double can be
added, but what is clear is that in some cases the reflexive reading is impossible without it.
SE appears (compulsorily) in reflexives only when three syntactic conditions are met:

a) The predicate has at least two arguments

b) One of the arguments is the subject, agentive or not
c) The other argument would be the direct object or the indirect object

Let us move now to reciprocal SE structures, which we will treat –like the vast majority of
the literature– as built over reflexive SE structures.

38
 SE IN SPANISH

3.2. Reciprocal SE
The general take on reciprocal constructions is that they should be viewed as a particular
subcase of the reflexive structure, where there are additional semantic requisites that are met
(Heim, Lasnik & May 1991, Pollard & Sag 1992, Everaert 1999, Dimitriadis 2004, 2007,
Maslova 2007). Even if some languages differentiate reciprocals and reflexives
morphologically (Nedjalkov 2007), Spanish uses the same set of clitics for reflexive and
reciprocal structures, which makes it plausible that reciprocity is an interpretation of reciprocal
structures that can be favoured or forced by specific constituents. Let us start with the evidence
that reciprocal SE is a subtype of the general reflexive SE and then let us move to the specific
semantic conditions that differentiate it.

3.2.1. Similarities with reflexive structures

In §3.1 we have seen that reflexive SE does not appear in any kind of reflexive predicate,
but only in those where reflexivity involves a verb and specifically the subject and the direct
or indirect object. This property also applies to reciprocal SE structures. (97a, b) shows that
reciprocal non-verbal predicates lack SE, and (97c-g) shows that SE appears in these cases also
when the subject and the direct / indirect object are involved; note in (97g) that if the two
individuals are contained in the object position and are disjoint with the subject, SE is
impossible.

(97) a. el parecido del uno con el otro

the resemblance of the one with the other
b. diferentes el uno del otro
different the one from.the other
c. Juan y María se miraron.
Juan and María SE looked
'Juan and María looked at each other'
d. Juan y María se dijeron cosas muy graves.
Juan and María SE said things very serious
'Juan and María said very serious things to each other'
e. Juan y María (*se) dependen uno del otro.
Juan and María SE depend one on-the other
Intended: 'Juan and María depend on each other'
f. Juan y María (*se) fueron a la fiesta uno con el otro.
Juan and María SE went to the party one with the other
Intended: 'Juan and María went to the party with each other'
g. Pedro (*se) juntó a Juan y María.
Pedro SE joined DOM Juan and María

As in the case of reflexive SE structures, the reciprocity –which we will provisionally define
as a mutual relation between two entities, performing the same event on each other– is not
directly associated to SE, but rather to a number of syntactic expressions that force the
reciprocal reading, and which can appear without SE in the cases where the syntactic relation
between the two involved arguments is not satisfied. Two of such expressions are invariable:
the adverb mutamente 'mutually' and pronouns introduced with entre 'between', the last of
which does not have a reciprocal reading when introducing other types of nominal constituents.
These elements generally combine naturally only in cases where the coreferentiality involves
the subject and the direct object, although the preposition entre has a broader range of context
where it can be used.

39
 ANTONIO FÁBREGAS

(98) a. Juan y María se odian mutuamente.

Juan and María SE hate mutually
'Juan and María hate each other'
b. ?Juan y María se dicen cosas horribles mutuamente.
Juan and María SE say things horrible mutually
c. *Juan y María hablan mutuamente.
Juan and María talk mutually
(99) a. Juan y María se pelearon entre {sí / ellos}.
Juan and María SE fought between themselves
'Juan and María fought against each other'
b. ??Juan y María se dijeron cosas horribles entre ellos.
Juan and María SE said things horrible between them
c. Juan y María hablan entre ellos.
Juan and María talk between them
'Juan y María talk to each other'

The constituent which, however, has a broader range of syntactic contexts where it can be
applied is the sequence el uno P el otro, where the preposition is compulsory and is generally
selected by the verbal predicate.

(100) a. Juan y María se miraron el uno al otro.

Juan and María SE looked the one DOM-the other
b. Juan y María se dicen cosas horribles el uno al otro.
Juan and María SE say things horrible the one to-the other
c. Juan y María dependen el uno del otro.
Juan and María depend the one from-the other
d. Juan y María van a la fiesta el uno con el otro.
Juan and María go to the party the one with the other

As in the case of reflexives, absence of these elements might involve losing the reciprocal
reading; in (101) the sentence can be interpreted, at least, as reflexive or reciprocal, and it is
the doubles that would disambiguate.

(101) a. Juan y María se dicen cosas horribles.

Juan and María SE say horrible things
b. Juan y María se dicen cosas horribles a sí mismos. Reflexive
Juan and María SE say horrible things to themselves
c. Juan y María se dicen cosas horribles el uno al otro. Reciprocal
Juan and María SE say horrible things the one to-the other

We will not discuss here the nuances in the grammar of these constructions in terms of
variation in gender, number and definiteness (cf. RSAE & ASALE 2009: §16.5g-k), beyond
the fact that the presence of the preposition is compulsory in these structures, even in cases
where the predicate would not compulsorily combine with a preposition. RAE & ASALE
(2009: §16.5n) offers the following minimal pair.

(102) a. Una rama del árbol tocaba (a) al otra.

a branch of.the tree touched DOM the other

40
 SE IN SPANISH

b. Las ramas del árbol se tocaban unas *(a) las otras.

the branches of.the tree SE touched one DOM the others
'The tree branches touched each other'

However, there are no cases in which the preposition is present in a context where the verb
would not allow presence of that preposition. This, like in the case of reciprocal structures,
suggests that the syntactic position where el uno P el otro is introduced should be related to the
base syntactic position where the argument would have been introduced in a disjoint reference
construction. Parallelisms like (103), and the absence of cases where a preposition not allowed
by the predicate is introduced, suggest that the reciprocal double should be related to the same
position in (103b) as the disjoint object in (103a).

(103) a. Pedro habló con María.

Pedro talked to María
b. Pedro y María hablaron uno con el otro.
Pedro and María talked one with the other
'Pedro and María talked to each other'

In the case of reflexives we noted that one prefix, auto-, forces that reading into the
predicate. The same goes for reciprocals, in this case with the prefix inter- (Feliu 2003); as in
the case of reflexives, the presence of this prefix does not preclude presence of SE, which is
blocked under the same syntactic conditions.

(104) a. Juan y María se comunican.

Juan and María SE communicate
b. Juan y María se inter-comunican.
Juan and María SE inter-communicate

Just like there are naturally reflexive verbs, the notion of 'naturally reciprocal' verb has also
been proposed, sometimes under the label 'symmetric verbs', which are verbs whose semantics
implies that the event described should involve mutual participation of two or more entities
(Lakoff & Peters 1969, Levin 1993), forcing a non-distributive reading where the entities
involved in the event do not establish a relation with external entities. These verbs include
verbs of mutual contact (105), union or separation (106, 107), interpersonal relations (108) and
relations of similarity and difference (109). In all these cases, the reciprocal reading is
interpreted by default, without the need of adding any double. Note that SE is not always
present in this cases, and in (107), for instance, it should rather be viewed as an anticausative
marker (see §4 below).

(105) Juan y Pedro chocaron (uno contra el otro).

Juan and Pedro hit one against the other
(106) El agua y el aceite ligan uno con el otro.
the water and the oil combine one with the other
(107) Pedro y María se divorcian.
Pedro and María SE divorced
(108) Pedro y María conversan.
Pedro and María talk
(109) Pedro y María se parecen.
Pedro and María SE look-alike

41
 ANTONIO FÁBREGAS

Like in the case of reflexives, the reciprocal construction requires that one of the theta roles
is reduced, descriptively speaking, which in this case results in the property that the referential
expressions must be conjoined under the same syntactic constituent (110), not each one located
in a different constituent (111).

(110) a. Juan y María se miraron uno al otro.

Juan and María SE looked one to.the other
b. Juan miró a María.
Juan looked to María

Let us now talk about the semantic conditions that differentiate reciprocals from reflexives
(Fiengo & Lasnik 1973, Dougherty 1974, Langendoen 1978, Higginbotham 1980, Roberts
1987, Darlymple, Kanazawa, Mchombo & Peters 1994). The informal definition of reciprocals
is that they express a mutual relation between at least two different entities, such as each one
of those entities relate to all the other entites by the same relation, expressed by the predicate.
This is known as 'strong reciprocity', and (111) illustrates it: there must be at least 2 distinct
individuals, and each one of those individuals relate to all other individuals in the set by the
same relation, in this case 'know him or her'.

(111) Todos los estudiantes de la clase se conocen unos a otros.

all the students of the class SE known each other

(112) |A| ≥ 2 and ∀x,y ∈A(x≠y → Rxy)

('given a set whose cardinality value is equal or higher than two, for every pair of x
and y members belonging to the set, if they refer to distinct entities, then the relation
R must hold between x and y')

[Langendoen 1978]

Trivially, this imposes, in contrast to reflexivity, the condition that the argument position is
occupied in reciprocal sentences by constituents that, syntactically or semantically, denote
groups of two or more people. This includes, of course, coordinated constituents as most of the
examples provided above, plural nominal constituents (111) or collective nouns in the singular
(113).

(113) La pareja se peleó.

the couple SE fought

However, strong reciprocity is not obtained in many other cases which could be classified
as syntactically reciprocal given the presence of the relevant constituents.

(114) a. Los postes están separados diez metros los unos de los otros.
the posts are separated ten meters the ones from the others
'The posts are separated ten meters from each other'
b. Los diputados se pegan unos a otros.
the congressmen SE hit one DOM other
'The congressmen hit each other'

In (114a) we can have –and this is the most natural reading– a linear relation between posts,
such as that each two adjacent posts are separated from the other by ten meters; obviously the

42
 SE IN SPANISH

third pole and the first pole would then be separated by 20 meters: the relation only applies to
two adjacent members in a sequence that is either linear or can be conceptualised,
geometrically, as a set of interconnecting lines, as in the example (115) (RAE & ASALE 2009:
§16.5r):

(115) Las hojas de los árboles se tocaba unas a las otras.

the leaves of the trees SE touched one DOM the other
'The tree leaves touched each other'

The intuition is that in this type of linear(isable) disposition, two entities relate to each other
in the same way directly, or relate to each other indirectly by the intermediation of third entities
that separate them in the linear organisation. Langendoen (1978) calls this 'intermediate
reciprocity', but in the literature it is more common to call this simply 'weak reciprocity'.
The example in (114b) is also not strong reciprocity, because its truth conditions are satisfied
if, given 6 congressmen, it is not the case that each one of the six hit the other five, but rather
that at least during the fight the same congressman is both hit and hits another congressman.
This type of reading is called 'partitioned strong reciprocity' in Langendoen (1978), because it
involves partitioning the set of a cardinality value higher than 2 into subsets involving possibly
different pairs.
There are other definitions of subtypes of reciprocity that do not directly conform to the
strong definition of reciprocity (see in particular the overview in Darlymple, Kanazawa,
Mchombo & Peters 1994), but we will not discuss them. What is relevant for us in this state of
the art is, however, that the vagueness that they presuppose in the interpretation assigned to a
reciprocal structure strongly suggests that reciprocity should not be viewed as a different type
of SE structure –at least for the languages, like Spanish, that do not differentiate reciprocals
and reflexives with different sets of clitics or morphemes–.
The so-called reciprocal SE is rather a particular type of coreferential interpretation made
available by some predicates when they use a reflexive syntactic structure that involves a plural
entity. This is suggested by the vagueness of structures like Se hacen mucho daño 'they SE give
a lot of pain', which in the absence of doubles like a sí mismos 'to themselves' or el uno al otro
'each other' could be interpreted as either structure. As far as SE is concerned, SE does not
relate to reciprocal readings just like it does not relate to reflexive readings, and it is the doubles
that can be responsible for imposing that interpretation.
It seems, then, that the same general operation –leaving one theta role position open, or
insufficiently saturated– that produces a reflexive structure can also be used for a reciprocal
structure: the double can appear in that position, and the two (or more) entities involved in the
interpretation must appear in the same syntactic position, the subject in the case of SE-marked
structures. The same options that we pointed out for the analysis in the case of reflexives, then,
can be extended to the case of reciprocals, and the same problems noted there would apply to
these cases.
To summarise this §3 section, in the case of reflexive and reciprocal SE we could establish
the following generalisation:

(116) SE in reflexives and reciprocals is related to the supression or insufficient saturation of a

direct or indirect object

It is possible to connect (116) with SE as a clitic as shown in §2.2-§2.4. Why would the
indirect or direct object position play a role in whether SE is present or not? From the
perspective of the clitic sequences in §2.4, one plausible answer to this is that direct and indirect
objects are the only two syntactic functions that in Spanish are marked with clitics, which in

43
 ANTONIO FÁBREGAS

turn only appear in verbal structures. The realisation of SE as an element that marks the
potential suppression of a theta position –or its insufficient saturation– is restricted to those two
syntactic functions precisely because they are the only two that have access to the clitic area
and therefore the ones that would form some kind of syntactic sequence of projections with
SE.
Let us leave the discussion of the first type of SE here and move now to the second predicate-
affecting SE: anticausative.

4. Predicate-affecting SE uses (2): anticausative

The so-called anticausative SE –which in other terminologies gets the perhaps confusing
name of 'middle SE'– is the use of SE where the presence of the clitic marks the so-called
inchoative, non-causative or non-externally caused version of an otherwise causative predicate.
For this reason, to talk of anticausative SE one presupposes that the same verb without SE can
be used, in a causative construal where there is an externally caused event that affects what
would be the direct object in the SE marked version.

(117) a. Juan movió la mesa. Causative

Juan moved the table
b. La mesa se movió. Inchoative
the table SE moved

Roughly, the interpretation of (117a) is 'John made the table undergo some movement' and
(117b) is 'the table underwent some movement on its own'.
The verbs of the shape in (117a) are transitive, and the verbs of the class of (117b) are
generally analysed as unaccusative for the case of Spanish, and also in other languages.
However, as it was the case with reflexive predicates (§3.1), the unaccusative nature of
anticausatives is far from clear and quite problematic. Anticausative predicates have a strong
tendency to be interpreted in their causative version in absolute participle structure: Rota la
televisión, Juan fue a la tienda a comprar otra 'Broken the TV, Juan went to the shop to buy
another one' is rather interpreted as 'Once someone broke the TV...' than as 'Once the TV broke'.
Importantly, moreover, anticausatives do not allow bare nominals: Se rompieron ventanas 'SE
broke windows' can be interpreted as a passive SE, but not as an anticausative SE. Perhaps, as
in the case of the reflexive SE, this can be analysed as involving movement of the internal
argument to a specifier position, which would not be the case for the passive SE structure. We
will get back to this problem, and discuss it more in detail, when we compare the anticausative
and the passive structures in §8.1, and leave it for the moment.
There is a huge body of work about anticausative structures, the relations between the
causative and the anticausative version and the correct notion of 'causation' that should be
involved in such cases (see, just to name a few, Jespersen 1927, Nedjalkov 1969, Fillmore
1970, Smith 1970, Chierchia 1989, Labelle 1992, Haspelmath 1993, Levin 1993, Levin &
Rappaport-Hovav 1995, Reinhart 2000, Doron 2003, Alexiadou & Anagnostopoulou 2004,
Dobrovie-Sorin 2006, Schäfer 2008, 2009, Koontz-Garboden 2009, Labelle & Doron 2010,
Rappaport-Hovav & Levin 2010, Lundquist 2011, Mendikoetxea 2012, Vivanco 2015, to name
just a few examples).
From the perspective of Spanish, there are some specific problems that deserve to be
highlighted, beyond the complexity of the uses of SE.

a) First of all, not all verbs that allow the semantic and syntactic alternation between
causative and anticausative take SE in Spanish. From this perspective, anticausative SE is
lexically conditioned. Remember that reflexive SE is largely not lexically conditioned, but

44
 SE IN SPANISH

rather one could establish a reasonably solid set of syntactic constraints that condition its
presence –eg., the subject must be coreferential with the direct or indirect object–; this set of
factors is not enough for the case of anticausative SE.

b) Secondly, the directionality between the causative and the anticausative pair is not easy
to diagnose because the syntactic / semantic evidence does not match well with the surface
marking introduced by SE. Spanish SE marks the anticausative version, not the causative
version. Most initial analyses of verbs conforming to the semantic alternation between a
causative and a non-causative construal proposed a directionality like in (118), where the
unaccusative version is more basic (syntactically and semantically) and the causative is built
on top of it –for instance, as in Dowty (1979), by introducing a cause subevent–.

(118) non causative ---> causative

However, the directionality in (118) is at odds with the morphological marking in Spanish:
the extra element, SE, appears in the more basic structure according to that proposal, and
disappears when the more complex structure is built. This is in fact the reason why this type of
SE is called anticausative, suggesting that it has been added to the causative version in order to
derive from it its 'opposite', a version where there is no possible place for an external argument.
And yet, anticausative SE is very frequent in Spanish, with literally hundreds of cases –see
Vivanco (2015)–, in contrast to other lexical uses, like antipassive or the so-called aspectual
with intransitive predicates, which is much more restricted.
This apparent mismatch between the semantic relation and the formal marking has made
several authors –most significantly, Koontz-Garboden (2009)– propose that in languages using
SE to mark anticausatives one has to derive the anticausative construal from a reflexive
structure; we will get back to this proposal later.

4.1. Anticausative SE and the lexical properties of the predicate

Vivanco (2015), without any doubt, has the most detailed and updated overview of this
problem for the case of Spanish. In this section and in the rest of this section we will be
following closely the observations made in her work.
As this author explains, all anticausative SE verbs in Spanish belong to the broad class of
change of state verbs, with three subtypes:

i) Change of a physical property

(119) a. El sol secó la ropa.

the sun dried the clothes
b. La ropa se secó.
the clothes SE dried

ii) Change of mental property

(120) a. La noticia entristeció a Eulogio.

the news saddened DOM Eulogio.
b. Eulogio se entristeció.
Eulogio SE saddened

45
 ANTONIO FÁBREGAS

iii) Change of location

(121) a. Un iceberg hundió el Titanic.

an iceberg sank the Titanic
b. El Titanic se hundió.
the Titanic SE sank

In all three cases, the SE version is interpreted as marking an event that happens due to the
internal properties, conditions or capacities of the internal argument, without being triggered
or initiated by an external participant.
In fact, as an additional argument that anticausative SE is closely related to change of state
structures, we have the fact that the Spanish equivalents of BECOME as a verbal function
generally are marked with SE, even when the is causative equivalent is expressed with a
different root, as in (125):

(122) a. Juan se puso enfermo.

Juan SE put sick
'Juan got / became sick'
b. El mal tiempo puso a Juan enfermo.
the bad weather put DOM Juan sick
'The bad weather made Juan sick'
(123) a. Juan se hizo cristiano.
Juan SE made Christian
'Juan became Christian'
b. El accidente hizo a Juan cristiano.
the accident made DOM Juan Christian
'The accident made Juan Christian'
(124) a. Juan se volvió loco.
Juan SE turned crazy
'Juan got crazy'
b. El trabajo volvió loco a Juan.
the work turned crazy DOM Juan
'The work made Juan crazy'
(125) Juan se quedó ciego.
Juan SE became blind
'Juan became blind'

As Vivanco (2015) notes, the fact that anticausative SE is restricted to (some) change of
state verbs should be related to the presence of a property with a gradable set of values or path
of movement, two notions that have the common notion of scale underlying to it. In her
proposal, in fact, the presence of SE presupposes that the dimension that the verb uses to
quantify the change of state is scalar, a property to which we will come back in §5.5, as it
connects anticausative SE with aspectual SE.
The scalarity is not the only semantic notion that has been proposed as underlying SE verbs
with anticausative semantics: Haspelmath (1993) proposes that anticausative versions are only
possible with verbs that denote changes of state that human cognition can conceptualise as
happening spontaneously, without the need for an external agent or causer that initiates that
change independently of the affected entity –hence the name 'middle' that was traditionally
assigned to them–. According to this dichotomy, the causative version of these verbs denotes
an externally caused change of state, while the anticausative version expresses an internally-

46
 SE IN SPANISH

caused change of state, where the event happens 'spontaneously' just by virtue of the internal
properties of the subject, without the intervention of an external participant. Unaccusative verbs
(Levin & Rappaport 1995) would be verbs that denote internally-caused events, and the same
goes for purely intransitive verbs like 'play' or 'talk'. Transitive verbs would be, in essence,
those that require an external cause to make conceptual sense, given the linguistic
conceptualisation.
Vivanco (2015: 100) shows that in this sense one could propose a scale of spontaneity
(Haspelmath 1987, Creissels 2006, Heidinger 2014) that orders the conceptual value of
individual change of state predicates according to how spontaneously the change expressed by
them can happen (cf. 126, adapted from Vivanco's 2015: 100, example 11).

(126) asesinar 'murder' - spontaneous

calentar 'warm'
romper 'break'
envejecer 'to become older'
nacer 'to be born' + spontaneous

The idea is that verbs that are closer to the extremes do not alternate between a causative
and the non causative version: a verb that is conceptualised as a purely spontaneous change
like 'to be born' would not have a causative version, and a verb with a purely non spontaneous
change of state like 'murder' would not have an unaccusative version. Verbs in the middle differ
with respect to how broadly and how naturally they allow one or the other construal. Many
speakers reject (127a) in the anticausative reading, while others accept it, and the same goes
for the causative (127b).

(127) a. #La leche se calentó.

the milk SE warmed
Intended: 'the milk got warm'
b. ??El trabajo envejeció a Juan.
the work made.older DOM Juan
Intended: 'the work made Juan older'

Interestingly, when the verb does not alternate, anticausative SE might appear, depending
on what one assumes about the relation between so-called inherent SE and anticausative SE
(§6.3 below). The problem, as in many approaches that propose more or less flexible scales
with different degrees of a given property, is to determine the criteria that should be adopted to
assign a specific degree in the scale to each predicate.
One option that has been argued to be relevant to operationalise the distinction is 'manner'.
If the notion of scale or scalar change underlies the concept of unaccusative change of state /
anticausative SE structure, according to Levin & Rappaport (1995), the notion that underlies
verbs that do not allow the anticausative version is 'manner': the concepts expressed by the
verbs are defined as different manners of performing an event, or as introducing a significant
component of manner, which requires being controlled by an external agent. This would
explain, for instance, that matar 'kill' can alternate, because it lacks a manner component, while
guillotinar 'to guillotine' does not alternate, because it specifies the manner in which the killing
happened.

(128) a. Juan mató a Pedro.

Juan killed DOM Pedro
b. Pedro se mató en el accidente.

47
 ANTONIO FÁBREGAS

Pedro SE killed in the accident

'Pedro died in the accident'
(129) a. Juan guillotinó a Pedro.
Juan guillotined DOM Pedro
b. #Pedro se guillotinó en el accidente.
Pedro SE guillotined in the accident
Intended: 'Pedro died by guillotine in the accident'

However, even when the verb can alternate, the presence of SE is not granted. Vivanco
(2015) identifies three situations where an alternating verb lacks anticausative SE marking. We
will discuss the first and the second here, and leave the third –presence of SE only with some
internal arguments– to section §5.5, as it is plausibly related to the notion of scale.
The first one involves alternating verbs that even when used as anticausatives reject SE –
Vivanco (2015: 174) notes that there can be dialectal variation with respect to this, so her data
reflect the Madrid variety, and we should in fact expect some degree of variation across
geographical and diachronic varieties–.

(130) adelgazar 'to get thin', aumentar 'to increase', blanquear 'to become white' (and many
other verbs in -ear), disminuir 'to decrease', empeorar 'to worsen', engordar 'to get fat',
enfermar 'to get sick', enloquecer 'to get crazy', enmudecer 'to become mute', envejecer
'to become old', hervir 'to boil', rejuvenecer 'to become young', resucitar 'to resurrect',
cambiar 'change'

The second one involves verbs that in the anticausative version can have SE or not, for the
same speakers. Again, there is dialectal and even idiolectal variation, but for the Madrid variety
the examples volunteered by Vivanco (2015: 57) include the following.

(131) cuajar 'to become solid', enmohecer 'to get moldy', ennegrecer 'to become black',
mejorar 'to become better', reventar 'to explode'

The idea, in any case, is that there seems to be the need to introduce a lexical component in
the analysis of verbs that take anticausative SE. In the subsections related to aspectual SE
structures we will see a few proposals that have attempted to find a specific rule for the presence
of SE in only some anticausative verbs, but not all of them, but we will see that they require
access to lexical information that is not always clearly reflected in the syntactic structure. As
we will see, these other ways of predicting the presence or absence of SE involve
deconstructing the notion of aspectual SE, proposing –among other things– that the presence
of SE with anticausatives has something to do with the aspectual information.
Beyond this, we would like to highlight that just like in the case of reflexive SE, which was
not compulsory to express reflexive meanings, this lexically-conditioned distribution of
anticausative SE means that anticausative SE is not compulsory to express the anticausative
version of an otherwise causative predicate.

4.2. Syntactic conditions: theories that eliminate the agent from the anticausative version
Let us now examine the syntactic conditions and the proposals about the syntactic structure
of the anticausative SE structures. The first property that we will highlight is pretty trivial, but
connects it directly with reflexive SE: the anticausative SE verbs invariably involve bases that
corresponded to transitive verbs, where superficially the argument that one expects to find as
the direct object of the causative version becomes the subject of the anticausative version.

48
 SE IN SPANISH

(132) a. El sol derritió el hielo.

the sun melted the ice
b. El hielo se derritió.
the ice SE melted

Interestingly, though, the anticausative SE is never used when one could argue that the
indirect object of the causative version becomes the subject of the 'anticausative version'.
Consider the two sentences in (133).

(133) a. Juan le dio el libro a Pedro.

Juan him gave the book to Pedro
b. Pedro recibió el libro.
Pedro received the book

In (133a) we have a verb that could have been glossed as 'to make someone receive
something', so (133b) could be seen as its anticausative version. However, Spanish does not
use in this case (or other similar conceivable cases) SE: a different root is used –the so-called
'suppetive alternation' to express the contrast between causative and unaccusative, cf. Spanish
entrar 'to enter' and meter 'to make something enter'–. Why cannot (134) be used to replace
(133b)? The fact is that anticausative SE is restricted to alternations where the direct object,
and no other argument, is promoted to the subject position, in fact just like we saw in the case
of reflexives and we will again see in the case of passive.

(134) *Pedro se dio el libro.

Pedro SE gave the book
Intended: 'Pedro got the book'

Beyond this, many of the internal properties of anticausative SE differentiate it from passive
SE (cf. §8.1). While passive construals allow an external agent introduced as a prepositional
complement (135a), anticausative structures do not allow it. The reason for several authors
(Reinhart & Siloni 2005, but see Labelle & Doron 2010) is that the anticausative structure does
not project syntactically a position for the external agent that can be recovered by the
prepositional complement, and for others that the external agent is semantically incompatible
with the internal causation associated to anticausatives (Schäfer 2008).

(135) a. El barco fue hundido por los alemanes.

the boat was sunk by the Germans
b. *El barco se hundió por los alemanes.
the boat SE sank by the Germans

In Spanish, the internal causation of anticausatives can be tested by its compatibility with
por sí mismo 'by itself', por sí solo 'by it alone' or solo 'alone, without the help of anyone'.

(136) a. El barco se hundió por sí mismo.

the boat SE sank by itself
b. *El barco fue hundido por sí mismo.
the boat was sunk by itself

49
 ANTONIO FÁBREGAS

What anticausatives allow are causative complements, which Schäfer (2008) uses to argue
that they also involve external arguments, even if they cannot be agentive –remember that the
subject could be a causer, only that an internal causer–.

(137) El barco se hundió por la tormenta.

the boat SE sank by the storm
'The boat sank because of the storm'

Other tests show that anticausative structures lack agentivity: impossibility of combining
the predicate with an agent-oriented adverb (137) and impossibility of licensing a final clause
with anticausatives (138).

(137) #El barco se hundió queriendo.

the boat SE sank willingly
Intended: 'The boat sank on purpose'
(138) #El barco se hundió para cobrar el seguro.
the boat SE sank to get the insurance
Intended: 'The boat sank in order to get the insurance money'

The absence of an agent entailment in these cases has been related to the so-called ‘dative
causer’ construction (cf. Cuervo 2003, Schäfer 2008: 41-45). Dative causers are non-
argumental datives that get interpreted as the human entity that sets in motion an event without
involving any kind of volitionality, where the notion of 'set in motion' is quite flexible and
might involve cases where it is difficult to imagine some kind of intervention of the subject
into the event. (139b) illustrates one prototypical case for Spanish, where the dative causer has
been added to the anticausative SE structure in (139a). (139c) is one case where it is difficult
to conceive such intervention, even without willful intent, and the affected entity interpretation
seems to be more salient.

(139) a. Se rompió la televisión.

SE broke the TV
‘The TV broke’
b. A Luisa se le rompió la televisión.
to Luisa SE her.dat broke the TV
‘Luisa involuntarily broke the TV’
c. A Pepe se le ha roto la bañera.
to Pepe SE him has broken the bathtub

The anticausative structures’ apparent rejection of external agents but compatibility with
causers, as we already advanced, can be interpreted in two ways, which are relevant for the
crucial question related to anticausative SE: whether the agent argument is eliminated from the
anticausative structure together with any other external argument or there is some implicit or
explicit external argument.
The general consensus is that the agent is in essence eliminated from anticausatives or not
projected in such structures; the exceptions to this general consensus will be presented in §4.3,
as they involve the so-called reflexivisation analysis of anticausatives. Beyond the apparent
impossibility of licensing final clauses or agent-oriented arguments, another factor that plays a
role in this discussion is the presence, precisely, of dative causers in some anticausative cases.
The initial assumption is that the dative causer is only licensed when the agent is not present in

50
 SE IN SPANISH

the configuration. As noted in Schäfer (2008) and Vivanco (2015), the dative at best receives
an affectedness reading in a transitive clause (140).

(140) a. Juan cambió la mesa de sitio.

Juan changed the table of place
b. A María Juan le cambió la mesa de sitio.
to María Juan her changed the table of place
‘Juan moved the table and that affected María’ (not ‘María unwillingly caused Juan to
move the table’)

Relevantly, the passive version of a transitive verb rejects the dative causer reading, while
the anticausative reading of (at least some) verbs allows it.

(141) a. Pedro fue despedido (por su jefe).

Pedro was fired (by his boss)
b. *A María le fue despedido Pedro.
to María her was fired Pedro
Intended: ‘María caused involuntarily that Pedro got fired’
(142) a. Se abrió la puerta.
SE opened the door
b. A María se le abrió la puerta.
to María SE her opened the door
‘María caused involuntarily that the door opened’

Prima facie, this suggests that passive and anticausatives contrast precisely in the presence
vs. absence of an agent argument, with important caveats that we will get back to in a few
moments. This position is explicitly advocated by Reinhart & Siloni (2006), who argue that the
passive is a syntactic operation that starts from a transitive predicate with an agent external
argument, while the anticausative is a lexical operation –assuming by ‘lexical’ that it is an
operation that applies before syntax, on individual lexical items– that inactivates the external
argument in a way that when that verb is projected in syntax it does not have an agent position
to fulfill: the agent, from the perspective of syntax, is not there.
The idea that there is no agent projected in anticausatives is quite cannonical in the literature,
but with important nuances. Schäfer (2008) adopts the view that passives and anticausatives
are different in terms of their Voice head. In his assumptions, Voice is interpreted as a head
distinct from little v; little v can be agentive or causative, among other possible values (see also
Alexiadou et al. 2006), and generally anticausatives would have a causative little v. However,
the main distinction between passives and anticausatives takes place at the level of Voice.
Schäfer (2008: 175) proposes four values for voice which contrast in the presence or absence
of two feautres: agent and D. The ‘agent’ feature imposes that interpretation to an external
argument, while the D features imposes the syntactic requisite that the specifier of Voice should
be occupied by an overt DP argument. (143) presents the four values.

51
 ANTONIO FÁBREGAS

(143) a. Thematic active voice

VoiceP

DP Voice

Voice …
[agent, D]

b. Thematic passive voice

VoiceP

Voice …
[agent, ø]

c. Non-thematic active voice

VoiceP

Expletive Voice

Voice …
[D]

d. Non-thematic passive voice

VoiceP

Voice …
[ø]

The thematic (143a,b) vs. non-thematic voices (143c,d) differ in whether the specifier of
Voice receives a theta role (agent) or not. In the cases in which the voice is non-thematic, if
there is an argument in the specifier of Voice, it is expletive. Schäffer (2008) proposes that
German marked anticausatives with the reflexive zich ‘self’ are instances of the situation in
(143c). Based on the fact that anticausatives with zich do not license dative causers, Schäfer
proposes that they have the syntax of an active voice, only that no agent theta role: the reflexive,
being inherently dependent on an antecedent, does not require a theta role and is therefore
interpreted as an expletive that simply fills the position of the Voice required by the D feature.
As both Schäfer (2008) and Vivanco (2015) show, the situation in Spanish is radically
different. As we have seen above, SE-marked anticausative verbs do license the dative causer
structure. This means, within the theory of different flavours for voice that Schäfer proposes,
that anticausative SE structures in Spanish must be an instance of the (143d) Voice head, the
head that is non transitive and does not impose an agent reading to any implicit or explicit
argument. This would mean, therefore, that Spanish SE marked anticausatives lack any agent
argument –note, however, that Alexiadou, Anagnostopoulou & Schäfer (2015) revise Schäfer’s
(2008) analysis, proposing that Spanish SE is an instance of (143c), where SE is a reflexive
pronoun–.

52
 SE IN SPANISH

This does not mean, however, that anticausatives lack any kind of causing argument. In this
analysis, the anticausative form can involve a causer, which is not defined in Voice but in little
v. This causer argument remains implicit, but can be recovered through prepositional
structures, which in Spanish can be marked for instance with con ‘with’ or por ‘by’.

(144) a. La puerta se abrió con el viento.

the door SE opened with the wind
b. La televisión se rompió por el humedad.
the television SE broke by the humidity
‘The TV broke because of the humidity’

For the purposes of this overview, the nature of SE, the crucial fact is the one noted by
Vivanco (2015: 265-275), specifically that anticausative structures that do not involve SE do
not license dative causers, and unaccusative verbs do not license them either. This leads, as we
will see, to the generalisation that not all anticausative structures are of the same type, with SE
marked structures having a special status. The examples in (145) and (146) allow an
interpretation of the dative as an affected argument, but not as an involuntary causer. Note that
adding SE can license that reading.

(145) a. El trabajo cambió.

the job changed
b. #A María le cambió el trabajo.
to María her changed the job
‘The job changed, and that affected María’ (not ‘María inadvertedly made the job
change’)
c. A María se le cambió el canal.
to María SE her changed the channel
‘María inadvertedly made the TV-channel change’
(146) a. Desapareció un libro.
disappeared a book
b. #A Juan le desapareció un libro.
to Juan him disappeared a book
‘A book disappeared, and that affected Juan’ (not ‘Juan unwillingly made a book
disappear’)

Crucially, from the perspective of the typology of Voice in Schäfer (2008), this should mean
in principle that SE marked structures correspond to the structure of (147a), while
anticausatives that lack SE should correspond rather to the structure in (147b) or (147c), where
the last one does not even project Voice. Schäfer (2008: 176) suggests that unmarked
anticausatives rather correspond to (147c), but note that a covertly transitive structure like
(147b) where the internal argument and the external argument correspond to the same
constituent –as Ramchand (2008) proposes for unaccusatives– would equally explain that
unmarked anticausatives reject the dative causer.

(147) a. VoiceP

Voice vP
[ø]
v …
[cause]

53
 ANTONIO FÁBREGAS

b. VoiceP

DP Voice

Voice vP
[D]
v DP

c. vP

v DP

It is important to say that, in any case, Schäfer (2008) does not propose that SE is any
element that removes the agent argument. In his analysis, the SE element would be a
manifestation of the highly defective Voice head in (143d), that lacks both a theta role and a
syntactic feature forcing presence of a specifier, and which appears instead of a thematically
agentive voice head like (143a). SE is associated with a defective head –also if one follows
Alexiadou, Anagnostopoulou & Schäfer (2015), where SE is the expletive specifier of a
transitive Voice head that does not assign an agent theta role to it–, but it would not be an
element that removes a theta role; rather it would be the element that emerges when there is no
theta role to assign, and perhaps a nominal element must be merged anyways in the structure.
Then, to conclude this part of the section, and before we move to the relation between
anticausatives and reflexive SE structures, this seems to be the general situation:

i) There is agreement that anticausative structures lack an agent in the strict sense
that there is no clear argument that receives that theta-role; the controversy is with
respect to whether there is another external implicit argument with the ability to
trigger an event, a causer, that can be syntactically expressed or whether that
causer is simply introduced as an adjunct without the verb keeping that theta role.
ii) It seems that SE marked anticausatives are substantially different from unmarked
anticausatives or unaccusatives. Two options present themselves: (i) SE is the
manifestation of a highly defective Voice head that does not project any external
argument and does not impose an agentitivity requisite and (ii) SE is an expletive
argument that does not receive an agentive theta role. The combination with the
dative causer in SE marked anticausative structures suggests that option (i) is
initially in the right track; option (ii) would require some independent explanation
of why German zich structures reject dative causers and Spanish SE structures
accept them.

4.3. Syntactic conditions: the relation between anticausative SE and reflexive SE

We have already seen that SE anticausatives in Spanish provide evidence that the agent theta
role is not being assigned. This fact can be in principle interpreted in two ways that are
compatible with the morphological markedness relation that we mentioned in §4.1 above.
Schäfer (2008) and Vivanco (2015) –see also Schäfer & Vivanco (2016)– propose that SE
alternates –that is, substitutes– another head that otherwise would be spelled out as zero,
creating this the impression that more morphology has been added when in actuality a zero
morpheme has been replaced with SE. The second option is to in fact try to derive the
anticausative version from the causative version by building additional structure on top of it.
From that perspective, SE appears to signal the presence of additional syntactic structure that

54
 SE IN SPANISH

somehow has managed to inactivate or remove the agent argument that we know is not
accessible in the case of anticausative SE structures.
The existing proposals that treat the anticausative element of the pair as derived from the
causative have in common that they assimilate anticausative SE –or equivalent lexical elements
in each language– to reflexive SE (Chierchia 2004, Koontz-Garboden 2009, Beavers &
Koontz-Garboden 2013). The main fact that for this authors makes this relation plausible is the
cross-linguistic generalisation that, in languages that overtly mark the anticausative predicate
–as it is the case in Spanish–, it is common that the same element used to mark anticausatives
is also used to mark the reflexive. Haspelmath (1990: 36) reports that this coincidence is found
in typologically unrelated languages like Latin, Tigre, O'odham, Kanuri, Nimboran or Udmurt.
Spanish of course also shows this type of coincidence, and moreover a more or less standard
grammaticalisation path has been proposed between Latin and Romance (148, from Michaelis
1998) which immediately supports the view that anticausative SE uses are derived from
reflexive SE uses, in a way that anticausative SE is nothing but a more grammaticalised (=more
abstract, involving the bleaching of agent entailments) use of reflexive SE. In Indo-European,
for instance (Rodríguez Molina 2010), the notion of reflexive was subsumed under the general
'middle' voice, for instance.
Of course, the assumption that one has to make with respect to (148) is that somehow the
ordering of grammaticalisation in syntax follows some kind of 'natural' path that is punctuated
by ontological properties of the constructions involved, with the initial steps corresponding to
more basic elements and configurations that are adapted to the other structures, which are –
then– ordered according to how much they diverge from the starting point.

(148) reflexive > anticausative > passive > impersonal

The idea that anticausatives are reflexives or derived from them was suggested among others
by Lakoff (1971), Haspelmath (1990), Reinhart (1996) and Wunderlich (1997), but Chierchia
(2004) and, particularly, Koontz-Garboden (2009) are credited with its techinical
implementation. Following Chierchia (2004: 29), Koontz-Garboden (2009) takes
reflexivisation as an operation whereby a predicate takes a relation R as its argument, and which
sets both arguments of the relation to be the same.

(149) Denotation of the reflexivisation operator (Koontz-Garboden 2009: 83)

lRlx[R(x,x)]

In this context, R is an operator that ranges over arguments of a transitive verb. Importantly,
this would correspond to the SE part of the reflexive –thus, restraining the type of arguments
to transitive predicates, perhaps including in a broader definition indirect objects, correctly
gives an account of the distribution of SE discussed in §3.1–, not to the double, which as we
saw sometimes introduces additional entailments, such as agentivity.
The core proposal is that, if (150a) is the denotation of a causative verb like abrir 'open',
(150b) is the denotation of the anticausative version, abrirse. Note that crucially the second
case has identical arguments for theme and causer ('effector', in the terminology assumed by
Koontz-Garboden 2009).

(150) a. λxλyλsλe[∃v[CAUSE(v, e) ∧ EFFECTOR(v, y) ∧ BECOME(e, s) ∧ THEME(s, x) ∧

open(s)]]
b. λxλsλe[∃v[CAUSE(v, e) ∧ EFFECTOR(v, x) ∧ BECOME(e, s)∧ THEME(s, x) ∧
open(s)]]

55
 ANTONIO FÁBREGAS

In this way, Koontz-Garboden (2009) explains an anticausative as a predicate that does not
lack a causer, but as a predicate where the reflexive operator imposes the interpretation that the
theme and the causer correspond to the same entity. In principle the denotation of the
anticausative is a proper subset of the denotation of the causative version, and not vice versa:
the instances where the external argument coincides referentially with the theme.
Note that in (150) above there is no agent theta role. This proposal is combined with a theory
about theta-roles: verbs that allow an anticausative pair should be causative verbs which specify
the argument of cause as an effector, not as an agent; verbs that specify the external argument
as an agent do not allow the anticausative structure triggered by the reflexivisation operator,
because the presence of SE identifies the external and the internal argument and the entailment
of an agent are not satisfied in the reading that the internal argument undergoes the process by
its own internal properties; the reading that emerges is purely reflexive.

(151) a. Juan asesinó al embajador.

Juan murdered DOM-the ambassador
b. λxλyλsλe[∃v[CAUSE(v, e) ∧ AGENT(v, y) ∧ BECOME(e, s) ∧ THEME(s, x)]]
(152) a. #El embajador se asesinó (en el accidente).
the ambassador SE murdered in the accident
'The ambassador murdered himself', not 'The ambassador underwent murder'
b. λxλyλsλe[∃v[CAUSE(v, e) ∧ AGENT(v, x) ∧ BECOME(e, s) ∧ THEME(s, x)]]

Note that, like always, verbs that are semantically close might differ with respect to the theta
roles assigned. (151)-(152) contrast with (153)-(154) below, where the verb does allow an
anticausative version. For the theory we are reviewing now, this means that the subject cannot
be an agent, but an effector, a property that might be confirmed by (155) below, where we see
that the first verb rejects non-animate subjects that cannot act volitionally:

(153) a. Juan mató al embajador.

Juan killed DOM-the ambassador
b. λxλyλsλe[∃v[CAUSE(v, e) ∧ EFFECTOR(v, y) ∧ BECOME(e, s) ∧ THEME(s, x)]]
(154) a. El embajador se mató (en el accidente).
the ambassador SE killed in the accident
'The ambassador died in the accident'
b. λxλyλsλe[∃v[CAUSE(v, e) ∧ EFFECTOR(v, x) ∧ BECOME(e, s) ∧ THEME(s, x)]]
(155) El cáncer {mató / #asesinó} al embajador.
the cancer killed / murdered DOM-the ambassador
'Cancer killed the ambassador'

This is compatible with an interpretation where the event is internally caused by the
properties of the entity that undergoes the process in the sense that there cannot be an external
agent that triggers the event externally. The propopsal that effectors and agents are different,
combined with the reflexive interpretation, jointly produce the 'middle' interpretation, not
agentive and not passive, of anticausatives.
Like this, against analyses that directly treat anticausatives as derived from causatives, the
presence of SE marks an additional operation that applies a reflexive operator over the
causative verb. The problem mentioned in §4 above, where the morphologically marked
element is more simple than the morphologically unmarked element of the pair, dissolves. The
same operator is interpreted as reflexive when the external argument is an agent, and as
anticausative when the external argument is an effector.

56
 SE IN SPANISH

The evidence in favour of the relation in (150) divides into two groups: (i) tests that show
that the anticausative contains the causative and (ii) tests that show that the additional property
of the anticausative pair is indeed a reflexive one. Let us review the evidence provided in
Koontz-Garboden (2009) for each of these two aspects; as we will see, this evidence is quite
controversial.
The main test is the one in (152): the anticausative can be denied without denying the
causative member. In an analysis where the anticausative is more simple than the causative this
should be impossible, because the causative contains the anticausative and one cannot deny the
change of state without denying also that some agent has caused the change of state. However,
if (150) is correct, (152) is possible because one denies the reflexive part of the anticausative
without necessarily denying that someone caused a change of state –or in other words, because
the anticausative contains the causative and not vice versa–.

(152) La puerta no se ha abierto, la has abierto tú.

the door not SE has opened, it have.2sg open you
'The door didn't open, you opened it'

Some speakers, however, find this not natural and want to add sola 'alone' in the first
sentences, so that the negation can be interpreted as constituent negation applying to sola. At
the same time, this test is a bit puzzling: as can be seen in the English translation, the sentence
is also fine in English, a language where the anticausative should not contain the causative
because there is no reflexive SE morpheme involved and therefore the anticausative should not
be a reflexivised version of the causative.
For these reasons, some researchers have noted that (152) is not your standard clausal
negation, but in actuality an instance of metalinguistic negation, that is, the negation of
something that someone stated –'that the door opened alone'–, not the negation of a particular
state of affairs (see Horvath & Siloni 2011, 2013; Schäfer & Vivanco 2016). Conscious of this
problem, Koontz-Garboden (2009) noted that the context in (152) licenses Spanish ningún
'none', which he treats as a negative polarity item similar to English any, which cannot be
licensed in metalinguistic negation contexts.

(153) a. John didn't manage to solve some of the problems, he solved them all.
b. *John didn't manage to solve any of the problems, he solved them all.

(152) accepts ningún, as witnessed by (154), which for Koontz-Garboden (2009) means that
(152) cannot involve metalinguistic negation.

(154) No se rompió ningún vaso, los rompiste todos tú.

not SE broke none glass, them broke all you
'No glass got broken, you broke them all'

However, Schäfer & Vivanco (2016) note that Spanish ningún is not an NPI, as it can be
licensed in clear cases of metalinguistic negation.

(155) Juan no odia a ningún niño, los aborrece a todos.

Juan not hates DOM no child, them loathes DOM all
'Juan doesn't hate chidren, he loathes them'

57
 ANTONIO FÁBREGAS

A real Spanish NPI that cannot be used also as a negative inductor is siquiera 'even'. This
one is, as expected if (152) is a case of metalinguistic negation, is rejected in (152) as well as
in other clear instances of metalinguistic negation.

(156) a. *Juan no odia a los niños siquiera, los aborrece.

Juan not hates DOM the children even, them loathes
Intended: 'Juan does not just hate children, he loathes them'
b. *La puerta no se abrió siquiera, la abriste tú.
the door not SE opened even, it opened you
Intended: 'The door didn't just open, you opened it'

An element that forces metalinguistic negation is solo 'just', which is compatible with (152),
again providing evidence that it is a case of metalinguistic negation and therefore does not
constitute a test in favour of the derivation of anticausatives from causatives.

(157) La puerta no solo se abrió, la abriste tú.

the door not just SE opened, it opened you
'The door didn't just open, you opened it'

A second test that Koontz-Garboden (2009) proposes is the fact that what could be taken to
be an anticausative SE can control a final clause without the passive interpretation. (158a) is
the causative version of the bona fide anticausative in (158b); note that a final clause can be
licensed by (158b), as seen in (158c).

(158) a. Juan ahogó a su hijo.

Juan drowned DOM his son
b. Juan se ahogó.
Juan SE drowned
c. Juan se ahogó para salvar a su hijo.
Juan SE drowned to save DOM his son

However, this effect seems to be lexically restricted and does not generalise to all SE marked
anticausatives, which is what the reflexivisation analysis predicts: in those cases and only in
those cases is there an effector argument that can control the final clause. Remember, for
instance, (159), where the presence of the final clause forces a passive reading.

(159) #El barco se hundió para cobrar el seguro.

the boat SE sank to cash the insurance

It is also difficult to ascertain that, beyond the apparent contrast between (158a) and (158b),
the structure in (158c) is indeed anticausative. Note that Koontz-Garboden (2009) takes
Spanish por sí solo 'by itself' as a sign of anticausativeness, an element to which we will go
back almost immediately. However, (160a) does not sound natural, even though (160b) is in
principle possible.

(160) a. ??Juan se ahogó por sí solo para salvar a su hijo.

Juan se drowned by himself to save DOM his son
b. Juan se ahogó {solo / por sí solo / sin ayuda}.
Juan drowned alone / by himself / without help

58
 SE IN SPANISH

One possibility is that the type of SE in (158b) and (158c) is not the same one, and crucially
that while (158b) might be anticausative, (158c) is not –perhaps it could be analysed as an
aspectual SE or, lacking that class, agentive SE, meaning that the person did it on purpose–.
The second set of tests refer to the reflexive nature of these predicates. Koontz-Garboden
(2009), following Chierchia (2004) (see also Beavers & Koontz-Garboden 2013), proposes that
the elements in (161), which typically differentiates passive SE from anticausative SE (162),
is itself reflexive because the anticausative, but not the passive, is obtained with a reflexive
operator.

(160) a. por sí solo

by self alone
b. por sí mismo
by self same
(161) a. La puerta se abrió {por sí sola / por sí misma / sola}.
the door SE opened by itself / by itself / alone
b. *La puerta fue abierta {por sí sola / por sí misma / sola}.
the door was opened by itself / by itself / alone

Crucially, Chierchia (2004) argues that the Italian equivalent of (160) can only be bound by
agents (or effectors, we may add); (161a) is fine because the subject is indeed an external
argument (effector), and (161b) is out because the subject is not an external argument.
There are two problems here. Importantly, the prediction that the reflexivisation analysis
does is that the distribution of (160) in anticausative cases must match with SE marking: only
SE marked anticausatives should license it, and not the ones that lack marking. This prediction
is not borne out, as witnessed by (162):

(162) El tiempo cambió por sí solo.

the weather changed by it self

See Horvath & Siloni (2011, 2013), where they argue that some of the cases that were
supposed to show that non-SE marked anticausatives are incompatible with 'by itself' in fact
refer to processes that cannot be triggered easily by external forces, and therefore where 'by
itself' is redundant, as in (163; see also Mendikoetxea 1999).

(163) *El niño creció {por sí mismo / por sí solo / #solo}

the boy grew.up by himself / by himself / alone
Intended: 'The boy grew without the help of anyone'

This basically casts doubt on the first side of the test, namely that the presence of this
element suggests a reflexive construction: if SE marks reflexivity, (162) should not allow this
marker. Secondly, there is a quite extensive body of work, among others Reinhart (2000),
Schäfer (2007) and Alexiadou et al. (2015), refuting the claim that 'by itself' requires
necessarily agentive subjects; this denies the second part of the test, namely that the SE
anticausative verbs involve projection of an agent external argument. For instance, a non
agentive holder of a state like in (164) can be the antecedent of 'by himself'.

(164) Juan sabe la respuesta {por sí mismo / por sí solo / solo}.

Juan knows the answer by himself / by himself / alone
'Juan knows the answer without the help of anyone else'

59
 ANTONIO FÁBREGAS

Thus, under the light of these tests, it seems that there is still not enough evidence in favour
of an analysis of anticausative SE structures that assimilates them to reflexive structures.
Without any doubt, the fact that cross-linguistically reflexive markers become often markers
of anticausativity should be telling us that there is a non-trivial relation between both structures,
but stated in the way that Chierchia (2004) and Koontz-Garboden (2009), it does not seem that
the right predictions are being made for the entailment relations between the causative and the
anticausative or for the expected distribution of the 'by itself' element in the relevant meaning
of 'no external causer can be identified'.
The general consensus is then that anticausatives are probably derived from causative
structures by operating in its Voice head. SE could then be treated as an expletive pronoun that
fills the specifier position of Voice without receiving an agentive theta role, as proposed by
some authors (165), or the manifestation of a Voice head without any features (166), as
proposed by others. The analysis, then, might connect with reflexives in the sense that SE
would be a defective element that simply satisfies the position of the agent without further
semantic effects or, in the sense of Labelle (2008), if SE marks a type of voice where at least
one argument position is left unexpressed.

(165) VoiceP

se Voice

Voice vP

v ...

(166) VoiceP

Voice vP
se
v ...

With these options in mind, we are almost ready to now move to the third type of lexical
SE, aspectual SE. However, before we do that we have to briefly discuss a generally
understudied class of anticausative predicates: transitive anticausatives, whose analysis
connects with one of the main problems of SE in traditional grammars, the dative reflexive
analysis.

4.4. Transitive anticausatives and the problem of reflexive datives expressing affectedness
The analysis of sentences such as the ones in (167) is not straightforward.

(167) a. Juan se cortó el dedo.

Juan SE cut the finger
'Juan cut his finger'
b. Juan se rompió el brazo.
Juan SE broke the arm
'Juan broke his arm'

These sentences have been considered instances of transitive verbs that are interpreted as
anticausatives. Like anticausatives in combination with dative causers, the predicates have the
flavour of events that happen in a non volitional way and that affect the subject. The standard

60
 SE IN SPANISH

interpretation of (167) is that Juan accidentally got a cut in the finger and that his arm broke,
without him setting in motion that change of state volitionally. Note that these sentences
typically have three ingredients:

a) A direct object that corresponds to a body part, the most prototypical class of inalienably
possessed entities
b) A subject that invariably denotes the human entity whose body part is being discussed in
the predicate
c) A change of state verb that expresses a change that generally does not interrupt the
relation between the possessor and the possessee: (167a), for instance, is not interpreted as if
the finger gets severed from the hand.

However, the accidental or non willful interpretation of (167) is not forced. Unlike the
anticausative predicates that we have examined in this section, (167) cannot easily be combined
with the usual markers (168).

(168) Juan se cortó el dedo {#solo / ??por sí solo / ??por sí mismo}.

Juan SE cut the finger alone / by himself / by himself

Even when the predicates are combined with the usual markers of anticausative, the
interpretation that emerges is rather 'without anybody else helping him', which is not
incompatible with Juan doing the task on purpose. In fact, (167) can combine with a final clause
and there does not seem to be additional grammatical markers that distinguish the non-willful
interpretation from the willful one.

(169) Juan se rompió el brazo para cobrar el seguro.

Juan SE broke his arm to cash.in the insurance
'Juan broke his arm in order to cash the insurance in'

There is one property, though, that differentiates (167) from usual transitive predicates: the
impossibility of undergoing a passive, as illustrated in (171), in contrast with other uses of the
same verb.

(170) a. Juan cortó el pan.

Juan cut the bread
b. Juan se cortó el dedo.
Juan SE cut the finger
(171) a. El pan fue cortado.
the bread was cut
b. #El dedo fue cortado por Juan.
the finger was cut by Juan

(171b) would imply something like a complete separation between Juan and his finger, that
is, the finger would be severed, and the notion of inalienable possession is lost.
Of course, the potential existence of anticausatives with direct objects would directly deny
the unaccusative analysis of the anticausative verbs in Spanish –along the same lines as we saw
at the beginning of the section–, but would also complicate in a radical way the nature of
anticausative predicates, at least as far as the structure in (167) seems to involve a transitive
predicate with an external argument that is distinct from the internal argument.
There are three main ways out from this situation.

61
 ANTONIO FÁBREGAS

a) The first one is to treat (167) as some kind of anticausative structure, similar to the dative
causer structure that we have analysed, but where the dative is expressed through a reflexive
SE pronoun that allows the dative to be expressed as a nominative argument. The parallelism
would be of the kind expressed in (172), where perhaps the inalienable possession of the object
is what determines that the relation between the possessor and the possessee can involve a
nominative argument.

(173) a. A Juan se le rompió la televisión.

to Juan SE him broke the TV
b. A Juan se le rompió la mano.
to Juan SE him broke the hand
c. Juan se rompió la mano.
Juan SE broke the hand

Note that in this analysis SE would be an anticausative marker in (173a, b) and a dative
reflexive in (173c). In this analysis, (173c) would be built from a predicate like (174), only that
adding a reflexive dative that blocks the expression of the anticausative SE.

(174) La mano se rompió.

the hand SE broke

b) As in the case of factitive or causative SE (§6.1), the SE that we see in (167) is not related
to an anticausative or to a dative reflexive, but to a marker that licenses the interpretation of
the subject as not an agent, but some kind of less directly involved participant. This would
explain that the same predicate can be interpreted, depending on context, as expressing a
volitional event or an accidental one, in the same way that the predicates that allow the factitive
reading typically also allow an agentive one and context determines between the two. However,
this theory cannot explain why the direct objects are restricted to body parts or other closely
inalienably possessed entities, and why the subject must be the possessor of the object. In other
words, this theory does not relate the SE form that appears in (167) with a reflexive dative
interpreted as a possessor.

c) Like in (a), the example in (167) involves a dative reflexive, but it is not derived from the
anticausative, but from a causative. What the SE in (167) does is to introduce the possessor of
the body part in the same way that the dative possessor (Gutiérrez Ordóñez 1999) does in (175).
The only difference is that in (167) the dative possessor is coreferential –that is, reflexive– with
the subject.

(175) a. Le rompí la mano.

him.dat broke the hand
'I broke his hand'
b. Se rompió la mano.
SE broke the hand
'He broke his hand'

Of course, what this theory does not explain is why an involuntary interpretation emerges
in (167). Note, in this respect, that (175a) does not have to imply that the breaking of the hand
was intended, in the same way that (167) does not force the unintentional interpretation. If the
two readings are in principle available, that the unintentional interpretation is prefered in (167)

62
 SE IN SPANISH

might simply follow from a pragmatic component that makes us expect that someone would
not harm himself wilfully, although perhaps this is not so clear.
The case of transitive anticausative predicates in traditional Spanish grammar is generally
treated as (c), that is, as involving a dative argument that is reflexive to the subject and
expresses in this particular case, the possession of an inalienably possessed element. However,
perhaps the clearest manifestation of this type of dative reflexive pronoun is the one that other
authors have classified as aspectual: the case of an argument that seems to involve the
affectedness of the subject in the event expressed by the predicate. Let us then move to the next
section, where we will discuss these cases.

5. Predicate-affecting SE uses (3): aspectual

Perhaps the most controversial of all the SE functions that have been recognised in
grammatical analyses internal to Spanish is the notion of aspectual SE, which is first noticed
in Bello (1847: §758). This term is used to define a class of predicates that can be facultatively
marked with SE, in the strict sense that the same verb might appear without SE without
affecting in principle the number of arguments that are syntactically active or their case
licensing conditions. As the name suggests, the presence of SE in these cases is related to an
aspectual difference, where 'aspect' should be interpreted as 'lexical aspect'.

(177) a. Juan leyó dos capítulos.

Juan read two chapters
b. Juan se leyó dos capítulos.
Juan SE read two chapters

This aspectual label, however, substitutes other, more traditional, terms to refer to this
optional SE, such as 'quantificational SE' (Fernández Ramírez 1986), 'stylistic SE' (Klein
1987), 'benefactive SE' (Rigau 1994), or analyses that relate this SE with ethical datives
(Sánchez López 2002). This benefactive use of SE identifies the SE form that appears in (177)
with a dative participant that is not compulsory in the sentence but is introduced in order to
achieve some semantic effect. Thus, the dative reflexive analysis is characterised by the
following claims:

a) The SE that appears in examples like (177) is optional

b) Its presence involves a notion of involvement, where the subject that is coindexed with
the SE form is more involved in the event
c) This involvement is generally interpreted through a notion of 'intensity', where the subject
performs the event in a manner that requires more energy or effort, or alternatively where the
effect of the event, after its completion, is pleasant or affects the subject positively or negatively
in other ways.

On the other hand, the analyses that treat the SE in (177) as aspectual, propose the following
properties:

a) The SE that appears in (177) is not strictly optional, but required to express a particular
interpretation of the event where it must forcefully culminate as a telic event
b) The presence of SE has grammatical effects in the nature of the direct object
c) Possible 'involvement', 'affectedness' or 'intensity' readings must be the epiphenomenon
of the aspectual and grammatical changes just described.

63
 ANTONIO FÁBREGAS

The traditional discussion of this facultative (or apparently facultative SE) in Spanish
grammar has oscilated at least since Fernández Ramírez between these two poles: from a view
that relates the presence of SE with some notion of expressiveness, special agent implication
in the event or some affectedness on the subject, and a view that denies or reinterprets the
previous effects as an epiphenomenon of a deeper property, related to aspect. The structure of
this section reflects this discussion.
The two main syntactic classes of predicates where this aspectual SE have been proposed
are asymmetrically treated. Transitive verbs like (177), generally involving some form of
consumption, and where the direct object acts as an incremental argument that measures the
development of the event across time are the class where the notions of dative reflexive and /
or aspectual SE have been most used. The second class where some authors have used the label
'aspectual SE' is one where nobody has attempted to treat the SE form as a dative reflexive: a
poorly-defined set of intransitive verbs, including many cases of bounded achievements
involving changes of state or directional movements. The class of transitive verbs is much more
extensive than the class in (ii), which is perhaps restricted to less than 20 cases.

(178) a. Juan cayó dentro del pozo.

Juan fell inside of-the well
b. Juan se cayó dentro del pozo.
Juan SE fell inside of-the well

As we will see, this category is controversial for three main reasons. The first one is that
there is no agreement with respect to which types of SE constructions should be considered
aspectual; as we will see several authors have proposed that some of the prototypical examples
of aspectual SE uses should be classified as other types of structures, such as agentive SE,
anticausative SE or figure SE. Secondly, even assuming that there is some proper notion of
aspectual SE, the type of aspectual change that this SE induces into the predicate is not
described in the same way by all the authors. Third, as so-called aspectual SE comes
accompanied in many cases of other relevant changes in the predicate, some have even
proposed that a class of aspectual SE structures should not be considered at all.
In this section, we will first present the properties of aspectual SE assuming that it exists
(§5.1). Then we will revise some of the analyses that have been proposed for the location of
SE on the assumption that aspectual SE is a real class (§5.3). Then we will move to the
proposals that have reanalysed the prototypical aspectual SE cases as instances of other
constructions (§5.3, §5.4, §5.5, §5.6), including those that have argued that this SE should be
seen in essence as an anticausative SE because anticausative SE is distributed according to
some aspectual restrictions.

5.1. The description of aspectual SE: predicates and objects

The so-called aspectual SE has been proposed mainly for a well-defined set of transitive
predicates (Nishida 1994, Sanz 1995, Zagona 1996, De Miguel & Fernández Lagunilla 2000,
Kempchinsky 2004, Basilico 2010, among others), with a much more controversial extension
to a less clearly defined class of intransitive predicates –something that does not mean that the
label of aspectual SE is not controversial with transitive predicates–.
The classes of transitive predicates where this SE has been proposed include the four
following classes (Di Tullio 2012); following Nishida (1994), the idea is that what these classes
have in common is that their direct object is an incremental object. The object in this verbs is
isomorphic with the development of the event, measuring it: the progression of the event of
eating matches, for instance, different mereological parts of the apple, such as that as the event

64
 SE IN SPANISH

progresses more parts of the apple are consumed, and the event culminates when there are no
more (edible) parts of the apple left.

a) Verbs of intellectual or physical consumption

leer 'read', comer 'eat', beber 'drink', fumar 'smoke', estudiar 'study', aprender 'learn',
memorizar 'memorise', tomar 'have, take'...
b) Verbs of creation
escribir 'write', pintar 'paint', construir 'build', dibujar 'draw', preparar 'prepare', componer
'compose', filmar 'film', esculpir 'esculpt'...
c) Verbs of acting which are also measured by the direct object
bailar 'dance', cantar 'sing', representar 'act', interpretar 'play', tocar 'play', copiar 'copy',
fotografiar 'photograph'...
d) Verbs of incremental change
lavar 'wash', planchar 'iron', teñir 'dye', sembrar 'plant', regar 'water', serruchar 'saw'...

With these verbs, the effect of SE is described in different ways, but with the common
intuition that what SE does is to force a telic reading where the event culminates. Nishida
(1994) proposes that SE involves here that the event is completely measured by the incremental
theme, up to its total completion –thus, SE imposes a telic reading (see Sanz 1995), or in other
words implies that the subject and the object coincide at the end of the event (Zagona 1996)–.
This involves that in many cases the presence of SE with these verb is paraphrased in English
with the same predicate in combination with a particle involving culmination, as up in (179b,
which is telic.

(179) a. Luis comió un bocadillo (durante un rato).

Luis ate a sandwich (for a while)
'Luis ate a sandwich for a while'
b. Luis se comió un bocadillo en diez minutos.
Luis SE ate a sandwich in ten minutes
'Luis ate a sandwich up in ten minutes'

Beyond telicity, and in addition to it, it has been also noted that in (179a) one could be
describing an unordered processes of eating where one nibbles here and there, while in (179b)
one expects an ordered and systematic consumption.
Another corollary of this approach is that the presence of SE with these verbs is sensitive to
the bounded or unbounded nature of the object. In principle, aspectual SE rejects incremental
objects that are unbounded (180b) or transitive predicates that take a bona fide implicit objects
(180c), because the nature of these objects is incompatible with the interpretation that the event
culminates –as an unbounded object does not define a boundary that can be used as a reference
for the endpoint of the event–. This explains contrasts as in (180).

(180) a. Pedro (se) cantó una canción.

Pedro (SE) sang a song
'Pedro sang a song (until the end)'
b. Pedro (*se) cantó tangos.
Pedro SE sang tangos
'Pedro sang tangos'
c. Pedro (*se) cantó.
Pedro SE sang
Intended: 'Pedro sang'

65
 ANTONIO FÁBREGAS

The correlation between aspectual SE and incrementality predicts that some verbs will only
accept it when they are combined explicitly with a (bounded) incremental entity, no matter its
nature. For this reason, this aspectual SE can in principle be extended to movement predicates
where the incremental entity is a locative path (García Fernández 2011); some verbs seem to
satisfy this requisite by its lexical meaning (181a,b), while others require an overt incremental
constituent; note that the presence of an endpoint is not enough for SE (181c,d), supporting
Nishida's (1994) main claim about these verbs.

(181) a. Juan llevaba un libro.

Juan carried a book
'Juan walked around with a book'
b. Juan se llevó un libro.
Juan SE carried a book
'Juan took a book (from somewhere) with him'
c. Juan se nadó cien metros.
Juan SE swam 100 meters
d. Juan (*se) nadó hasta la orilla.
Juan SE swam until the shore

So far so good; note crucially that the verb classes abopve share the property of belonging
to the same lexical aspectual class: accomplishments that regularly can alternate with an
activity reading. With these facts in mind, it is in principle reasonable to think that the SE used
in these contexts is aspectual, because the class of predicates that allow it share in fact crucial
aspectual properties, the effect that SE has in them has predictable and systematic effects on
aspect and the predictions about the incremental nature of the predicate, together with
complements and modifiers, and the bounded nature of that incremental component.
The approach, clearly, puts the aspectual contribution of SE as primary and the possible
agent effects (involvement of the agent, affectedness, etc.) that traditional grammar recognises
for these cases are, at best, derived from the aspectual properties. However, this happens in
ways that are perhaps not very systematic: Sanz & Laka (2002) for instance suggest that the
agent involvement is interpreted because typically accomplishment verbs are agentive; other
authors more or less informally suggest that when the event is telic, and has culminated, it is
easier to infer that it has affected the subject in some significant way. But in any case: this
approach treats the role of SE as inherently aspectual.
However, the aspectual SE label has also been extended, by Nishida (1994) and more in
particular in De Miguel & Fernández Lagunilla (2000), to a set of intransitive verbs, many of
them denoting directional movement, including irse 'leave', salirse 'to get in', morirse 'to die',
dormirse 'to fall asleep', caerse 'to fall down', sentarse 'to sit down' (see also Jiménez-
Fernández & Tubino 2014, 2019).
The list is not much longer than this. Note two facts, extensively discussed by García
Fernández (2011, 2015) and others: unlike the other classes of predicates, this class is not
aspectually homogeneous without SE: dormir 'to sleep' is atelic, while salir 'to exit' is telic.
However, De Miguel & Fernández Lagunilla (2000) propose that the presence of SE with these
verbs forces an aspectual reading that consists necessarily of an achievement followed by a
result state. Although they never provide a compositional analysis of this contrast, according
to their analysis, in (182a) there should not be a result state –and in fact the event can be
conceptualised as atelic– while in (182b) the event must be telic and should involve a result
state.

66
 SE IN SPANISH

(182) a. Juan cayó (por la grieta durante algunos segundos).

Juan fell by the crevice for some seconds
b. Juan se cayó al suelo.
Juan SE fell to-the ground
'Juan fell down to the ground'

The problem is that not all verbs in the list above, or in general all intransitive verbs with a
putative aspectual SE, can be argued to move from atelic to telic. Note, for instance, the contrast
in (183a), which does not depend on telicity, but rather on whether one can imagine that
lightning comes from a previous location or not. (183a) suggests that what SE does is not to
create telicity, as the SE-less version can also be telic, but rather to define some movement that
must come from an established previous source location, as García Fernández (2015) proposes.
This is not the only example. The verb salir 'to exit' has in principle without SE already a result
state, that can be measured with a for-phrase (183b).

(183) a. (*Se) cayó un rayo.

SE fell a lightning
'Lightning fell'
b. Juan salió de la reunión durante unos minutos.
Juan exited from the meeting for some minutes
'Juan left the meeting and stayed outside for some minutes'

Thus, this analysis faces one clear complication: is aspectual SE with these verbs an element
that creates an achievement + result state interpretation in verbs that do not have it without SE
or is it an element that highlights the result state in verbs that already have it? The short list
above seems to indicate that both options should be true at the same time.
The second problem is that the treatment of SE as an aspectual element in these intransitive
verbs has a problem with the predictions of the theory –unlike the clearer predictions of the
aspectual SE analysis for the incremental predicates–. As García Fernández (2011) notes, the
verb salir 'exit' contrasts with the verb entrar 'enter' in the availability of the combination with
SE. There are many varieties of Spanish that have the contrast in (184).

(184) a. Juan (se) salió un rato de la reunión.

Juan SE exited a while from the meeting
'Juan left the meeting for a while'
b. Juan (*se) entró un rato en la reunión.
Juan SE entered a while in the meeting
'Juan joined the meeting for a while'

Why would these verbs contrast so sharply? De Miguel & Fernández Lagunilla (2000)
would be forced to say that the two verbs, despite their obvious lexical relation with makes one
the reverse of the other, have different aspectual structures, but their behaviour seems to be
parallel also in the crucial respect that both verbs allow the measurement of a result state: (185)
is parallel to (183b).

(185) Juan entró en la reunión durante unos minutos.

Juan entered in the meeting for some minutes
'Juan joined the meeting for some minutes'

67
 ANTONIO FÁBREGAS

If aspectual SE marks or highlights the result state, why would (185) reject SE?
Alternatively, and more crucial: even if entrar was really aspectually different from salir, why
could we not add SE in order to change its aspect, as we do with caer or dormir?
These complications lead García Fernández (2011, 2015) to partially reject the aspectual SE
hypothesis at least for movement verbs and other transitive predicates, as we will see later.
However, other authors have in fact rejected aspectual SE altoghether from different
perspectives (Di Tullio 2012, Armstrong 2013, De Benito 2021), as we will also cover in a few
pages. However, before doing so, let us revise briefly the different analyses that aspectual SE
has received in the literarture.

5.2. The structural location of aspectual SE

To be fair, there are not many formal analyses of aspectual SE that relate their syntactic
properties to other uses of SE. De Miguel & Fernández Lagunilla (2000) simply treat the
element as an aspectual operator –loosely interpreted as a function that alters the aspect of the
base it combines with– and do not discuss its structural position, or the relation between this
use and the reflexive, anticausative or passive uses. In terms of formal analyses of aspectual
SE, in this section we will highlight several recent analyses that propose an explicit position
for aspectual SE and attempt to relate it to other uses of SE, in two of these cases specifically
the reflexive one.
The analyses can be grouped in two classes: (i) a set of analyses, most significantly and
explicitly represented by Kempchinsky (2004), where SE has a main aspectual use which
derives the reflexive or anticausative uses, and (ii) a set of analyses that treat aspectual SE as
reflexive and explain its presence because adding additional subevents to the predicate force
movement of an argument from one subevent structure to the other, hence introducing
reflexivity (see also Cuervo 2014).
Kempchinsky (2004; see also 2000 and 2006) proposes that SE is inherently an aspectual
element, in the sense that it is introduced always in an Asp(ect)P that is internal to the vP,
specifically sandwiched between v and V. The different uses of SE are explained by two
parameters, which partially correlate with the nature of SE as a head or a specifier –on the
assumption that, being a maximally impoverished element that gets no arguments of its own,
it can be ambiguously defined as an X and as an XP–:

a) whether the role of SE is to add a subevent, link to a subevent or absorb a subevent

b) whether SE is introduced as a specifier of AspP, in which case it blocks the relation
between a direct object and the verb, or is introduced in the head of Asp, in which case it does
not have an effect on the argument structure of the predicate.

(186) represents the structure of aspectual SE according to this author.

(186) vP

DP v

v AspP

DP Asp

Asp VP
se
V DP

68
 SE IN SPANISH

The idea is that in this case SE is projected as part of the head of Asp, where it introduces
the subevent that defines the event as telic (if extended to intransitive predicates, it would
introduce a result state). Being in a head position, it leaves the specifier of AspP free for the
internal argument to move, where it measures the event –thus explaining why the aspectual SE
structures with transitive verbs are sensitive to the boundedness of the object–, and the direct
object can from there check case of v. See also, in this sense, De Miguel (1992), who associated
SE to a functional aspect node, and Sanz & Laka (2002), who propose that aspectual SE is
linked to an Event Phrase which they use to define the internal aspect of the predicate.
In contrast to this, for Kempchinsky reflexive SE would be an instance of SE projected as
the specifier of Asp (187) –note that she follows Torrego (1995) in proposing that the doubling
cases of the reflexive, including a sí mismo, have a different structure where SE is introduced
as part of a complex DP structure together with the mismo structure, which we will not discuss
here–.

(187) vP

DP v

v AspP

se Asp

Asp VP

V (*DP)

Once SE is in the spec, AspP position, it links to a transition subevent, and it blocks the
case-checking relation between the internal argument and the head little v, by intervening
between v and DP. This is what forces a reflexive construction not to introduce a direct object
in the derivation: if present, as in (187), the derivation will not be convergent because the
internal argument has no way to receive case.
The anticausative SE is a situation where SE projects as part of little v, absorbing the
initiation subevent –the causative component–, thus delinking also the argument that would be
associated to that subevent, the agent or causer.

(188) vP

v AspP
se
DP Asp

Asp VP

V DP

Thus, SE is inherently aspectual in all cases: the ability to remove arguments is either due
to its absorbing a subevent, and with it the argument related to it, or explained by blocking of
a case-assignment relation; the aspectual effect follows directly from its role as an aspectual
element that can introduce more subevents. However, the aspect that Kempchinsky (2004) does

69
 ANTONIO FÁBREGAS

not address, going back to (186), is the variation as a reflexive element. Remember that
aspectual SE coincides in features with the subject, not the object:

(189) a. Alberto se bebió un café.

Alberto SE drank a coffee
'Alberto drank a coffee up'
b. Yo me bebí un café.
I ME drank a coffee
'I drank a coffee up'

Assuming that in (186) the internal argument displaces to a projection in whose head SE is
introduced, one would expect that SE matches features with the internal argument, not the
external argument. The solution could be not to displace the internal argument to DP and letting
it check boundedness with aspect without movement; this would force treating boundedness as
an agreement feature, a move that is possible but not clearly motivated for a language like
Spanish, where agreement does not seem to be sensitive to boundedness.
For this reason, several approaches have proposed a reflexive analysis of the aspectual SE.
They have in common the idea that the presence of SE does not trigger the aspectual change.
Rather, the aspectual change is triggered by introducing a more complex verbal structure; the
complexity forces that one single DP occupies more than one argument positions, and SE is
introduced to mark that the DP is acting as two arguments at the same time, as a reflexive.
The analyses that behave in this way generally try to incorporate to the explanation also the
affectedness or agent involvement effect that since Bello (1847) has been associated to this
type of facultative SE.
García Pardo (2021) is a very clean and explicit example of the first type of analysis. His
concern is the pseudo-copulative verb in (190), which has a SE-less pair that is aspectually
different.

(190) a. quedarse
remain-SE
b. quedar
remain

While (190b) is stative, (190a) has two interpretations: one that involves a change of state
(191a) and one that is stative (191b) but where it is entailed that the subject is involved as an
agent in maintaining that state, that is that the subject puts effort in staying quiet (see Bull
1952).

(191) a. Juan se quedó dormido.

Juan SE remained asleep
'Juan fell asleep'
b. Juan se queda callado.
Juan SE remained quiet

His proposal in that quedar has a simple stative structure as in (192). No SE is necessary
here.

70
 SE IN SPANISH

(192) ResP

DP Res

Res AP

La silla quedó vacía.

'The chair remained empty'

SE marks, as a head, the presence of a second verbal subevent where the specifier DP is
identical to the DP in the lower aspectual event. The change of state reading is obtained where
a projection corresponding to a dynamic event, Proc, is projected over the stative head.

(193) ProcP

DP Proc

Proc ResP
se
DP Res

Res AP

The agent implication in (191b) is obtained when the second subevent corresponds to a
causative subevent, Init, that defines its subject as an agent.

(194) InitP

DP Init

Init ResP
se
DP Res

Res AP

A similar analysis is done by Jiménez-Fernández & Tubino (2014), who try to relate also
the aspectual effect to the presence of additional structure, as in (195a) –simplified with respect
to their original structure, but where crucially the small clause (SC) introduces a predication
between the subject and a particular location, entailing permanence on the result location
expressed by it–.

71
 ANTONIO FÁBREGAS

(195) a. VoiceP

DP Voice

Voice vP

v VP
CAUSE
se V SC
llegar
DP PP

a casa

b. El hombre se llegó a mi casa.

the man SE arrived to my house
'The man came to my house and stayed there some time'

The analysis is complicated by the geographical variation on how (195) is interpreted, buyt
here we focus on the reading that this predicate has in some varieties of Andalusian Spanish –
other varieties, such as the one in Madrid, take (195) as meaning 'to approach' without
necessarily arriving to the place–. Let us assume the intended reading in the original article we
are reviewing.
As can be seen in (195a), it can be argued that the approach shares spirit with García-Pardo's,
specifically in that SE is related to the presence of an argument in two different positions in the
structure; here, SE is introduced in a verbal head that involves causation, highlighting the
contribution of the aspectual SE to the involvement of the agent: (195b), beyond the aspectual
meaning that forces the presence of a result state where the subject stays for some time, we
have the effect that the subject performed the movement event involved as an agent that
controls it. In fact, (195b) cannot mean that the man arrived to my house unless he had the
explicit intention of doing so, that is, it cannot be that he mistakenly arrived to my house when
he intended to go somewhere else.
These different analyses present different views of the aspectual SE cases, but do we really
have an aspectual SE? The proposal that aspectual SE should be removed from the list of uses
of SE, or at least that the range of cases where that label can be applied with cause should be
severly restricted, has also been made in the literature. The next three subsections discuss
different aspects of this approach.

5.3. Aspectual SE, transitive SE and agentive SE

Transitive cases of aspectual SE have been reinterpreted as cases of other uses of SE in
different works. There are two main types of reinterpretations of the notion of aspectual SE: (i)
an analysis that treats the aspectual cases as instances of reflexives, specifically dative
reflexives and (ii) analyses that reduce aspectual SE in part or wholly to a type of SE that
defines an agent.
The first option is an instance of the analysis of aspectual SE as a reflexive, but not relating
it to an additional aspectual head. This analysis connects with the traditional view of these SE
structures as dative reflexives. MacDonald's (2017) proposal emphasises the subject
affectedness in the event related to aspectual SE uses, and is illustrated in (196).

72
 SE IN SPANISH

(196) VoiceP

DPi Voice

Voice ApplP
se
Appl Appl
proi
Appl VP

DP V

V PP

Pø pro

In a nutshell, aspectual SE transitive predicates involve a more complex argument structure

than their SE-less counterparts. They involve a covert PP structure which introduces an empty
pronoun that is coindexed with the external argument in VoiceP. They additionally involve an
applicative structure that defines an indirect object which corresponds to the affected pronoun,
which takes the external argument –the agent– as its referent. Like this the structure explains
that the aspectual SE structures involve the implication of the agent as an entity that is somehow
affected, positively or negatively, by the event itself. SE is introduced in the voice head, and
as it gets its features assigned by the external argument, it marks the dative pro argument with
which the agent is coindexed: SE spells out voice because the dative and the agent are
coreferential, therefore as an instance on reflexivity.
A strong argument in favour of an analysis like this comes from Rigau (1994), that notes
that the presence of the so-called aspectual SE in fact precludes another beneficiary or dative
affected entity from appearing, as expected if in fact SE corresponds to a dative:

(197) a. Juan le leyó un libro a su madre.

Juan her.dat read a book to his mother
b. Juan se (*le) leyó un libro (*a su madre).
Juan SE her.dat read a book to his mother

Starting from this type of analysis, MacDonald (2017) derives the other effects from the
configuration without positing a specific aspectual head. The boundedness requirement that
aspectual SE imposes in the direct object is not treated in this analysis as a requisite of telicity;
note that in this analysis SE relates to a dative or goal that is affected by the event, and does
not enter into any direct relation with the internal argument, unlike Kempchinsky (2004) and
other analyses. In fact, MacDonald (2017) notes that the aspectual SE can appear with stative
verbs, hence atelic, with the same effect on the boundedness of the internal argument. The set
of verbs involved here is quite restricted, and they share the property of being verbs expressing
intellectual states. Note that other stative verbs not belonging to this class lack this option, no
matter how delimited the object (198e).

(198) a. Juan se cree lo que has dicho.

Juan SE believes that what have said
'Juan believes what you have said'

73
 ANTONIO FÁBREGAS

b. Juan se sabe la lección.

Juan SE knows the lesson
c. Juan se conoce muy bien a María.
Juan se knows very well DOM María
d. *Juan se sabe cosas.
Juan SE knows things
e. *Juan se tiene todos los libros.
Juan SE has all the books
Intended: 'Juan has all the books'.

Thus, what derives the requisite that unbounded bare nouns cannot appear in aspectual SE
cases? MacDonald (2017) proposes that this is because the VP structure required in this
construction is in fact different and more complex than the one in the absence of SE. In (196)
above, the direct object is the subject of the VP structure, in a specifier position. If bare nouns
are generally blocked from specifier positions, the constraint that aspectual SE imposes in these
objects naturally follows. In the SE-less cases, the object is in the complement of V position,
thus allowing bare nouns, and therefore unbounded cases.

(199) VoiceP

DP Voice

Voice VP

V DP

Even though througout the article MacDonald uses the label 'aspectual SE', one could in fact
argue that MacDonald's (2017) analysis makes the label lose its original meaning, as far as the
absence of a systematic aspectual effect on the predicate goes. Even though MacDonald (2017)
uses the boundedness of the direct object as a criterion to determine which predicates belong
to this class, his analysis treats the relevant constructions as instances of reflexive SE cases
where the agent and the dative end up being coreferential and the boundedness effect is not
directly related to aspect. What MacDonald (2017) codifies in the syntactic structure is, then,
the affectedness effect of 'aspectual' SE with respect to the subject, not the telicity information
–remember the stative examples in (198)–.
There is a second, different kind of critique of aspectual SE, going back to DiTullio (2012)
and Armstrong (2013), specifically that transitive verbs with aspectual SE should be
reinterpreted as verbs that have a use of SE that we could label 'agentive'.
Here are the arguments for the agentive approach, closely following DiTullio (2012; see
also Armstrong 2013):

i) The range of verbs that accept the so-called aspectual SE share a significant property of
the external argument: they cannot be anticausativised. This suggests (remember also the
discussion by Koontz-Garboden 2009) that these verbs force the presence of an external
argument distinct from the internal argument (note that in 200 it is possible to interpret for
instance a passive or middle SE).

(200) a. #El bocadillo se come.

the sandwich SE eats

74
 SE IN SPANISH

Intended: 'The sandwich gets eaten by its internal properties'

b. #El bolero se baila.
the bolero SE dances
Intended: 'The bolero becomes danced by its internal properties'

ii) Verbs that allow an anticausative pair reject the so-called aspectual SE:

(201) a. El televisor se rompió.

the TV SE broke
'the TV broke'
b. *Juan se rompió dos televisiones.
Juan SE broke two televisions

iii) Verbs that allow the so-called aspectual SE only do so when the external argument is an
agent, not a causer or any other type of initiator.

(202) a. Juan se lavó cincuenta platos.

Juan SE washed 50 plates
b. #La lavadora se lavó cincuenta calcetines.
the washing-machine SE washed 50 socks
(Unless personified)
c. *Este jabón se lava cincuenta calcetines.
this soap SE washes 50 socks

iv) Verbs that are otherwise transitive and might in principle satisfy the aspectual requisite
do not allow the so-called aspectual SE if their subject is not an agent.

(203) a. *Juan se recibió catorce cajas.

Juan SE received 14 boxes
b. *Juan se oyó el disparo.
Juan SE heard the noise
c. *Juan se reconoció su error.
Juan SE admitted his mistake

v) The so-called aspectual SE imposes the entailment that the subject performs the event
with willful intent. In (204a) the subject might smell the roses by accident, while in (204b) one
rather interprets that Ana's job is to check that roses smell fragantly or something along these
lines where Ana smells them on purpose and intending explicitly to do so. Similar points can
be made about other perception verbs, like ver 'see', oír 'hear'.

(204) a. Ana olió cincuenta rosas.

Ana smelled 50 roses
b. Ana se olió cincuenta rosas.
Ana SE smelled 50 roses

vi) One can have sentences where the so-called aspectual SE appears but where there are
bare nouns or other unbounded expressions. The facts are more controversial here from the
perspective of grammaticality judgements. Armstrong (2013) offers the example in (205b),
which he –as well as the author of this article– considers grammatical. García Fernández

75
 ANTONIO FÁBREGAS

(2015), who treats transitive verbs with SE as cases of aspectual SE, finds the sentence
ungrammatical.

(205) a.Sócrates se tomó veneno. [Romero Pascual & Teomiro 2012]

Socrates SE took poison
b. El niño se tomó veneno.
the boy SE took poison

We would like to note in passing that there is some possibility that the examples in (198)
should be analysed as instances of agentive SE. The nature of the verbs as expressing knowing
or thinking makes it impossible to test whether only human subjects are allowed –as the verbs
anyways would involve this type of subject–, but it is possible to point out that (198) tends to
be perceived as involving more of a conscious effort on the part of the speaker to have acquired
the mental state through his or her personal involvement in studying it, memorising it, etc., and
perhaps maintaining the state denoted by the verb. In the reflexive version of creer 'believe',
one gets the feeling that the subject insists on holding that thought. Similarly, saberse
presupposes that there was a previous learning process where the subject tried to acquire the
knowledge on the expectation that someone might actually check that he has the knowledge:
Juan se sabe la lección 'Juan SE knows the lesson' suggests that Juan would make an effort to
study it again if he thinks that he is forgetting that knowledge.
In all fairness, not all aspectual SE cases have been claimed to be agentive SE cases in the
literature that defends the existence of this type of SE. Armstrong (2013) notes that (205b)
lacks agent entailments and proposes that it is an instance of a transitivising SE, which we will
discuss immediately. This weakens the argument against aspectual SE, but De Benito (2021)
identifies in her corpus study more cases of apparent aspectual SE with unbounded objects. It
is true that in some cases there are adjectives (206a) that might delimit the noun in an indirect
way –similarly to how some nouns modified by adjectives can appear as preverbal subjects
without determiners– but this is not so in all cases (206b).

(206) a. Se ha tomado leche cruda.

SE has taken milk raw
'He had raw milk'
b. Pasarnos noches en la cuadra.
pass-SE nights in the stable

These arguments support that at least some of the uses of SE that have been classified as
aspectual are rather considered as agentive or some other type of construction. Di Tullio (2012)
proposes reducing all cases of aspectual SE to agentive SE cases, without providing a formal
analysis. Armstrong (2013), in contrast, dissolves the notion of aspectual SE but does not claim
that all instances of transitive predicates with SE are agentive. For him, the cases above (with
the exception of 205b) are all cases of agentive SE, but he allows a second type of SE, transitive
reflexive SE, which accounts for the cases where there is no agentive entailment. (205b) can
be suplemented with a modifier that indicates absence of agentivity or willful intent (207).

(207) El niño se tomó veneno sin querer.

the boy SE had poison without wanting

This transitive reflexive SE use identified by Armstrong (2013) is defined through negative
properties: it does not impose an agentive reading to the subject and it does not have a
systematic aspectual effect on the predicate: its role is simply to define a different type of

76
 SE IN SPANISH

predicate by composing with the predicate. Other relevant cases of this transitive SE use are
provided in (208), none of which involves an agent with wilful intent.

(208) a. La mesa se come la habitación.

the table SE eats the room
b. Me dejé las llaves en casa.
SE left the keys at home
'I left the keys at home'
c. Juan se encontró a María.
Juan SE found DOM María
d. El viento se llevó las hojas.
the wind SE took the leaves

A very similar example to (208a) is in fact mentioned by DiTullio (2012), who claims that
it does not constitute a counterexample to her proposal as far as the verb comerse has a different
meaning as the verb comer, unlike the normal cases of agentive SE. The transitive reflexive SE
is, in Armstrong's analysis, a type of SE that occupies one of the argument positions of VP and
incorporates to the V head from there. This is what makes (208a) have a different meaning
from the SE less element; defining a new V, this transitive reflexive SE can alter the selectional
restrictions of the predicate, and will not have any defined aspectual effect on it, as it acts as
an argument. (206) shows the structure proposed in Armstrong (2013; see also De Cuyper
2006).

(209) vP

DP v

v VP

DP V

V+se se

Note that in (209) the direct object is necessarily located in a specifier position, just like in
the relevant examples by MacDonald (2017): this opens the door to explaining the difficulty
of having bare nominals in this context through the configuration and not through any kind of
aspectual role.
In contrast, the structure associated by Armstrong (2013) to agentive SE treats SE as an
instantiation of a particular type of little v head, a version of a DO little v that forces presence
of an agent, and moreover comes tagged with features that force the rest of the structure to
define an initial and a final subevent, which SE checks, in an approach similar to Kempchinsky
(2004).

77
 ANTONIO FÁBREGAS

(210) vP

DP v

vDO AspP
se
Asp VP

V DP

The features in the little v represented by SE determine that the internal argument has to be
bounded and that there is an accomplishment reading, in accordance with Nishida's (1994)
description. SE then has two different sources in the purely agentive cases and in the transitive
cases, where in the second SE is not part of the predicate and therefore cannot impose an
aspectual interpretation or force an agentive construction. In this sense, Armstrong contrasts
with DiTullio, who had no specific way to account for the aspectual effects in the transitive
clauses. The price to pay is that SE has to be treated either as an argument or as a predicate,
somehow circularly, depending on whether the structure has effects on aspect or not.

5.4. Possibly agentive statives

The possible reduction of aspectual SE to cases where what is relevant is in fact the
agentivity entailments –total reduction, as in DiTullio (2012), or partial reduction, as in
Armstrong (2013)– finds another battling ground in the case of the pair in (211).

(211) a. estar
beestar
b. estar-se
beestar-SE

The copulative verb that in Spanish is traditionally associated to stage level predicates is,
without SE, uncontroversially stative and non-agentive. The cases with SE, however, display
different properties which are most easily noticed in the imperative: imperative sentences with
this predicate require (212) –particularly in the second person singular; there is more variation
with respect to the second person plural–.

(212) Está-*(te) quieto.

be-SE still
'Be still!'

There are three other contexts where speakers identify some contrast, although not as strong
as the imperative: the first one is in combination with agent-oriented adverbs.

(213) Juan ??(se) estuvo callado a posta.

Juan SE was quiet willingly
'Juan was willingly quiet'

The second is as the infinitive of a control verb, although here the contrast is much less clear
for many speakers, who don't need the SE form in such cases:

78
 SE IN SPANISH

(214) Pedro obligó a María a estar-??(se) despierta.

Pedro forced DOM María to be-SE awake

Finally, some speakers note that the use of the eventive verbal pro-form hacerlo 'do it' is
more natural in combination with SE.

(215) Lo que hizo Juan en la reunión fue estar-??(se) callado.

that what did Juan in the meeting was be-SE quiet
'What Juan did at the meeting was to stay quiet'

Several recent works have addressed the issue of the pair in (208), and as in the case of the
transitive aspectual SE cases, the controversy is whether this is an instance of an aspectual or
an agentive SE (Morimoto 2008, 2011; Gómez Vázquez & García Fernández 2013, García
Fernández & Gómez Vázquez 2015, Marín & Fábregas 2018). In a nutshell, Morimoto (2008)
interprets the facts above as a sign that the addition of SE alters the aspectual properties of the
copula by forcing it to denote a result state; in order to be a result, the predicate now has to
presuppose a previous event that leads to that result, and Morimoto proposes that this
presupposition involves an agentive event, which explains the combination with control verbs
or imperatives. Thus, for Morimoto (2008) the verb with the reflexive is still stative, but
presupposing a previous event which might introduce some agentivity entailments.
García Fernández & Gómez Vázquez (2015) closely follow the general approach to
aspectual SE in De Miguel & Fernández Lagunilla (2000) and propose that estarse is a complex
event composed of an achievement and a result state –thus operationalising the conclusion in
Gómez Vázquez & García Fernández (2013) that the verb with SE displays mixed properties
between states and events. In contrast to Morimoto (2008) this means that this verb contains
an eventive component, but one that does not involve any type of incremental change because
it involves an achievement. This explains that the verb disallows progressive periphrases,
although marginally giving a preparatory stage reading ('is about to become quiet') and rejects
modifiers that presuppose incrementality:

(216) a. *Juan se está callado poco a poco.

Juan SE is quiet little by little
b. ??Juan ya iba estándose callado.
Juan already went being-SE quiet
Marginally interpretable as 'Juan already was about to become quiet'

Having an achievement component, the verb cannot in principle have agent entailments, as
generally achievements are non-agentive.
Marín & Fábregas (2018) argue against the analyses with a result state involving an
achievement, showing that the predicate is atelic, unlike what the achievement + result state
would predict:

(217) *Juan terminó de estarse callado.

Juan ended from be-SE quiet
Intended: 'Juan finished the process of becoming quiet'

In the analysis of these authors, the aspectual structure of estarse is affected with respect to
the SE-less member, but in the sense that the state gets added an initial boundary that sets a
starting point of change, without defining the final point where the change culminates.

79
 ANTONIO FÁBREGAS

(218) [-----------------------

This initial boundary is associated to the presence of an agent which is added to the structure,
in a sense close to what García Pardo (2021) considers an agentive state where the state defined
by the copulative verb is incremented by a second head, higher, which introduces a causative
relation between an agent and the state. This causative relation is what defines the initial
boundary, explaining that estarse gets an agentive reading and gets its aspectual properties
altered so that it does not denote a pure state, but rather an inchoative state, adding some degree
of eventivity to the verb without defining a full telic event or properly a result state following
the culmination or endpoint of the previous event. Thus, the agentivity entailments are more
relevant in the pair than the aspectual effects, adding to the type of analysis that treats aspectual
SE as an agentive SE whose aspectual effects are identifiable.

5.5. Aspectual SE and the intransitive uses

In two separate works, García Fernández (2011, 2015) has questioned the specific analysis
of aspectual SE in the case of intransitive, typically achievement predicates. His research leads
to two conclusions: (i) many of the cases of intransitive predicates with SE that have been
analysed as aspectual are probably instances of other types of SE –while admitting that
aspectual SE exists in combination with transitive incremental verbs, typically
accomplishments–; (ii) in the cases that remain of real aspectual SE with intransitive predicates,
the correct characterisation of such cases is the notion of traversing a boundary, in many cases
the initial boundary of movement, the source where movement is initiated. Let us review this.
García Fernández (2011) notes first that the list of intransitive verbs with aspectual SE is
pretty short in the literature, and secondly that there are no less than 8 aspectual structures
proposed for them in the works that extend the notion of aspectual SE to intransitive verbs,
without any attempt to delimit the conditions for belonging to one or the other class. García
Fernández (2011) proceeds to reanalyse the most typical cases, which are listed in (219):

(219) a. Juan se murió.

Juan SE died
b. Juan se durmió.
Juan SE slept
'Juan fell asleep'
c. Juan se fue de casa.
Juan SE went from home

These, in fact, involve the three main ways in which García Fernández (2011, and then more
specifically for movement verbs, 2012) reanalyses the aspectual SE cases.
The case of (219a) is an instance of a case where aspectual SE apparently would not have
any visible aspectual effect in the predicate, as we already discussed above, because the
predicate without SE is also telic, achievement-like and with exactly the same difficulty of
expressing a reversible result state. Given the aspectual neutrality of this aspectual SE, García
Fernández (2011) rejects the option that it simply marks an aspectual property that was already
present in the predicate and proposes that it should be reanalysed as closed to anticausative SE,
in the sense that this SE marks the notion that the event happened by accident, without the
intervention of any external entity that causes the death, and with the sole internal participation
of the subject, which is not an agent either. Though grammatical, (220) is deemed as false
because we know that García Lorca did not die by accident, but was murdered.

80
 SE IN SPANISH

(220) García Lorca se murió en Granada.

García Lorca SE died in Granada

This anticausative SE involving through its presence the notion that the event happens by
accident, without any kind of external control, is also proposed in other cases of SE, now with
movement verbs:

(221) a. Durante el accidente, el muerto se salió del ataúd.

during the accident, the corpse SE exited the coffin
'During the accident, the corpse fell out of the coffin'
b. Si abres el grifo, se sale el agua.
if open the faucet, SE exits the water
'If you open the faucet, water spills over'

The case in (219b), involving dormirse, has the problem that, if right, the aspectual SE
analysis would be a unique instance in Spanish where aspectual SE moves the predicate from
an activity (dormir = to sleep) to an achievement (dormirse = to fall asleep). García Fernández
(2011) proposes that this verb is in fact an instance of an anticausative SE which forms pair
with the causative verb in (222).

(222) a. Pedro durmió a su hijo.

Pedro make.sleep DOM his son
'Pedro made his son sleep'
b. Pedro se durmió.
Pedro SE slept
'Pedro fell asleep'

What gives the impression of aspectual SE is the existence of a second intransitive verb
dormir. García Fernández proposes that this verb, in fact, is a different verb from dormirse,
although morphologically related to it, and notices that in other Romance languages the verbs
are differentiated by something more than the presence of SE: cf. French dormir / s'endormir
and Italian dormire / addormentarsi. Thus, the presence of SE in the case of (222b) is not
aspectual, and the aspectual difference is unrelated to it.
The case in (219c) combines with the 'by accident' or 'without the participation of an external
agent' explanation with a locative notion of source. García Fernández (2011, 2015) proposes
that the presence of SE in a robust group of movement verbs is related to the existence of a
source of motion where the movement starts. In the following examples, the absence of SE
makes it impossible that a from-phrase is present.

(223) a. Juan *(se) fue de su casa.

Juan SE went from his house
'Juan left his house'
b. La estatua *(se) cayó de la cornisa.
the statue SE fell from the cornice
c. El dinero #(se) me sale del bolsillo.
the money SE me exists from.the pocket
'The money falls from my pocket'
d. *(Me) llevé el libro de la biblioteca.
SE carried the book from the library
'I took with me the book from the library'

81
 ANTONIO FÁBREGAS

In other words: if one wants to express the origin of the movement, one must mark with SE,
which means that in such cases the presence of the clitic refers to the presence of some
boundary that has to be traversed. One could extend this analysis to the pair dejar 'leave' and
dejarse 'to leave-SE', where the second involves that the speaker had the intention to take
something with him but fails to do so, and moves away from the location where the object has
been left (Me dejé las llaves en casa 'I left the keys at home', where the subject must move
away from the location of the keys).
In fact, this is expanded in García Fernández (2015), where the aspectual interpretation of
these movement predicate facts is maintained or even restored with respect to (2011). García
Fernández (2015) proposes that the presence of SE with movement verbs introduces a notion
of origin, which explains the oddness of the following cases, because we do not conceptualise
these entities as coming from any location.

(224) a. #Se salió el sol.

SE came.out the sun
b. #A Juan se le salió un grano.
to Juan SE him came.out a pimple
c. #Se cayó un rayo.
SE fell a lightning
d. #La nieve se cayó en el carro.
the snow SE fell on the cart
(unless the snow was already located on some roof, from where it falls)

This SE, then, alters the argument structure of the predicate by marking movement as
coming from some particular place. Following Sánchez López (2002), García Fernández
(2015) also proposes that the addition of SE in these cases makes the predicate move from an
unaccusative to an unergative verb where the subject is defined as an external argument.

(225) a. Sale humo (de la chimenea).

comes.out smoke from the chimney
b. *Se sale humo (de la chimenea).
SE comes.out smoke from the chimeney
(cf. El humo se sale de la máquina 'The smoke SE comes out of the machine')

The proposal that (225b) is an unergative verb explains the contrast between (225a) and
(225b) in a similar way than MacDonald (2017) explained the presence of boundedness in
direct objects with the transitive aspectual SE cases: the subject can only be a bare noun in a
complement position.

(226) a. SE-less directional movement verb

VP

V NP
come out smoke

82
 SE IN SPANISH

b. SE-marked directional movement verb

VoiceP

DP Voice
the smoke
Voice VP
se
V PP
come out
P ...

This author adopts Kratzer's (1996) hypothesis to introduce the external argument with
Voice rather than vP, but beyond this his approach is similar to Basilico's (2010) and the rest
of approaches that treat SE as one manifestation of specific flavours of the predicate that
introduces the external argument. In (226b), the SE marker introduces a voice head that projects
the subject as the external argument. This has the syntactic effect of forcing it to be introduced
by a determiner and of liberating the complement position of the VP to introduce the origin PP.
Admittedly, this does not explain why that PP must be origin, but one could especulate that
there is a semantic correlation between an external argument taken as the origin of the event,
because it initiates it, and the locative notion of origin of the path of motion. Moreover, one
could think that the structure of the VP is more complex and a directional preposition
expressing movement to a location is also present in both structures, so that the internal
argument position that the Voice head liberates must be occupied by an origin PP, because the
directional PP is already present in the structure.
Once proposed, that structure can also be extended to some of the other examples of
apparent SE with verbs that are not of movement. The effect obtained with morirse, where the
subject is the sole responsible of the event, and the effects related to the accidental happening
of the event without intervention of an external agent, could also be explained in the structure
(226b), because the argument that participates in the event is already occupying the external
argument position, so there is no place for another entity that is interpreted as the causer of the
eventuality. It is true, however, that the notions of accidental event and 'subject as the only
responsible of the event' are not coextensive, and should be viewed at best as two alternative
readings of the same structure, which one would be forced to apply to different predicates,
perhaps due to their lexical differences. For instance, the contrast in (227) suggest that with
salir the accident reading is imposed –a corpse cannot be responsible of coming out of a place–
, while with ir the sole responsible entity is imposed, as there is no accidental event reading
that gets imposed and without SE one can interpret that some other entity moves the corpse.

(227) El muerto #(se) salió de la tumba.

the corpse SE came.out of the tomb
(228) El muerto (#se) fue de Madrid (a Bilbao).
the corpse SE went from Madrid to Bilbao

García Fernández (2015) also extends this interpretation of SE as a head in Voice imposing
an external argument reading to the transitive aspectual cases, like (229), where he argues that
there SE-marking is optional because the verb alone already contains a Voice head that defines
an external argument that initiates the event. Thus, SE is not necessary to impose that argument
structure, but it can be added –sometimes with lexical effects on the meaning of the predicate–
without substantial changes on the syntactic construal.

83
 ANTONIO FÁBREGAS

(229) Juan (se) comió un bocadillo.

Juan SE ate a sandwich

Thus, to recapitulate so far: we have seen that the notion of aspectual SE is controversial.
There is a discussion with respect to (i) to what extent this aspectual SE can be reanalised as a
reflexive SE that is introduced in extra argument positions created by additional verbal heads,
which are responsible for the aspectual effect; (ii) to what extent the effects on the subject that
this SE has, as inducing forms of affectedness to the subject, or increasing the agentivity of the
subject as sole responsible of the event, or excluding the possible participation of additional
external agents, can be used to replace the aspectual interpretation and (iii) what specific
aspectual change does this SE produce, with transitive verbs being somehow more
homogeneous –they are accomplishments which involve incremental notions– than intransitive
verbs –where the effect has been argued to be to add a result state to an achievement, but where
the effect is not always as clear–. The analyses vary with respect to whether SE here is taken
as a reflexive argument filling a position created by an additional verbal or prepositional head,
with possible agent effects through coindexation, or as the manifestation of a verbal head that
introduces arguments in different configurations than the SE-less version. In the second case,
there is more or less explicit connection with the anticausative SE. Thus, the analyses that do
not want to acknowledge a separate category of aspectual SE reduce it to a reflexive or to an
anticausative SE. See also §6.4 below, where we analyse the case of figure SE as one possible
way to reanalyse some of the intransitive 'aspectual cases', but where we will show that the
reduction cannot be complete.

5.6. Relation between anticausative SE and aspectual SE

The analyses that we have seen so far try to reduce the aspectual SE to reflexives or
anticausatives, but there is an additional option in the literature: to treat at least part of the
anticausative SE cases that we revised in §4 as actually aspectual.
The core empirical facet of the phenomenon that gives grounds to attempt this kind of
unification is the existence of a class of anticausative verbs that can be marked with or without
SE without incidence on their argument structure. Let's repeat the list of verbs here, taken from
Vivanco (2015), although remember that this list is subject to variation among speakers and
varieties.

(230) caramelizar 'to become caramel', cicatrizar 'to scar', cristalizar 'to chrystalise', cuajar 'to
curdle', despertar 'to wake up', encoger 'to shrink', enmohecer 'to become mouldy',
ennegrecer 'to become black', enrojecer 'to become red', mejorar 'to get better', reventar
'to explode'

Given that there are no effects in the argument structure of pairs such as (231a) and (231b),
this opens the possibility that the difference between such pairs is aspectual.

(231) a. Juan despertó de la siesta.

Juan woke.up from the siesta
b. Juan se despertó de la siesta.
Juan SE woke.up from the siesta

A number of scholars have argued that the difference between the SE-marked and the SE-
less version of anticausative verbs that alternate is in fact aspectual in Italian and French (Zribi-

84
 SE IN SPANISH

Hertz 1987, Labelle 1992, Folli 2001, Labelle & Doron 2010): the idea is that the SE-less
version can be telic or atelic, but the SE version must be telic:

(232) a. Le ciment a durci pendant trois heures.

the cement has hardened for three hours
b. *Le ciment s'est durci pendant trois heures.
the cement SE is hardened for three hours

Notice that in French (232) or Italian, however, there is a second difference: the choice of
the auxiliary, which might have a non-trivial effect on the aspectual information, affecting the
extent to which this generalisation can be extended to modern Spanish, which does not
differentiate between perfect auxiliaries. However, the authors mentioned above take the
choice of the auxiliary as a direct consequence of the presence of SE, and thus reduce the effect
to the presence or absence of SE, not the auxiliary. Basilico (2010) and Cuervo (2003, 2014)
extend the pattern to Spanish and argue that SE concentrates on a result, while its absence,
marked as ø, concentrates on a process.
Extending these facts to Spanish is initially problematic because, as Vivanco (2015, 2021)
notes, the SE-less version of the verbs in the list of alternating verbs might be telic, like the SE
version of the same verbs. Some of them are in fact achievements.

(233) a. Juan despertó.

Juan woke.up
b. La herida cicatrizó.
the wound became a scar

The approach is also problematic from the perspective of obligatorily SE-marked verbs: not
all of them are always telic, so if SE marked telicity (or a result), these predicates should always
be telic.

(234) El té se enfrió durante diez minutos.

the tea SE cooled for ten minutes

It is also far from obvious that verbs which never mark with SE their anticausative version
tend towards an atelic construal, which one should expect if in principle SE is related to a telic
reading. The list in (235) contains verbs with clear achievement interpretations, like resucitar
'resurrect' or enmudecer.

(235) adelgazar 'to become slim', aumentar 'to increase', blanquear 'to whiten', cambiar 'to
change', disminuir 'to diminish', empeorar 'to worsen', engordar 'to become fat', enfermar 'to
get sick', enloquecer 'to become crazy', enmudecer 'to fall silent', envejecer 'to grow old',
hervir 'to boil', rejuvenecer 'to become younger', resucitar 'to resurrect'

Even acknowledging that the original proposal is that SE-less verbs can be telic or atelic,
these facts are problematic to the extent that the claim is that SE relates to telicity and its
absence, to atelicity at least in a group of alternating verbs: we should expect that the role of
SE remains homogeneous across anticausative verbs if its role is truly aspectual. Moreover,
one cannot simply claim that the absence of SE directly correlates with atelicity, even for the
class of anticausatives that alternative between the two construals.
Vivanco (2021), in contrast to Vivanco (2016) who rejected entirely the aspectual role of
SE in anticausatives, proposes a new vision of the aspectual role of SE in anticausatives. Her

85
 ANTONIO FÁBREGAS

claim is that SE in anticausatives is related to predicates that express change that goes through
a multi-point scale of values that is closed or bounded, hence telic, while its absence simply
signals the absence of a closed or bounded multi-point scale of values. This means that the
absence of SE is compatible with two kinds of changes: changes across a multi-point scale, but
which remains open and therefore gives an atelic reading, or changes across a single transition
in two points, as in the case of an achievement.
Like that, resucitar 'resurrect', despertar 'wake up', and the other anticausative achievements
that are not marked with SE fall within the predictions of the analysis: they express single
transitions from one point (not having the property) to another point (having the property),
without intermediate values; the predicate is telic, but there is no multi-point scale that
underlies it. The atelic reading of the SE-less predicates, such as enmohecer 'to become moldy',
directly follows: there is a multi-point scale with different values of moldiness, but that scale
is open and the verb is atelic. In the case of SE-marked verbs which seem to be atelic, such as
(234) above, Vivanco (2021) proposes that the predicate is indeed telic, but that the durative
modifier introduces an additional interpretation that atelicises the event by selecting only a part
of the internal closed scale.
Finally, with respect to the case of achievement predicates that can be marked with SE or
not, such as (231) above, repeated here for convenience, Vivanco proposes that the effect of
SE is in accordance with the multipoint scale requisite: the SE has the effect of producing an
extended achievement where the initial and the final point of the change do not coincide in
time, so that the running time of the event –in an intuitive sense– becomes extended.

(236) a. Juan despertó de la siesta.

Juan woke.up from the siesta
b. Juan se despertó de la siesta.
Juan SE woke.up from the siesta

The extension, more specifically, means this: the SE-less version denotes a mere transition
between not having a property and having it, while the SE-marked version extends the
achievement so that it denotes the endpoint of the previous state and the initial point of the
following state. This has the effect of separating in time the initial and final point of the
transition, which in a normal achievement are simultaneous because the change is
instantaneous. Therefore, with these predicates the SE-version allows combinations with phase
verbs singling out the initial or the final stage:

(237) a. Juan comenzó a despertar-??(se).

Juan began to wake.up-SE
b. Juan terminó de despertar-*(se).
Juan finished of wake.up-SE

The SE version, in allowing a temporal extension of the achievement due to its preference
in marking multi-point scales, also makes the verb compatible with tardar 'take long' in the
reading where the internal duration of the event is measured –not the one where the starting
point of the event is delayed by some time–.

(238) a. Juan tardó una hora en despertar.

Juan lasted one hour in wake.up
'Juan's waking up was delayed by one hour', not *'The process of waking up took Juan
one hour'
b. Juan tardó una hora en despertar-se.

86
 SE IN SPANISH

Juan lasted one hour in wake.up-SE

'Juan's waking up was delayed by one hour' or'The process of waking up took Juan one
hour'

Thus, it seems that SE might be related to a specific aspectual value, although that value
cannot be directly stated in terms of a coarse distinction between telic and atelic readings: SE
correlates with a multi-point scale which incrementally defines a change of state. This is of
course quite directly related to the main interpretation identified for aspectual SE in transitive
verbs since Nishida (1994), which suggests that a unification of the two is possible.
The case of intransitive verbs with aspectual SE, discussed in the previous subsection, is
still difficult to unify from this perspective, unless one takes the presence of an origin
component in the movement verbs as a sign that the multi-point scale is also defined for the
case of the path, defining both an origin and a goal of motion that close the scale while letting
it be extended in space. Note, in this same respect, that one could try to extend the analysis to
cases like morir / morirse, where the contrast in (239) seems to be partially replicated.

(239) a. Juan tardó una hora en morir.

Juan lasted one hour in die
'Juan's death was delayed by one hour', not ??'The process of dying took Juan one hour'
b. Juan tardó una hora en morir-se.
Juan lasted one hour in die-SE
'Juan's death was delayed by one hour' or'The process of dying took Juan one hour'

The unification is also promising from the perspective of the analysis that Vivanco (2021)
proposes for the anticausative SE; similarly to the approaches that treat aspectual SE as the
manifestation of a verbal head, this author places SE in little v, and proposes –like
Kempchinsky (2004)– an aspectual head immediately below v whose information –in terms of
scalarity and boundedness– is sensitive to the existence of a degree phrase below it, which is
what determines the presence or absence of SE: when there is a multivalue closed scale, SE
spells out v, and when it is lacking, it is spelled out as ø.

(240) a. vP No multivalued scale (achievement)

v AspP
ø
Asp PredP

DP Pred

Pred AP

87
 ANTONIO FÁBREGAS

b. vP Multivalued scale, open (atelic)

v AspP
ø
Asp PredP

DP Pred

Pred DegP

Deg AP
(open)
c. vP Multivalued scale, closed (telic)

v AspP
SE
Asp PredP

DP Pred

Pred DegP

Deg AP
(closed)

d. vP Extended achievement

v AspP
SE
Asp PredP

DP Pred

Pred DegP

Deg AP
(x2)

This ends our discussion of aspectual SE; let us now move to other lexical values of SE that
affect the nature of the predicate.

6. Predicate-affecting SE uses (4): other

Without doubt, among the so-called paradigmatic SE uses, which we have relabeled as
'lexical', the three most studied and established types are the reflexive-reciprocal, the
anticausative and the aspectual, at least for the small set of cases where it might be present,
which we will come back to. However, in the literature there are at least three other minor
classes, the causative or factitive SE (§6.1), the antipassive SE (§6.2), the inherent SE (§6.3),
and the figure reflexive SE (§6.4). Beyond these, and given the high degree of idiosyncrasy
that the presence of SE has with some predicates, we will discuss what can be considered

88
 SE IN SPANISH

lexical –therefore fully unpredictable– in the case of SE (§6.5). This section ends with a brief
interim summary of all the lexical uses of SE reviewed up to now (§6.6).

6.1. Causative or factitive SE

The so-called causative or factitive SE (Cano Aguilar 1981, Kovacci 1986, Sanz & Laka
2002, DiTullio 2012, Armstrong & Kempchinsky 2021) has received relatively less attention
than the other uses that we have discussed so far. This value of SE is illustrated in the examples
in (241).

(241) a. Juan se hizo un traje con este sastre.

Juan SE made a suit with this taylor
b. Juan se cortó el pelo en esta barbería.
Juan SE cut the hair in this barbershop
c. Juan se construyó una casa.
Juan SE built a house

The crucial criterion to determine that this is a factitive use of SE is the equivalence in
meaning between (241) and (242): the subject is not the agent of the event, or even the person
that controls the development of the event, but a causer that initiates the event indirectly by
commissioning it to someone else, who acts as the real agent that develops the process.

(242) a. Juan se hizo hacer un traje con este sastre.

Juan SE made make a suit with this taylor
'Juan made a taylor make him a suit'
b. Juan se hizo cortar el pelo en esta barbería.
Juan SE made cut the hair in this barbershop
'Juan made someone give him a haircut in this barbershop'
c. Juan se hizo construir una casa.
Juan SE made build a house
'Juan made someone build a house for him'

There is an agreement in the literature that this use of SE comes from the reflexive use: only
predicates that have an empty slot for the direct object or the indirect object can participate in
this construction. Kovacci (1986), in fact, treats factitive uses of SE in two categories: factitive
intransitive (243), where the SE form appears in the place of a direct object, and factitive-
transitive (244), where the SE form is in a bona fide indirect object position.

(243) a. Juan se retrató en esta tienda.

Juan SE photographed in this shop
b. Juan se afeitó en esta barbería.
Juan SE shaved in this barbershop
(244) a. Juan se operó la mano en este hospital.
Juan SE operated the hand in this hospital
b. Juan se hizo un tatuaje en esta tienda.
Juan SE made a tattoo in this shop

Even though this is a clear connection between reflexive SE and factitive SE –remember
that SE in reflexives and reciprocals is restricted to coreference between subject and direct or
indirect object–, factitive SE never allows the form sí mismo, as shown in (245), sentences

89
 ANTONIO FÁBREGAS

which force Juan to be the agent of the event and not the instigator that makes someone else
do it.

(245) a. Juan se retrató a sí mismo en esta tienda.

Juan SE photographed himself in this shop
b. Juan se operó la mano a sí mismo en este hospital.
Juan SE operated the hand to himself in this hospital

The way in which this fact should be taken is far from clear. One could claim that this is an
epiphenomenon of the property of sí mismo, that imposes an agentivity constraint on the
subject, in such a way that it excludes the interpretation of the subject as instigator of a different
agent. Remember however that the form is not incompatible with the bona fide experiencer
subject of psychological verbs (246), so this restriction at best should be taken as a ban on
causers rather than as a requirement on agents.

(246) Juan se odía a sí mismo.

Juan SE hates DOM himself

The situation is further complicated by the existence of reflexive verbs that cannot accept
doubling with sí mismo, as we discussed extensively in §3.1 –remember the grooming verbs
and posture verbs, such as afeitarse 'shave', levantarse 'stand' or tumbarse 'lie down', which
tend to reject the strong reflexive form too–. In fact, the factive interpretation appears typically
with verbs that either are themselves grooming verbs, or that share with them the property of
expressing events that typically affect the subject or are performed for the subject's benefit
involving entities of the personal sphere, like operar 'operate', hacer un traje 'to make a suit',
etc.
Another point of contact with other uses of SE is the agentive SE, reinterpreting the
aspectual SE as in DiTullio (2012). Like factitive SE, agentive SE cannot be incremented with
sí mismo, and involves an event that somehow affects the subject. Just like agentive SE,
factitive SE requires human subjects that in principle could have been defined as agents. There
is no way of interpreting (247) as factitive without personalising the subjects.

(247) a. La empresa se ha hecho un logo nuevo.

the company SE has made a logo new
'The company made someone design a new logo'
b. El gusano se ha operado.
the worm SE has operated

The difference is of course that in factitive SE the reading is that of an instigator, not of the
agentive entity that performs the event. In fact, DiTullio (2012) argues that this is directly
expected if the grammar does not differentiate strictly between agents and causers within the
argument structure of a predicate: if causers and agents are syntactically identical, together with
other possible interpretations of the subject, and the difference between them is left to a
conceptual semantic component, one could take factitive SE to be of the same type as agentive
SE, with the only difference that here we are talking of different readings of the same subject
and the general notion of 'initiating an event'. The agent initiates the event volitionally and
controls the development of the action, while the causer initiates the event also volitionally but
does not control the development of the action, with the possibility of delegating into someone
else the specific carrying out of the event.

90
 SE IN SPANISH

One syntactic fact supports the view that factitive SE is not different from agentive SE even
if it triggers the interpretation that the event has a different agent: in contrast with the syntactic
causative structure, there is no space to introduce an agent.

(248) a. Juan hizo al peluquero cortarle el pelo.

Juan made to.the barber cut-him the hair
'Juan made the barber give him a haircut'
b. *Juan se cortó el pelo al peluquero.
Juan SE cut the hair to.the barber

What one can do is to introduce an adjunct marked with a strong lexical preposition –
instrument, company, place; never agent-introducing prepositions– that introduces another
semantic participant interpretable as the notional agent, or a space where some worker carry
the event out.

(249) a. Juan se afeita siempre con este barbero.

Juan SE shaves always with this barber
b. Juan se afeita siempre en esta barbería.
Juan SE shaves always in this barbershop

A lexical fact also supports this view that the agent and instigator interpretations should be
viewed as different conceptual semantic interpretations and not as receiving different structures
with two locations for SE: as noted in Sanz & Laka (2002), the availability of the factitive
reading depends on how we conceptualise the event denoted by the predicate.

(250) a. Pedro se arregló la chaqueta en Valladolid.

Pedro SE fixed the jacket in Valladolid
b. Pedro se hizo la chaqueta en Valladolid.
Pedro SE made the jacket in Valladolid
c. Pedro se cosió la chaqueta en Valladolid.
Pedro SE sew the jacket in Valladolid
d. Pedro se tiñó la chaqueta en Valladolid.
Pedro SE dyed the jacket in Valladolid
e. Pedro se planchó la chaqueta en Valladolid.
Pedro SE ironed the jacket in Valladolid
f. Pedro se lavó la chaqueta en Valladolid.
Pedro SE washed the jacket in Valladolid

The factitive interpretation, in my own judgement, is very easy to obtain in (250a) and
(250b), but more difficult –but not completely out, pace Sanz & Laka (2002)– in (250c,d,e),
and totally out in (250f). Note that in all these examples we have transitive predicates which
do not select an indirect object goal, and where the SE can be argued to act as a beneficiary; I
am not aware of any syntactic difference between these verbs, although one can imagine
conceptual differences related to how much expert skill is required to perform the event in
particular, and therefore how much sense it makes that someone commissions someone else to
perform the event for him instead of trying to do it by himself. I am claiming, indeed, that
putting together, sewing or dying a jacket require more specialised knowledge than washing or
ironing it, and therefore these last two verbs are less easy to interpret in a factitive construal.
In some other cases, our world knowledge telling us that some people make a living shaving
or giving haircuts to others allows us to interpret the event as factitive.

91
 ANTONIO FÁBREGAS

Finally, if the factitive SE is not a real class of SE structurally different from agentive,
reflexive or aspectual SE, what we expect is that the factitive reading of the subject can be
present in predicates without the intervention of SE, and this is actually confirmed.

(251) a. Carlos III construyó la Puerta de Alcalá.

Carlos III built the Gate of Alcalá
b. Nerón quemó Roma.
Nero burnt Rome
c. Hitler abrió campos de concentración en Polonia.
Hitler opened camps of concentration in Poland
'Hitler opened concentration camps in Poland'

In all these cases, without SE, we see that we place as subjects of an event individuals that
our world knowledge tells us were not the agents of the events: construction workers, paid
mercenaries or individual soldiers and officials were the entities responsible for carrying out
these events, and the subjects were the instigators that gave the order and never got their hands
dirty. Now, if this reading can be obtained without SE, we confirm indirectly that the distinction
between agent and causer / instigator does not depend on the presence of SE, and therefore that
factitive SE should not be considered as an independent class of SE constructions. Rather, the
evidence that we have just revised points out to the conclusion that the so-called factitive SE
structures are additional interpretations that the subject can receive in the reflexive, agentive or
aspectual SE configuration, simply in cases where the notion of external argument can be read,
given our world knowledge, as the instigator of an event that is commissioned to someone else.
The presence of SE, from this perspective, could be viewed simply as a benefactive dative SE,
which means that the subject is the one that gets a benefit out of the action being performed.
We are only aware of one analysis where factitive SE is different from other syntactic uses
of SE in terms of configuration. Masullo (1992: 236-238) in fact adopts a clearly reciprocal
analysis of factitive SE where, crucially, he argues that this is a use of SE that is syntactically
different from the reflexive, although related to it. He takes factitive SE to be a reciprocal
predicate that is coindexed with the subject and with the (covert) agent that can however be
expressed obliquely with con 'with' or other prepositions (remember 249). This SE is
introduced as a second VP layer which dominates the one headed by the lexical verb, and
therefore can define a different type of subject, as in (252).

(252) VP

V VP
SEi+j
Juani V

V PP
afeita
P el barberoj
con

Being coindexed with both arguments, SE controls the presence of the oblique agent; later
on, the DP argument raises to the subject position and the lexical V incorporates to the V
materialised as SE, building a new predicate with different subject entailments. Beyond the
properties mentioned above, that show that there are no reasons to propose a distinct syntactic
configuration for factitive cases, one major problem of this approach is that it is unclear how

92
 SE IN SPANISH

SE ends up establishing the normal c-command relation with the PP and how the presence of
two VP layers does not produce an increase in the number of DP arguments that could be
projected. Let us now move to antipassive SE, where Masullo (1992) also made a major
contribution related to the same type of incorporation analysis.

6.2. Antipassive SE
In the cases that we have seen so far, the role that SE has in impoverishing the predicate that
it combines with can be seen as essentially disabling an argument position –with the exception
of aspectual or agentive SE, which does not alter the argument structure of the predicates–,
either by forcing coreferentiality of some internal argument with the external argument or by
removing an external argument position. In the case of anticausative SE, an alternative
explanation of why one argument disappears could be that the SE removes the capacity of the
verb to assign accusative case to the internal argument. This could have the effect of forcing
the internal argument to become the subject of the clause, to get case at that level, thus blocking
the presence of another argument that could act as subject.
The term antipassive SE, which originally can be traced back to Deguchi (1978) –see also
Masullo (1992), Bogard (1996-1997, 1999)– is reserved in some works to a SE that, apparently,
when combining with a predicate also blocks its ability to assign accusative case to the internal
argument but where, in contrast to anticausative SE, the solution adopted to solve the case
problem is not to remove the external argument, but to introduce the internal argument with a
prepositional structure that, presumably, is what assigns case to it. (253) illustrates one of such
pairs, where the antipassive SE is in (253b).

(253) a. Juan deshizo el nudo.

Juan undid the knot
b. Juan se deshizo del nudo.
Juan SE undid from-the knot
'Juan got rid of the knot'

Note two relevant properties: between (253a) and (253b) there are no differences with
respect to the number of arguments that the verb requires. The difference is in how the internal
argument is expressed syntactically: in (253a) it is a direct object defined by accusative case,
while in (253b) the same internal argument, which is compulsory for the interpretation of the
predicate, is introduced by a PP lexically selected by the predicate. The second relevant
property is that there is a non-systematic difference in meaning in the type of event denoted in
each one of the two cases, which in terms of the conceptual semantics refer to different types
of actions: to undo something and to get rid of something. While in broad terms one can
imagine more or less metaphorical extensions of the meaning of (253a) that would give us the
meaning of (253b), here we don't have the systematic difference in meaning obtained in
anticausatives (internally caused event or event causes without external participation),
aspectual cases (telic increase, result state, unergativity) or agentive SE cases (wilful intent).
(254) provides a longer list of predicates where the literature has identified the antipassive
value; they all behave like (253) in terms of the presence of a transitive SE-less version and a
SE-marked version where the internal argument is introduced by a PP.

(254) acordar-se 'to remember', arriesgarse 'to risk', burlarse 'to make fun of someone',
aprovecharse 'to take advantage', compadecerse 'to feel pity', despedirse 'to say bye',
evadirse 'to flee away from', lamentarse 'to regret'

93
 ANTONIO FÁBREGAS

The list is increased a bit if one includes, as Masullo (1992) does, other classes of verbs that
have a family resemblance with (254) but do not correspond to the same contrast as in (253):
(i) verbs that lack an internal argument without SE but that can include a PP argument with SE,
such as reír 'laugh' vs. reírse de alguien 'to laugh-SE at someone' and (ii) verbs which are
transitive without SE and have at least one use with SE in which they introduce an internal
argument with a PP, even if they allow a second use where the internal argument can be
accusative, such as olvidar 'to forget' vs. olvidar-se, which allows both Se la olvidó 'SE it forgot,
he forgot it' and Se olvidó de ella 'SE forgot of her, he forgot about her' (similarly, encontrar
'to find'). Even though in this second case we might have two different uses of SE –perhaps
aspectual like the one in dejarse 'leave' in the first case, and antipassive in the second–, we will
not include these verbs in our list in order to keep the description as clear as possible to
determine whether the class is really a grammatically relevant one. Similar considerations make
less straightfoward whether soltarse 'to let go of', which could be taken to be an anticausative
version of soltar algo de un lugar, 'to release something from a place' or an antipassive from
the otherwise transitive soltar un objeto 'to let an object go'.
In some of these cases, the meaning of the predicate does not vary much between the two
versions, such as in (255a,b) and (255c,d), while in other cases the meaning difference is
extremely sharp (255e,f).

(255) a. Juan soltó la cuerda.

Juan let.go the rope
b. Juan se soltó de la cuerda.
Juan SE let.go of the rope
c. Juan despidió a los invitados.
Juan said.bye DOM the guests
d. Juan se despidió de los invitados.
Juan SE said.bye from the guests
e. Juan acordó salir de casa.
Juan agreed exit from home
'Juan agreed to leave the house'
f. Juan se acordó de salir de casa.
Juan SE agreed from exit from home
'Juan remembered leaving the house'

In the cases where the meaning difference is not so sharp one can find the notion of origin
of movement that was relevant in the case of aspectual SE in §5.4 above, although note that in
these cases the difference is not in the presence of an origin of movement, but in the fact that
that origin is expressed with a PP: in (255d) one gets the feeling that the subject is the one that
would leave the place where the guests are present, but the separation between Juan and the
guests is also present in (255c); the same can be said about (255a): the separation is already
there without SE, but one gets the feeling in (255b) that it is Juan and not the rope the entity
that moves away from the other –the rope might stay in the same location, and Juan lets go of
it, moving away from it, while in (255a) Juan might stay in the same location and the rope can
fall down. Despite this interesting connection, not all predicates where antipassive SE does not
trigger a lexical change in meaning can be subsumed to a notion of separation (see 256) and it
cannot be said that separation defines antipassive SE in the less systematic cases either: at least
one case the separation component disappears with SE (257).

(256) a. Juan lamentó esa muerte.

Juan regretted that death

94
 SE IN SPANISH

b. Juan se lamentó de esa muerte.

Juan SE regretted of that death
'Juan complained about that death'

(257) a. Juan burló a la policía.

Juan evaded DOM the police
b. Juan se burló de la policía.
Juan SE evaded from the police
'Juan made fun of the police'

Bogard (1996-1997, 1999) has argued that this construction is properly antipassive in the
sense that, like in the languages where the antipassive is a productive grammatical operation
applied to transitive verbs, the internal argument becomes optional. His claim is based on
minimal pairs like (258).

(258) a. Juan aprovechó *(tu experiencia).

Juan profitted (your experience)
'Juan made the most out of your experience'
b. Juan se aprovechó (de tu experiencia).
Juan SE profitted from your experience
'Juan took advantage of your experience'

The argument has two sides, one of which is more controversial than the other. The first
claim is the observation that the SE-less pair of the antipassive SE verbs rejects the object-less
construction; that is, that without SE the direct object of the verbs in (254) must be syntactically
expressed. This claim is clearly correct in empirical terms:

(259) *Juan {burla / acuerda / resuelve / compadece / evade / suelta...}.

Juan evades / agrees / solves / pities / evades / lets go...

The second part of the generalisation is empirically less solid: that the PP constituent is not
a compulsory argument of the predicate. Undoubtedly, the PP constituent can be removed in
the sense that it might stay tacit or be inferred from context, as in (260), something that the
direct object could not do –although not with identical ease in all cases–.

(260) Juan se {burla / acuerda / ??resuelve / compadece / evade / ??suelta...}.

Juan SE makes fun / remembers / decides / pities / evades / lets go...

However, even in the cases where removing the argument is more natural, one requires a
generic reading or some specific previous mention of the entity; this, then, is a case equivalent
to not expressing overtly other internal arguments, as in (262), which is possible given the right
conditions, but where the predicate is still interpreted as requiring an internal participant.

(261) a. Juan se ha aprovechado.

Juan SE has taken.advantage
b. Juan se evade.
Juan SE evades
(262) a. Juan no ha comido.

95
 ANTONIO FÁBREGAS

Juan not has eaten

b. Juan lee.
Juan reads

Considering both sides of the test, and particularly that these verbs without SE do not admit
easily that the internal argument is tacit, one could speculate that these verbs are formed from
verbal stems that impose the requisite that their internal argument position must be occupied.
This opens the door for finding a direct connection between antipassive and anticausative SE
in terms of the removal of the capacity to assign accusative. The connection would go as
follows: in both cases, antipassive and anticausative, SE has the effect of removing the capacity
of the verb to assign accusative case. The verbs that end up being anticausative are verbs where
the internal argument, not receiving accusative case, can become the subject of the clause in
order to get its case checked; this prevents them from projecting a different external argument.
In contrast, the verbs that end up being antipassive are verbs which must have some material
within vP and therefore cannot project the internal argument as their subject: the only option
to overcome the case problem is to introduce the argument as a PP, in which case it can remain
as internal argument along the whole derivation. This intuitive approach, however, does not
explain properly how the requisite of keeping the internal argument within vP should be
implemented technically.
Alternatively, Deguchi (1982) proposes that the effect on the transitivity of the predicate is
actually triggered by the interpretation of the subject: according to him, the subject in the
antipassive SE cases is not a strongly agentive subject, something that forces its demotion as a
proper agent to a category closer (but not identical) to an object, without the verb becoming
unaccusative. This demotion reduces the transitivity of the predicate, which then introduces the
object as a PP. From this perspective, antipassive SE and anticausative SE would differ in that
antipassive SE keep an external argument that has been semantically weakened while the
anticausative SE remove the external argument entirely; antipassive verbs could be viewed as
verbs that must project an external argument, even if semantically bleached as a responsible
agent, and therefore where the reduction of case cannot be solved by turning the internal
argument into the subject.
In terms of the formal analysis of antipassive SE, the approach that is still predominant –
with minor tweaks– is the one adopted in Masullo (1992: 241). In contrast to his analysis of
factitive SE above, SE in the antipassive case is occupying the position of internal argument,
from which it incorporates into the V head and absorbs its case assigning possibilities without
saturating the theta-role, because by assumption SE is empty of features. Note that SE is
coindexed with the semantic internal argument –the structure in (263) tries to adapt Masullo
(1992), who does not specifiy the internal structure of the VP, to a more modern approach with
c-commanding relations in place–.

(263) vP

Juan v
ø
v VP

tu experienciai V

V sei
aprovechar

96
 SE IN SPANISH

From (268), SE incorporates to the V head, and as a result of it the DP direct object cannot
receive accusative case; a preposition is introduced –according to Masullo (1992)– at PF to
satisfy the case requirement. Note the crucial similarity between this analysis and Armstrong's
(2013) for the transitive SE uses (revised in §5.3). Like in Armstrong (2013), the incorporation
defines a new predicate that might be associated to a different lexical meaning than the one
related to the SE-less version; like in that analysis, here the presence of SE as an argument
occupies a syntactic position that forces the redefinition of the place occupied by the internal
argument. The difference is that the transitivising SE cases do not involve absorption of the
case assigning options of the verb, while this one does it.
Thus, the antipassive SE uses seem to establish connections with anticausative cases, via
the possible absorption of case, and with the transitivising SE cases able to redefine the lexical
semantics of the verb. Additionally, and like in the intransitive so-called aspectual SE cases,
the preposition in most of the antipassive cases is the separation preposition de 'from, of',
although the separation meaning is not always present in the predicates –note, also, that this
preposition would anyways be the underspecified preposition in Spanish–.
Let us now move to the inherent SE predicates, the last class within the group of lexically
defined SE values.

6.3. Inherent SE verbs

There is a number of verbs which in contemporary Spanish are only used with a SE version,
and which receive different names in the literature: inherently reflexive verbs, pure pronominal
verbs, inherently pronominal verbs or properly reflexive verbs (Martínez Martín 1989, Gómez
Torrego 1992, Otero 1999, Sánchez López 2002: 96 and folls.; cf. RAE & ASALE 2009:
§41.13k for an even longer list). As in the other lexical SE cases, keep in mind that the list
might be subject to dialectal or even individual variation.

(264) abalanzarse 'to run towards', aborregarse 'to become tame', abstenerse 'to abstain',
acurrucarse 'to get cozy with someone', adentrarse 'to get inside', adormilarse 'to get sleepy',
adueñarse 'to come to possess', agolparse 'to come to form a group', antojarse 'to fancy',
arrepentirse 'to repent', arrogarse 'to reclaim', atenerse 'to follow', atreverse 'to dare',
condolerse 'to show condolences', contonearse 'to wiggle', desentenderse 'to ignore',
desgañitarse 'to grow hoarse', desternillarse 'to laugh a lot', despelotarse 'to get naked',
dignarse 'to accept', empecinarse 'to insist', enamoriscarse 'to get infatuated', enfrascarse 'to
get absorbed into an intellectual matter', enfurruñarse 'to get grumpy', ensañarse 'to treat
brutally', esforzarse 'to put effort', fugarse 'to flee', guasearse 'to tease', inmiscuirse 'to
intervene', jactarse 'to boast', mofarse 'to laugh at', obstinarse 'to persist', pavonearse 'to
boast', pitorrearse 'to mock', portarse 'to behave', regodearse 'to gloat', suicidarse 'to kill
oneself', ufanarse 'to boast', vanagloriarse 'to boast'

The problem posed by these verbs is that, lacking a correlate without SE, it is difficult to
determine the semantic or syntactic contribution of the clitic in these cases. The strategy, then,
can be the following: to treat these verbs with inherent SE as instances of the other, more
established SE uses, with the only additional property that in their case the addition of SE is
compulsory, that is, that these verbs must necessarily be for one reason or the other included
in the broader class that is necessarily marked with SE: for instance, some of these verbs could
be anticausative verbs that have lost their causative pair and currently are only licensed
lexically or syntactically under the configuration that requires insertion of SE.
At this point it is appropriate to make a small methodological remark. The notion of
inherently reflexive SE predicate only makes sense under two questionable assumptions: (i)
that SE is basically a reflexive marker, something that we have already repeatedly shown to be

97
 ANTONIO FÁBREGAS

inexact, and (ii) that whether SE is compulsory or optional is a valid criterion from a
grammatical perspective. Under the assumption in (ii), reflexive SE is optional and its presence
or absence is dependent on the referential relations within the argument structure of the verb;
anticausative SE is less easy to classify as optional or compulsory because, given an
anticausative syntax and semantics, some verbs can have SE or not (remember §4.1 above)
while others forcefully have it; aspectual or agentive SE is optional in the same sense, but again
remember §5.3 above, where we saw that the presence of certain syntactic elements, such as
origin phrases, force their presence; finally, inherently reflexive SE would be the only proper
instance of compulsory SE. The problem becomes more accute when we include in the
discussion predicates that require SE to introduce certain types of direct object or meanings
that the SE-less version cannot express –for instance, some of Armstrong's (2013) transitive
SE cases–. Should we take this meaning difference as crucial and then say that SE is
compulsory in this predicates, as they cannot have this meaning without it, or should we
consider only the lexical verb without regard to its conceptual meaning and say that these verbs
have optional SE? Here we are already close to the area of lexicology rather than to grammar.

(265) a. saltarse un semáforo

jump-SE a traffic light 'not to stop at a traffic light'
b. ganarse la vida
win-SE the life 'to make a living'
c. perderse algo
lose-SE something 'to miss some event'
d. jugar-se algo
play-SE something 'to bet something'

Given how fuzzily defined the optionality is in the case of the allegedly non compulsory SE
cases, it is clear that a distinction between optional and compulsory SE will not take us too far
in analysing the properties of SE.
Thus let us rather examine the verbs from the perspective of whether the necessary SE can
be assimilated to other better described types of SE. As expected from a group that has been
established in traditional grammars using a questionable criterion of obligatory presence of SE,
there are very different verbs in the list, so the types of SE involved here are equally varied.
Once classified in the appropriate SE class we might ask ourselves what makes them have to
belong particularly to that class, without an alternating version without SE.
Let us start with the observation that the list in (264) does not include a homogeneous class
of predicates if we attend to the properties of the subject, as noted in Sánchez López (2002),
who presents the list from the perspective of whether these verbs can be considered
anticausative. The verbs in (266) act as unaccusatives in the sense that the subject seems to act
as a patient: the verbs allow the absolute participle construction in combination with the subject
(exemplified with a few cases in 267).

(266) aborregarse 'to become tame', acurrucarse 'to get cozy with someone, to curl up',
adentrarse 'to get inside', adormilarse 'to get sleepy', agolparse 'to come to form a group',
arrepentirse 'to repent', desentenderse 'to ignore', desgañitarse 'to grow hoarse', despelotarse
'to get naked', enamoriscarse 'to get infatuated', enfrascarse 'to get absorbed into an
intellectual matter', enfurruñarse 'to get grumpy'

(267) a. Acurrucados los enamorados...

curled.up the lovers...
'Once the lovers curled up with each other...'

98
 SE IN SPANISH

b. Adentrados los niños en el bosque...

got.inside the children in the forest...
'Once the children got inside the forest...'
c. Enfrascado Pedro en la lectura...
got.absorbed Pedro in the reading
'Once Pedro got into his reading...'
d. Desentendido Juan de sus obligaciones...
ignoring Juan of his obligations...
'Once Juan had ignored his duties...'
e. Arrepentido Juan de su decisión...
repented Juan of his decision...
'Once Juan repented of his decision...'

Other unaccusativity tests give a worse result; as it is the case with anticausatives in Spanish
(§4.1, see also §8.1 later), generally they do not allow postverbal subjects projected as bare
nominals (268); in the list there is only one case (268a) where bare nominal subjects are
possible.

(268) a. Se agolpó gente en el parque.

SE formed.group people in the park
'People gathered in the park'
b. ??Se despelota gente en esta playa.
SE get.naked people in this beach
c. *Se desgañitaron profesores durante la representación.
SE got.hoarse teachers during the representation
d. *Se enamoriscaron adolescentes durante las fiestas.
SE fell.in.love teenagers during the festivities
e. *Se arrepintieron acusados.
SE repented accused-people
Intended: 'The accused repented'
f. *Se desentienden padres.
SE ignore parents
Intended: 'Parents tend to ignore [their children]'

This is good news from the perspective or reanalysing inherent SE as other types of SE
structures: the behaviour is what one expects of anticausatives. Moreover, their semanticsis
also typically the one of anticausatives: change of state or change of location, where the subject
is the entity that undergoes that change. However, in some cases one identifies that the subject
might be potentially an experiencer –and hence not an internal argument–, as in the psych
predicates: arrepentirse, enamoriscarse, enfrascarse, desentenderse, etc., which in fact are
predominant in the group. If one accepts that perhaps experiencers are projected in external
argument positions (or move from an internal argument position to an external argument
position, losing the capacity to be bare nominals; see §8.1 below), perhaps these verbs can be
treated as anticausatives that lack a causative pair. In some cases, that causative version might
have existed in the past, or might be marginally active for some speakers, as it is the case of
the author of these lines with aborregar 'to make someone tame'. We will, however, come back
to this potential characterisation of a subgroup of inherent SE verbs as anticausative verbs
without a causative member.
In contrast with this, the verbs in (269) act as unergative verbs, with what seems to be an
agentive subject (Otero 1999). They systematically reject the absolute participle structure

99
 ANTONIO FÁBREGAS

together with the subject and their subject tends to be interpreted as an agent that controls a
process, with the few exceptions of antojarse 'to fancy', condolerse 'to show condolences',
where the external argument seems again to be an experiencer.

(269) abalanzarse 'to run towards', abstenerse 'to abstain', adueñarse 'to come to possess',
antojarse 'to fancy', arrogarse 'to reclaim for oneself', atenerse 'to follow', atreverse 'to dare',
condolerse 'to show condolences', contonearse 'to wiggle', dignarse 'to accept', empecinarse
'to insist', ensañarse 'to treat brutally', esforzarse 'to put effort', fugarse 'to flee', guasearse 'to
tease', inmiscuirse 'to intervene', jactarse 'to boast', mofarse 'to laugh at', obstinarse 'to
persist', pavonearse 'to boast', pitorrearse 'to mock', portarse 'to behave', regodearse 'to gloat',
suicidarse 'to kill oneself', ufanarse 'to boast', vanagloriarse 'to boast'

The semantic categories expressed by these verbs combine with the impossibility of getting
projected as absolute participle structures and make any analysis in terms of anticausativity
impossible: these verbs do not express inherently directional movement unless they combine
with a clear manner component expressing the speed or the intention of the subject when
moving (cf. abalanzarse), and they do not express changes of state unless combined with a
clear wilful intent of the subject (cf. adueñarse). More frequently, they express actions, not
changes of state, and different ways of behaving with respect to others, all prototypical
unergative verb classes. One identifies, even, some conceptual semantic tendencies, such as
the tendency to involve verbs of mocking someone (cachondearse, pitorrearse, mofarse,
burlarse...). So what are the options to treat these verbs as instances of other, independently
motivated classes of SE, when the anticausative analysis cannot be applied?
Let us start with reflexive or reciprocal SE. Among inherent SE verbs, there is at least one
obvious candidate to fall in this class as a predicate that by its internal meaning has to be
reflexive and where reflexive SE might be a clear option: suicidarse. The absence of a SE-less
pair here would be due only to the plausible fact that SE here marks reflexivity and the predicate
cannot be performed on an entity distinct from the subject. Remember in this respect that
reflexive verbs do not seem to act as unaccusatives in Spanish (§3.1): the absolute participle
structure, even with naturally reflexive predicates, tends to be interpreted with an external
subject –that is, the nominal expression is interpreted rather as a direct object than as a subject
in the clausal version–.

(270) a. Visto el bebé, apagamos la luz.

seen the baby, turned.off the light
'Once someone saw the baby, we turned off the light' (not *'Once the baby saw himself,
we turned off the light')
b. Peinado Juan, salió de casa.
combed Juan, exited from home
'Once someone combed Juan's hair, he left his house' (not ??'Once Juan combed his hair,
he left his house')

The ungrammaticality of (271) from this perspective would come from the absence of a
direct object interpretation of the verb, just as in the other cases in the list of (269) –of course,
(271) is grammatical for a speaker that has a transitive version of the verb meaning something
like 'to kill someone making it look like a suicide'–.

(271) ??Suicidado Juan, lo enterramos.

suicided Juan, him buried.we
Intended: 'Once Juan had committed suicide, we buried him'

100
 SE IN SPANISH

Other potential cases of naturally reflexive predicates in the list might include vanagloriarse
'to brag about one's own accomplishments' –and its perhaps less common synonyms jactarse
and ufanarse–. The interesting property of these three verbs is that one cannot (semantically or
syntactically) boast about something done by someone else:

(272) a. Juan se vanagloria de que ha conocido a Harry Potter.

Juan SE boasts of that has met DOM Harry Potter
'Juan boasts that he (himself) has met Harry Potter'.
b. #Juan se vanagloria de que María ha conocido a Harry Potter.
Juan SE boasts of that María has met DOM Harry Potter
Intended: 'Juan is very proud that María has met Harry Potter'

From this perspective, arrogarse 'to reclaim something for oneself' is also a good candidate
to be an instance of reflexive SE in a predicate where the meaning is naturally reflexive and
therefore there is no double.

(273) Juan se arrogó esos derechos.

Juan SE arrogated these rights
'Juan arrogated these rights to himself'

We might add to this list of potentially naturally reflexive SE verbs where there is no SE-
less version the case of quejar-se de algo 'to complain-SE of something', where there is a strong
implication that the thing that the subject complains about affects the subject personally (as
oppose to lamentar 'regret' or protestar 'protest', where one can complain about things affecting
others).
In a less obvious way than with suicidarse, then maybe these five predicates are instances
of reflexive SE marking verbs whose lexical meaning forces them to have a reflexive meaning.
Another clear candidate to reduce the inherent SE predicates in above is the antipassive SE
that keeps an internal argument but marks it prepositionally. Hernández Sacristán (1986) notes
that a subgroup of these verbs can be argued to be derived through prefixation (sometimes,
with parasynthesis) from bases that might be transitive verbs without the prefix.

(274) a. Juan se abstiene de eso.

Juan SE abstrains from that
b. Juan tiene eso.
Juan has that
(275) a. Juan se atiene a eso.
Juan SE abides to that
'Juan abides to that'
b. Juan tiene eso.
Juan has that

These are the two clearest cases where one could argue that the presence of the prefix creates
a verb that can only be intransitive but requires an internal argument that must be marked by a
preposition. Noting the relation between doler 'to hurt', dolerse de algo 'to feel pain about
something', and condolerse con alguien 'to feel sympathy for someone', the last involving a
prefix con- which correlates with the preposition used to introduce the internal argument
(condolerse con alguien, 'to feel sympathy with someone'), one could assimilate this verb
partially to the antipassive SE class, even if the base verb never takes an accusative object.

101
 ANTONIO FÁBREGAS

However, Hernández Sacristán (1986) notes that many of these inherent SE predicates are
parasynthetic, among them the following (276). Perhaps the parasynthetic structure creates in
these cases bases that are compulsorily combined with an internal argument, where the SE
forces that internal argument to receive prepositional marking.

(276) a. Juan se adueña de eso.

Juan SE appropriates of that
b. Juan se ensaña con alguien.
Juan SE treats.brutally with someone

Several of the possibly unaccusative verbs above are also parasynthetic, insisting in the
relation between inherent SE and parasynthesis: aborregarse 'to become tame', acurrucarse 'to
get cozy with someone, to curl up', adormilarse 'to get sleepy', agolparse 'to come to form a
group', desentenderse 'to ignore', desgañitarse 'to grow hoarse', despelotarse 'to get naked',
enamoriscarse 'to get infatuated' and enfrascarse 'to get absorbed into an intellectual matter',
enfurruñarse 'to get grumpy'.
That said, the two verbs in (276) have something else in common between them and with
other verbs in the group in (269), namely that they involve necessarily agentive events where
the subject must be an agent with wilful intent, along the lines of what agentive SE predicts.
The members of (269) that are the strongest candidates to design eventualities that must be
agentive and where the SE might be marking the restriction that these verbs impose on their
subjects, as agents, are in (277):

(277) abalanzarse 'to run towards', atreverse 'to dare', contonearse 'to wiggle', dignarse 'to
accept', empecinarse 'to insist', ensañarse 'to treat brutally', esforzarse 'to put effort',
guasearse 'to tease', inmiscuirse 'to intervene', mofarse 'to laugh at', obstinarse 'to persist',
pavonearse 'to boast', pitorrearse 'to mock', regodearse 'to gloat'

What these verbs have in common is that they express events that must be consciously and
purposefully controlled by the subject; in same cases the event does not consist in more than
expressing a conscious and insistent will to carry out the event (esforzarse, obstinarse,
atreverse) or to agree to do it (dignarse). Relating them to agentive SE does not sound crazy
from that perspective.
However, as we have noted repeatedly in §5 above, the agentive properties of the subject
and the aspectual properties of the predicate come related to each other in many cases, making
the division between aspectual and agentive SE (for those that admit both) or the reduction of
one class to the other (for those that only want to accept one category) more controversial. Not
surprisingly, from this perspective, there are cases in (269) which one could argue are agentive
or aspectual. The main case is fugarse 'to flee from somewhere', which in European Spanish
must carry SE. García Fernández (2011) relates this verb to the aspectual interpretation where
the origin of movement is expressed, noting also that surprisingly SE is compulsory with this
verb and impossible with its virtual synonym huir even if both require an origin complement.

(278) a. Juan *(se) fugó de la cárcel.

Juan SE fleed from the jail
b. Juan (*se) huyó de la cárcel.
Juan SE fleed from the jail

102
 SE IN SPANISH

One could argue that the verbal stem in (278a) is part of an idiom formed together with SE,
where the argument structure related to the verb is directly conditioned by the presence of SE
and the unergative construal, while in (278b) the verbal stem alone –not part of an idiom–
already defines the argument structure that the idiom fugar+se defines. However, the
connection with agentive SE can be noted in that (278b) allows non agentive constructions,
while (278a) rejects non agentive subjects.

(279) a. La vida {huía / *se fugaba} de dentro de él.

the life fleed / SE fleed from inside of him
'Life escaped from inside him'
b. El tiempo {huye / *se fuga}.
the time flees / SE flees

Thus, the global conclusion with respect to inherent SE verbs could be that the inherent SE
cases have in common that they are instances of verbs that form idioms with SE –in the sense
that the structure introduced by SE is lexically listed as the only structure that licenses the
verbal stems, in contrast to other verbs which allow the SE structure and its absence, with
changes in meaning–. Even if they can be considered idioms involving SE, the specific SE
structure that is part of the idiom for each verb varies depending on the subgroup of verbs:
some are anticausative idioms, some are naturally reflexive idioms, some are antipassive
idioms, etc.
A strong argument in favour of this reinterpretation of inherent SE verbs as verbs which are
part of idioms containing SE comes from the examples of inherent SE verbs which share their
verbal stem with other verbs, but where the meaning contrast is so strong that for some speakers
the connection is lost. Next to other cases where SE has a strong change in meaning and that
we mentioned at the beginning of this section, the clearest case of this is portarse 'behave', an
agentive verb that rejects non-agentive subjects (280) and which shares the verbal stem with a
SE-less predicate that, although less frequent, allows non-agentive subjects without wilful
intent (281).

(280) {Juan / el perro / *el examen} se portó bien.

Juan the dog the exam SE behaved well
(281) {Juan / este estuche / el tren} porta ese objeto.
Juan this case / the train carries this object

The difficulty of how to categorise the SE that is found in each verb reflects in many cases
the impossibility of finding a SE-less pair to compare the verb with, in order to determine which
properties of SE are affected in each case. The next list summarises some of the clearest cases,
with some potential options for other inherent SE verbs that might increase the list:

(282) Potential classification of the verbs that form idioms with SE

a. Anticausative idioms: aborregarse 'to become tame', acurrucarse 'to get cozy with
someone', adentrarse 'to get inside', adormilarse 'to get sleepy', agolparse 'to come to form
a group', arrepentirse 'to repent', desentenderse 'to ignore', desgañitarse 'to grow hoarse',
despelotarse 'to get naked', enamoriscarse 'to get infatuated', enfrascarse 'to get absorbed
into an intellectual matter', enfurruñarse 'to get grumpy'
b. Naturally reflexive idioms: suicidarse, jactarse, vanagloriarse, ufanarse, quejarse;
perhaps antojarse ('to perceive for onself'?), inmiscuirse ('to involve onself in something'?)
c. Antipassive idioms: abstenerse, atenerse; perhaps condolerse, adueñarse, ensañarse...

103
 ANTONIO FÁBREGAS

d. Agentive idioms: abalanzarse 'to run towards', atreverse 'to dare', contonearse 'to
wiggle', dignarse 'to accept', empecinarse 'to insist', ensañarse 'to treat brutally', esforzarse
'to put effort', guasearse 'to tease', inmiscuirse 'to intervene', mofarse 'to laugh at', obstinarse
'to persist', pavonearse 'to boast', pitorrearse 'to mock', portarse 'to behave', regodearse 'to
gloat', and maybe also adueñarse and ensañarse.
e. Aspectual idioms: fugarse, and perhaps jalarse, zamparse 'to eat up', which are almost
exclusively used with SE by speakers (García Fernández 2011).

6.4. Figure SE: reflexive SE and locative structures

Next to these cases, there is a final type of SE documented in the literature that might
connect with agentive uses (§5.3 above), but whose identification in Spanish is less clear: the
so-called figure reflexive SE. This type of SE –interpreted in the broad sense of a reflexive– is
identified in Icelandic by Wood (2014, 2015), and refers to constructions such as the ones in
(283) and (284), where the SE does not involve an agent-removing function: the subject is still
interpreted as a volitional agent, but additionally it is interpreted as an entity that ends up in the
location expressed by the prepositional structure.

(283) a. Han ruddi henni úr rödinni.

he cleared her out line
'He knocked her out of the line'
b. Han ruddi-st úr rödinni.
he cleared-SE out line
'He elbowed his way out of the line'
(284) a. Jon bödladi boltanum yfir línuna.
Jon bumbled ball over line
b. Ég bödladi-st yfir snjóskaflinn.
I bumbled-SE over snow.bank
'I struggled my way out of the snow bank'

These predicates with SE are related to a sense of effort and intended movement to the result
location. Like the aspectual SE cases, they have an immediate effect on the aspectual
information of the predicate, forcing a result location where the subject will stay for some time,
and are related to agent entailments involving wilful intent. Wood's analysis treats the reflexive
as an argument which appears, coindexed with the agent, in the prepositional position of figure
of movement, thus trearting the SE morpheme as a reflexive which occupies an argument
position which is added to the predicate by the locative structure.

(285) VoiceP

Agent Voice

Voice vP

v PP
clear
SE P

P out of the line

104
 SE IN SPANISH

It is not difficult to find some clear examples of this potential type of structure in Spanish,
keeping the relevant properties. In fact, some of the examples in Jiménez-Fernández & Tubino
(2014, 2019) might be reinterpreted as instances of this type of structure, with the only tweak
that the predicates in (219) do not get their agentivity entailments affected by the presence of
SE, which is not what happens in the Jiménez-Fernández & Tubino (2014, 2019) cases –
remember that in her analysis SE is part of the predicate, in a head position which can affect
the type of event–. All the examples in (286) are agentive, as the verb without SE was also
agentive.

(286) a. Juan se metió en la casa de María.

Juan SE entered in the house of María
b. Juan se pasó por la casa de María.
Juan se passed by the house of María
c. Juan se infiltró en la mafia.
Juan SE infiltrated in the mafia

In (287) one can even identify a contrast between being necessarily agentive and not
requiring the help of others to escape (287a) and maybe escaping not noticing that he does so
or helped by others in (287b), which again relate the necessarily agentive interpretation in
figure SE cases with the presence of the movement predicate.

(287) a. Juan se escapó de la cárcel.

Juan SE escaped from jail
b. Juan escapó de la cárcel.
Juan escaped from jail

However, not all movement verbs with SE have this type of agentivity entailment related to
the presence of SE; for instance, (288) lacks it.

(288) Juan se cayó en el pozo.

Juan SE fell in the well

This, in essence, means that figure SE cannot be used as the device to cover all types of
aspectual SE cases with intransitive verbs.
In order to be exhaustive, we should also note that the logic that makes possible the presence
of a reflexive figure SE would in principle have to imply that there should be a reflexive SE in
the position of ground of movement, that is, as the location where the subject finishes.
However, such 'ground SE reflexive' is, to the best of my knowledge, very restricted in Spanish,
perhaps completely unattested.
Some significant candidates for such 'ground SE' structure would be the one in (289b), if
one compares it with (288a)

(289) a. Juan puso el abrigo en la mesa.

Juan put the coat on the table
b. Juan se puso el abrigo.
Juan SE put the coat
'Juan put the coat on (himself)'

105
 ANTONIO FÁBREGAS

Other verbs that might fall in the same class are in (290), where their main common property
is that they express locations of clothing elements that someone generally puts on or removes
from oneself.

(290) a. Juan se colocó el sombrero.

Juan SE placed the hat
b. Juan se ciñó el cinturón.
Juan SE fastened the belt
c. Juan se quitó los zapatos.
Juan SE removed the shoes

As we can see, one way to interpret these verbs is as involving a subject that locates
something on himself, where the SE then would be a reflexive expressing that the result
location, the ground, is reflexive with the subject. However, there is a second interpretation:
these cases involve dative reflexive cases of the ditransitive location verbs such as (291).

(291) a. Juan le puso el abrigo al niño.

Juan him put the coat to.the boy
b. Juan le colocó el sombrero a su abuelo.
Juan him placed the hat to his grandfather
c. Juan le quitó los zapatos al paciente.
Juan her removed the shoes to.the patient

This interpretation seems simpler than the one involving a reflexive ground if only for one
reason: the ground in the locative structure should be the complement of a locative prepositon,
and would then be marked with a case different from accusative or dative, which are the two
cases that we have seen associate to SE in all the other cases reviewed so far. It seems, then,
that there are no instances of verbs where the ground reflexive SE is the best conceivable
analysis.

6.5. Wrap up: what is lexical about SE

In these first four sections (§3-§6) we have revised a number of uses of SE that have in
common that they are predicate affecting, and paradigmatic in the sense that they co-vary with
another argument of the clause –almost always, the subject– in person and number. Perhaps
the general conclusion, beyond the discussions about which value of SE is basic in these cases
and how specific examples are classified as one or the other, is that these uses of SE are lexical
in several relevant senses.

a) The first, obvious sense in which SE is lexical is that the presence or absence of SE affects
one or more of the following crucial properties of the predicate as a lexical item: its argument
structure, its internal aspect or Aktionsart and its lexical meaning. The following list details the
surface changes that are related to each one of the uses of SE revised here, including the most
controversial ones (eg., factitive one) but not the inherent SE cases, where SE and the verb
form an idiom (see 309 above).

a. Reflexive or reciprocal SE: affects the coreference relation between arguments

b. Anticausative SE: affects the argument structure, presumably removing one external
argument and potentially affecting the aspect of the predicate
c. Aspectual SE: affects the lexical aspect of the predicate and perhaps also the argument
structure of the predicate (only internal arguments affected)

106
 SE IN SPANISH

d. Agentive SE: restricts the theta role of the external argument

e. Factitive SE (if it exists): restricts the theta role of the external argument
f. Figure SE: affects the coreference relation between arguments
g. Antipassive SE: affects the case assignment capacities of the verb and potentially the
lexical meaning of the predicate
h. Transitive SE: affects the argument structure of the predicate and the lexical meaning
of the predicate

b) The second sense in which predicate-affecting or paradigmatic SE is lexical is with

respect to the optionality in marking. We have documented in particular that with
anticausatives one needs to make three groups of SE verbs, those that are marked with SE,
those that are SE-less in the anticausative member and those where the anticausative version
can be marked with SE or not. The verbs that belong to each specific group vary depending on
the geographical variety and sometimes on individual differences among speakers. To some
extent, one can make similar considerations with respect to aspectual SE and inherent SE verbs.
c) The third sense in which this family of predicate-affecting uses of SE is lexical refers to
the partial impredectibility of which verbs carry SE marking in this sense. Reflexives are the
exception here: even if not any reflexive or reciprocal relation is marked by SE, there is a
syntactic generalisation that can be made for them, namely that direct objects or indirect objects
must be coreferential to the subject –it is more lexical, because it is less systematically related
to a strictly defined notion, which verbs are naturally reflexive, on the other hand–. This is
unlike other SE uses: principles related to the prototypicality of the event as caused by external
forces or not, or principles related to aspect, have been invoked to explain why some
anticausatives take SE; leaving aside the more systematic case of transitive accomplishments
with incremental internal arguments, the predicates marked with aspectual SE are more
variable (cf. for instance the puzzling case of fugarse vs. huir); which verbs take agentive SE
are not determined solely by which verbs require agents –for instance, asesinar 'murder'
requires an agent but it does not allow agentive SE; not all verbs allow factitive SE, and the
range of verbs that allow antipassive SE is equally difficult to group in one single natural class.
d) In more cases than the overview above might suggest, we find cases of verb pairs with
and without SE where the syntactic and semantic contribution that SE makes seems to be non-
systematic and therefore difficult to classify in any of the uses above.

Among the cases of this asystematic lexical use of SE, we can highlight a few examples –
the reader can herself check that, flapping through a common dictionary, it won't take long
until she identifies her own examples of less systematic SE, as I did; it is a reasonably fun
game–. For instance, take the verb aparecer 'to appear'. With SE, it seems that only subjects
denoting animate entities that can decide to appear to someone are licensed; keys cannot appear
with SE.

(292) {Las llaves / Los fantasmas} aparecieron.

the keys / the ghosts appeared
(293) {*Las llaves / Los fantasmas} se aparecieron.
the keys / the ghosts SE appeared

We have already seen above the case of saltar 'jump' vs. saltarse 'to ignore', where the
meaning of the verb with and without SE has very little in common, but can still be somehow
related (to jump over something is a way of not being stopped by it). Other cases are worse to
the extent that the connection between the meanings is impossible to identify. The verb in (294)
and (295), which with SE means 'to be ashamed' in Old Spanish or to have an orgasm in

107
 ANTONIO FÁBREGAS

contemporary European Spanish, is a good example; it is very difficult to find a semantic

connection between the SE-less and the SE-marked versions of the verb that justifies this.

(294) Juan corrió una maratón.

Juan ran a marathon
(295) Juan se corrió.
Juan SE ran
'Juan had an orgasm'

We have also cases where the meaning seems to get specialised, or where there are nuances
in the meaning. In (296), as opposed to (297), it seems that what one forgets is a physical object,
that must be left in a location, and not its content –(297) allows both readings–.

(296) Se olvidó el libro.

SE forgot the book
'He left the book behind'
(297) Olvidó el libro.
forgot the book

In (298), one must have missed some show or something worth watching or attending, while
in (299) one must lose something physical, such as the physical copy of a movie in a DVD, or
the rolls with celluloid.

(298) Se perdió la película.

SE lost the movie
(299) Perdió la película.
lost the movie

Some contrasts are even more difficult to predict or to relate with established uses of SE. In
(300), one emphasises some notion of surprise or unexpected discovery that (301) does not
necessarily have, although one cannot claim to find things in a planned way.

(300) Juan se encontró una araña en el baño.

Juan SE found a spider in the bath
(301) Juan encontró una araña en el baño.
Juan found a spider in the bath

In (302), as opposed to (303), one seems to be emphasising some additional emotional

implication in the event that might favour readings where the invitation was unexpected,
unusually generous or other really difficult notions to pin down.

(302) Juan se invitó a unas copas.

Juan SE invited to some drinks
(303) Juan invitó a unas copas.
Juan invited to some drinks

In (304), one needs to interpret that the subject changes his clothes, while in (305) it must
be the personality, or the physical or mental traits that change.

(304) Juan se cambió.

108
 SE IN SPANISH

Juan SE changed
(305) Juan cambió.
Juan changed

In (305), one interprets that what the subject has invented is something that does not
correspond to the truth –the person has made up something–, while in (306) we can be talking
about a real invention.

(305) Juan se inventó eso.

Juan SE invented that
'Juan made that up'
(306) Juan inventó eso.
Juan invented that
'Juan invented that'

We cannot end this overview without discussing briefly the use of SE in (307), that has
acquired some popularity in recent times among young speakers of European Spanish. In
contrast to the use of SE described in §5.6, as can be seen, this SE does not force a source
phrase (which the verb would anyways license) or force the presence of determiners with the
subject, suggesting that this SE keeps the predicate as an unaccusative.

(307) Se vienen cosas.

SE come.3pl things

Speakers consulted who have this construction note that the meaning of (307) is, with
respect to the SE-less pair, an emphasis on the imminency of the arrival of those things. The
same speakers note that animate subjects are ungrammatical in this case.

(308) a. *Se vienen niños.

SE come.3pl children
Intended: 'Children are arriving imminently'
b. Se vienen quejas.
SE come.3pl complaints
'Complaints are about to arrive'

Similarly, the same speakers note that the imminency meaning associated to this
construction is not compatible with a gradual modifier; the arrival is imminent and just about
to happen, not something that happens little by little as the entity moves across the path.

(309) *Se vienen cosas poco a poco.

SE come.3pl things little by little

The reader can easily check that these properties do not fit with any of the SE classes
reviewed so far; at most, they might relate (only in part) with passive SE (§8), where the
argument can be a bare noun and there are restrictions on animacy that, however, do not arrive
to the point that the ban extends to bare nominals. As the other cases just reviewed, this is an
instance of a less systematic use of SE whose study should be deepened.
Importantly, this collection of facts (from a to d above) has led some authors –most
significantly De Benito Moreno (2015, 2021)– to the proposal that the paradigmatic uses of SE
should all be taken to express lexical facts: depending on variety and individual preferences,

109
 ANTONIO FÁBREGAS

speakers might display a strongest tendency to mark with SE predicates when the agent is
demoted, or when the event is completed, but this is not a systematic rule that reflects a real
syntactic generalisation. This approach is superficially supported also by the observation that
in many cases the meaning of the verb is affected substantially by the presence of SE, and there
are plenty of cases where the presence of SE seems arbitrary (inherent SE cases, or the
anticausative verbs that belong to each group).
However, despite the facts above, the conclusion that predicate-affecting SE is lexical is not
granted. An alternative, that we favour in this article, is that the predicate-affecting uses of SE
are instances where SE is introduced at a low position in the tree, within what one could call
'the syntactic space of the lexical verb'. Simplifying the analyses that we have revised in the
last four sections, we can find two common tendencies:

a) SE is a feature-deficient pronoun that is introduced in an argument position within VP,

vP –as well as other labels for the internal structure of the lexical verb, such as Proc, Res, Init,
Asp, etc.–, Voice, PP or some alternative argument structure; being an expletive, it does not
saturate the argument position but it does force a reconfiguration of the syntactic space, and it
might be incorporated to the verb or cancel the case-assigning possibilities.
b) SE is the manifestation of an impoverished verbal head –vP, Voice, Init, Asp, etc.– that
typically contrasts with the version not marked with SE in the number of arguments licensed,
the cancelation or modification of a subevent, etc. From this perspective, the agentive reading
might be the effect of an impoverished head where only the conceptually prototypical subject
interpretation is available, agent.

In other words: these values of SE locate the element that SE represents in the area of the
verb, not within the clausal projections corresponding to grammatical aspect, mood, tense, etc.
If SE in these configurations shares the same syntactic space with the lexical verb, the above
effects can be understood, if not fully explained: the presence of SE in the structures that
introduce verbal arguments can alter the argument structure of the predicates, and even the
case-assigning capabilities of the lexical verb for internal arguments; if these structures also
define Aktionsart, the presence of SE will be sensitive to the aspectual properties; if the domain
of meaning of the lexical verb is under VoiceP or the equivalent head in other frameworks,
presence of SE within that domain might also affect the lexical meaning of the verb, and create
idioms, as in the case of inherent SE. The spell out of this element as SE or as zero might also
be influenced by selectional idiosyncratic properties of the exponents and morphemes used to
spell out the lexical verb, because SE stays in the same syntactic domain with the lexical verb
material, and the interpretation of SE with each lexical verb might also be idiosyncratically
determined by the lexical verb material, as it can be the case with factitive or agentive cases.
In other words, what is 'lexical' about SE can be explained if SE is introduced in the area
below the curve, the domain of the lexical verb.

110
 SE IN SPANISH

(310) CP
CLAUSAL DOMAIN
C TP

T MoodP

Mood AspP

Asp VoiceP
LEXICAL VERB DOMAIN
Voice vP

v VP

V PP

P ...

As we will see in §7-§10, some of the clause-affecting uses of SE have been analysed as
instances of SE where the relevant element is located above VoiceP –particularly in the case
of passive and spurious– or at least controlled by elements that are located outside the
constituent defined by VoiceP –for instance, in the analysis of impersonal SE–, something that
can be correlated with the absence of predicate-effects or lexical effects of SE. However, this
does not mean at all that SE in clause-affecting uses is not located in some analyses in the same
head as some of the predicate-affecting uses, in particular in the case of VoiceP, which is
extensively used to analyse passive SE structures and to some extend impersonal SE structures.
We are now ready, then, to move to the four uses of SE that are clause-affecting: the
impersonal, the passive, the middle and the SE that substitutes the dative clitic. But before that,
given the complexity of the issues related to predicate-affecting SE, let us provide a short
interim summary.

6.6. Interim summary

The complexity of the uses of SE that have some lexical effect on the predicate is such that
it is appropriate to provide some small summary guide to the reader at this point. Let us start
with how these uses affect the predicate.

Table 4. List of predicate-affecting uses of SE and how they affect the predicate
Type of SE Coreference Removing one Removing Imposing an Lexical
among argument case aspectual meaning
arguments reading change

Reflexive or Yes Perhaps (cf. §3.1, Perhaps (cf. No No (but

reciprocal Labelle, Embick) Kempchinsky remember
2004, §5.2) naturally
reflexive
predicates,
§3)
Anticausative No (but Yes Yes Perhaps (cf. Yes
remember (alternatively, to (systematic: §5.5) (systematic:
Koontz- saturate an removal of the elimination
argument subevent of the

111
 ANTONIO FÁBREGAS

Garboden expletively, cf. related to the causation

2009, §4.3) Alexiadou, external cause) component)
Anagnostopoulou
and Schäfer
2015)
Aspectual No (but No (but there No Yes Normally
remember might be not
MacDonald restrictions on the
2017, §5.3) internal argument
and the theta grid
of the verb,
remember García
Fernández 2015,
§5.4)
Agentive No (no extra No (but the No No No
argument properties of the
added), unless external
analysed as a argument become
dative restricted)
reflexive
Antipassive No, unless No (in fact, it Yes No Often (not
analysed as a might involve the (systematic: systematic,
reflexive direct addition of an removal of the new
object argument) accusative meanings in
case capacities the SE and
of the verb) the SE-less
version)
Figure Yes (subject No (in fact, the No No, but the Yes
and entity PP adds argument PP might add (systematic:
displaced in the positions) a result state the agent is
PP structure) in the result
location)
Factitive (if No (unless No (but the No No Yes
different analysed as a subject gets its (systematic:
from others) dative interpretation the subject
reflexive) restricted) is an
instigator)
Transitive No (SE acts as No (but the No No Yes, non
part of the configuration systematic
predicate) might change) (new
meanings)

The table above shows two aspects that I believe deserve additional discussion. The first
one is that one cannot say that SE has the power to systematically impoverish a structure, contra
the claims in part of the literature. It is true that in most cases, SE has a surface effect that is
compatible with interpreting that it removes an argument position, a subevent or the capacity
to assign case to the internal argument. However, at least the agentive / aspectual / factitive SE
cases do not show any sign of reducing or impoverishing the verbal structure in any significant
way. The most recognisable effects of these SE types are (i) some effect on the aspectual
interpretation of the predicate and (ii) some effect on the thematic interpretation of the subject.
These additional restrictions could be operationalised as SE imposing additional requisites on
the predicate. Having additional restrictions, in principle, seems to be more compatible with

112
 SE IN SPANISH

adding more information to the predicate –information that is then used to restrict the
arguments more– than with removing capabilities in the predicate.
The second observation is that there is a predominant type of analysis that stands out in this
table: reducing the types of SE to cases of reflexive SE which are located at different points in
the tree structure. The most uncontroversial case where this strategy is adopted is the figure
reflexive SE –which is simply an instance of SE introduced in an argument position created by
a PP structure–, but as we have seen there are proposals that treat as reflexive virtually any
other kind of SE. Masullo (1992) treats the antipassive SE as a reflexive pronoun located in an
internal argument position, just like Armstrong (2013) treats the transitive SE. The aspectual
SE is analysed as a reflexive dative in MacDonald (2017), García Pardo (2021) proposes that
SE fills argument positions created by new verbal heads, and Kempchinsky (2004) proposes
that SE –due to its feature deficiency– can also be viewed as a reflexive that sometimes projects
as a head in Asp. The core idea that SE is a reflexive element used as the projection of a verbal
head –alternatively, is attached to the verbal head not occupying an argument position– has
also been used to capture agentive SE, anticausative SE and aspectual SE, among other.
Interestingly, the most controversial extension of the reflexive pronoun analysis is in the realm
of anticausative SE, as we saw in §4.3.
The common strategy to these analyses is to propose that SE is inherently a reflexive
element, perhaps reflexive precisely because of the very impoverished feature content that it
displays. The effects that SE has on 'impoverishing' the verbal structure through removing
arguments, etc., are produced because SE itself occupies their position –or is attached to the
head that should introduce them, or marks the head that should introduce them as defective–.
The non-impoverishing effects of SE are not triggered by SE itself: they are produced by
additional verbal heads that introduce further information –new subevents, prepositional
structures, etc.–. SE participates in the structure because the additional structures contain
argument positions where SE fills in. By introducing SE in those argument positions, one
guarantees that there is no increase in the number of participants in the event, despite the
additional structure –because SE, being reflexive, would fill the argument position and force
the interpretation that another argument occupies that position–.

Thus, the most popular analytic strategy with respect to predicate-affecting SE is, by far, to
reduce the different uses to reflexive cases where the differences depend on the structure that
introduces the reflexive pronoun and the location of the reflexive pronoun. However, we have
also seen in these four sections that there have been other attempts to simplify the number of
SE uses by merging together classes or distributing the cases of one type of SE among the more
established cases of SE.

Table 5. List of predicate-affecting uses of SE and how they have been reanalysed
Use of SE taken as basic Classes that fall there
Aspectual (Kempchinsky Anticausative (in both),
2004, Vivanco 2021) reflexive and others (only in
Kempchinsky)
Agentive (DiTullio 2012, Aspectual transitive (Di
partially Armstrong 2013 Tullio), aspectual
and García Fernández 2015) intransitive (García
Fernández, remember §5.4,
where SE makes them
unergative predicates)
Reflexive (most authors) Anticausative (Koontz-
Garboden 2009), aspectual

113
 ANTONIO FÁBREGAS

(MacDonald 2017),
antipassive (Masullo 1992)

With this, we finish our discussion of predicate-affecting uses of SE and move now to the
clause-affecting uses, which have in common that (i) they do not affect the lexical interpretation
of the predicate, (ii) they do not affect the number of arguments and (iii) they do not affect the
aspectual interpretation of the elements involved.

7. Clause-affecting SE uses (1): impersonal

The first clause-affecting use of SE that we will revise here is the impersonal, which is
illustrated with a few examples in (311).

(311) a. Se vive bien aquí.

SE lives well her
'One lives well here'
b. Se habló de este tema en la reunión.
SE spoke of this topic in the meeting
'Someone spoke of this topic at the meeting'
c. Se vio a los sospechosos.
SE saw DOM the suspects
'Someone saw the suspects'
d. Se la localizó en el parque.
SE her found in the park
'Someone found her in the park'

As can be seen, impersonal SE blocks the subject position, which cannot be satisfied by an
independent nominal expression, without altering the rest of the verb's arguments. The
impersonal structure is compatible with intransitive predicates (311a), predicates that introduce
the internal argument with a preposition (311b) or transitive predicates (311c). The case-
assigning capabilities of the verb are not affected by the presence of SE: if the verb is able to
assign accusative case, impersonal SE does not alter this (311d). There are additional
restrictions to the object, that we will partially revise in §7.5, and then later revisit in §8.2, as
they are relevant for the possible unification between passive SE and impersonal SE.
The impersonal use of SE has been broadly analysed in the literature (Oca 1914, Fernández
Ramírez 1957 [1986], Lozano 1970, 1972, Jordan 1973, Suñer 1973, 1974, 1983, 2002,
Llorente 1977, Jaeggli 1986, Otero 1986, Balari & Bel 1990, Moreno Cabrera 1990, Masullo
1990, Mendikoetxea 1992, 1999, 2002, 2008, Mendikoetxea & Battye 1990, Raposo &
Uriagereka 1996, Sánchez López 2002, Ordóñez & Treviño 2011, 2016, Ordóñez 2021,
MacDonald & Melgares 2021, to name just a few). What makes impersonal SE special within
the system of SE uses is that impersonal SE is the only instance where SE is related to a
nominative position, not a (bona fide) accusative or dative position, as it is the case in the rest
of uses (Dobrovie-Sorin 1998). In fact, not all Romance languages that have anticausative or
passive SE allow also impersonal SE –French or Romanian are instances of the absence of the
impersonal SE–, a fact that has been analysed as these languages lacking a nominative SE form.
In this section, we will first provide the main properties of impersonal SE, empirically
considered, from a syntactic and a semantic perspective (§7.1, §7.2). Then we will move to the
main analyses (§7.3, §7.4). What we will see as a common trend in the analyses available is
that (i) impersonal SE is taken to be a reflexive pronoun located in a high position or that moves
to a very high position, perhaps TP and (ii) given its reflexive nature, and how high in the
structure it is, impersonal SE is not c-commanded by any argument of the verb, so it must be

114
 SE IN SPANISH

associated to an empty pronominal category that determines its interpretation, or is itself

analysed as a defective pronoun, sometimes even equivalent to PRO. Then we will discuss the
properties of the object in the impersonal SE construction, something that will allow us to
present the analyses that have tried to unify impersonal and passive SE (§7.5), and we will end
with a discussion of how impersonal SE uses compare to the predicate-affecting ones. Let us
proceed.

7.1. Main properties

The impersonal SE constructions are characterised by two main properties that differentiate
it from passive SE (§8): the verb appears fixed in a default third person singular inflection (312)
and the case properties of the verb are not altered (313).

(312) a. Se recibió a los embajadores.

SE received DOM the ambassadors
b. *Se recibieron a los embajadores.
SE received.3pl DOM the ambassadors
(313) a. Juan los recibió.
Juan them.acc received
b. Se los recibió.
SE them.acc received

The non-paradigmatic nature of SE, which in the impersonal construction cannot alternate
with the person-marked clitics (me, te, nos...) might actually be an epiphenomenon of the
inflection of the verb being fixed to this 3sg value; in fact, as we will see, in theories where SE
is attached to T (or Flex, in an older terminology; cf. Belletti 1982, Jaeggli 1986, Otero 1986),
this fixation into the default value is forced by SE.
In contrast to the passive SE structure, impersonal SE can be combined not only with
transitive verbs (312, 313), but also with verbs that mark the internal argument obliquely with
a preposition (314) or intransitive verbs.

(314) a. Se confía mucho en ellos.

SE trusts much in them
'One trusts them a lot'
b. Se pelea mucho con ellos.
SE fights much with them
'One fights with them a lot'
c. Se discute mucho de este tema.
SE argues much about this topic

In intransitive verbs, impersonal SE allows in principle both unaccusative (315) and

unergative verbs (316), although as we will see later, unaccusative verbs have additional
temporoaspectual restrictions that transitive or unergative predicates lack.

(315) a. Siempre se llega tarde a estos sitios.

always SE arrives late to such places
'One always arrives late to these places'
b. Se nace y se muere.
SE is.born and SE dies
'One is born and one dies'
(316) a. Se nada demasiado en esta playa.

115
 ANTONIO FÁBREGAS

SE swims too.much in this beach

b. Se tose mucho con esta enfermedad.
SE coughs much with this sickness

Periphrastic passives and copulative verbs also allow impersonal SE, again with restrictions
shared with unaccusatives that we will detail later.

(317) a. Se es detenido inmediatamente si se hace eso.

SE is detained immediately if SE does that
'One gets immediately detained if one does that'
b. Se está bien aquí.
SE is fine here
'One is fine here'

This does not mean that any verb can have an impersonal SE version. First, only verbs that
accept a personal subject can combine with impersonal SE: lexically impersonal verbs, like
weather verbs, reject impersonal SE, just as predicates that combine with propositional subjects
(Mendikoetxea 1999).

(318) a. Llueve.
rains
'It rains'
b. #Se llueve.
SE rains
Intended: *'One rains'
(319) a. Está claro [que pasa eso].
is clear that happens that
'It is clear that that happens'
b. *Se está claro.
SE is clear

Verbs where the predicate is clasified as a dative experiencer psych predicate (Belletti &
Rizzi 1988) also reject impersonal SE (Ordóñez 2021).

(320) a. A Juan le gusta eso.

to Juan him likes that
'Juan likes that'
b. *Se gusta a Juan.
SE likes to Juan
Intended: 'Juan likes someone / everybody'
(321) a. A Juan le duele eso.
to Juan him hurts that
'That hurts Juan'
b. *Se duele a Juan.
SE hurts to Juan
Intended: 'Something / everything hurts Juan'

These verbs have been argued to have a structure where the nominative agreeing nominal is
an internal argument and the subject is the quirky marked dative experiencer. This has led

116
 SE IN SPANISH

several authors (eg., Belletti 1982, Suñer 2002, Ordóñez 2021, pace Dobrovie Sorin 1998) to
argue that impersonal SE must be related to subjects marked with nominative.
Finally, raising predicates reject impersonal SE too, perhaps because they do not select their
own subjects.

(322) a. Parece que Juan está enfermo.

seems that Juan is sick
'It seems that Juan is sick'
b. Juan parece estar enfermo.
Juan seems to.be sick
c. *Se parece estar enfermo.
SE seems to.be sick
Intended: 'One / everybody seems to be sick'
d. *Se parece trabajar mucho.
SE seems to.work a.lot
Intended: 'Everybody seems to work a lot'

Note, however, that this cannot be the only reason for (322) to be ungrammatical: auxiliary
verbs are supposed not to select their subjects either, but they can combine with a SE that is (in
principle) associated to the thematic subject of the subordinate infinitive (Se los puede invitar,
'SE them.acc can invite'). From this perspective it is unclear why (322c) or (322d) should not
license the subject in the same way, with the inescapable conclusion that auxiliary + infinitive
cannot be the same type of syntactic construction as parecer + infinitive.
Additionally, predicates that have a predicate-affecting SE cannot combine with impersonal
SE, something that follows from the already noted fact (§2) that one cannot have more than
one SE (or for that matter, more than one clitic of the same type) in the same clause: from se
acostumbran a eso 'SE get.used to that, They get used to that' one cannot make #Se acostumbra
a eso to mean 'One gets used to that', and from se arrepienten 'SE repent, They repent', one
cannot form #Se arrepiente 'One repents'. As explained in §2, two SE clitics cannot appear
together, and not surprisingly the one that seems to be preserved is the predicate-affecting one
–because it is part of the lexical structure of the verb–, blocking the clause-affecting impersonal
SE. We will get back later (§9) to situations where clitics that are too similar to each other
trigger relevant problems on the surface form of the elements involved in a structure.

7.2. The interpretation of the subject: existential and generic readings

A second aspect of the syntax and semantics of impersonal SE sentences is the interpretation
of the subject that SE imposes. The first aspect of this interpretation is that the subject has to
be interpreted as human, or at least as personified. Despite chosing verbs that are generally
predicated from animals, (323a-c) requires a human metaphorical interpretation due to the
presence of impersonal SE, and verbs without a possible human referent (323d-e) reject
impersonal SE precisely because of this (Ormazabal & Romero 2019).

(323) a. Se ladra mucho en esta clase.

SE barks much in this class
b. Se muerde mucho en esta clase.
SE bites much in this class
c. Se trota mucho en esta clase.
SE trots much in this class
d. *Se ocurre.
SE happens

117
 ANTONIO FÁBREGAS

e. *Se transcurrió.
SE passed.time

There are also relevant properties in terms of the reference of that subject. Since Cinque
(1988) there is a distinction between the quasi-generic and the quasi-existential reading of
impersonal SE sentences. Both of them have been labeled as 'non specific' in part of the
literature (Sánchez López 2002), although this term is somewhat confusing, because specificity
is generally related to the absence of a referent for the nominal expression (see Fábregas 2018
for an overview), when impersonal SE generally imposes the referential interpretation, even if
the reference of that entity is not definite and does not correspond to a previously unique group
introduced in the discourse.
The quasi-generic reading of the impersonal SE is associated to the paraphrase 'everybody,
people, one, each element in the group'. Every single type of predicate that accepts impersonal
SE allows this reading, which also correlates with the use of imperfective temporoaspectual
forms of the inflection (De Miguel 1992).

(324) a. Aquí se recibe a los embajadores.

here SE receives DOM the ambassadors
'Here people receive the ambassadors'
b. Se vive bien aquí.
SE lives fine here
'One lives fine here'
c. En los países desarrollados se nace sin infecciones.
in the countries developed SE is.born without infections
'One is born without infections in developed countries'
d. Se vuela bien con esta compañía.
SE flies well with this company
'One flies well with this company'
e. Se habla mucho de esto.
SE talks much about this
'People talk a lot about this'
f. Se es amable si se hace eso.
SE is kind if SE does that
'One is kind if one does that'
g. Se es examinado en esta aula.
SE is examined in this room

The quasi-existential reading, also termed arbitrary, means that there is at least one person
whose identity is not disclosed that corresponds to the reference of the subject. The normal
gloss is 'someone'. Crucially, this interpretation is out with the predicates where the subject is
assumed to be derived from an internal argument position: unaccusative verbs, passives and
copulatives.

(325) a. Se llamó a la puerta.

SE knocked at the door
'Someone knocked at the door'
b. Se habló del tema en la reunión.
SE spoke of the topic at the meeting
'Someone spoke of the topic at the meeting'
c. Se nadó en esta piscina.

118
 SE IN SPANISH

SE swam in this pool

'Someone swam in this pool'
(326) a. *Se nació en este hospital.
SE was.born in this hospital
Intended: 'Someone was born in this hospital'
b. *Se fue detenido por ese crimen.
SE was detained for this crime
Intended: 'Someone was arrested for this crime'
c. *Se fue amable en la fiesta.
SE was kind in the party
Intended: 'Someone was kind at the party'

As seen in (325), the quasi-existential reading is related to non-imperfective temporal

formsl, which means that the absence of an existential reading in (326) blurs together with the
impossibility of having a non-imperfective form with impersonal SE sentences in
unaccusatives, passives and copulative verbs. The impossibility of having existential readings
with impersonal SE with unaccusative predicates explains, moreover, that some existential
unaccusative predicates (faltar 'to lack', quedar 'to remain') cannot combine with impersonal
SE (Mendikoetxea 2002).
Beyond the fact that the data are not crystal clear for all speakers (Sánchez López 2002 gives
a sentence equivalent to 326b as grammatical), this double ban in the aspectual forms and the
existential reading of the subject has triggered two different types of explanations. In Cinque
(1988), the problem is that derived subjects coming from internal argument positions cannot
have an existential reading: he proposes that the existential reading is imposed before
movement to a specifier position. The subjects of the unaccusative, passive and copulative
verbs are merged before movement in the complement position of the verb, so they cannot
receive the existential interpretation. De Miguel (1992) proposes that the problem is the
aspectual forms: unaccusative verbs, in combination with impersonal SE, are inherently
procesual and are not compatible with specific tenses or aspects. Mendikoetxea (2002)
criticises both approaches and proposes that what causes the ungrammaticality in (326) has to
do with the interpretation of SE, which should be considered an equivalent of PRO, an empty
category that lacks its own reference, and where the reference has to be controlled by another
element. Sánchez López (2002) suggests that the ban is not actually absolute for the existential
reading of some derived subjects. However, before discussing this type of approach we must
sketch the main syntactic analyses of impersonal SE.

7.3. Impersonal SE as PRO or as an impoverisher of T

Leaving technical details aside that are not relevant right now, there are two major analyses
of impersonal SE in the market: an analysis where SE is itself in the subject position, as a non
referential subject equivalent to PRO, and an analysis where SE is in the head of T / Flex, and
has the role of impoverishing that node, forming a chain with an empty pronominal that is in
the specifier of TP / FlexP. Let us revise both approaches.
Oca (1914) is cited in Spanish as the first person that argued that impersonal SE should be
seen as a generic subject, equivalent to expressions like French on 'one', English one, or
German Man. In modern terms, for many of the proponents of this theory, this involves treating
impersonal SE as the equivalent of PRO, a pronoun without reference that appears in contexts
with low agreement, and which has to be assigned a reference by a controller. Teomiro (2005;
see also 2010, 2011) and, with qualifications, Mendikoetxea (2002), are among the authors that
in modern times have argued that this is the right analysis of impersonal SE: Teomiro argues
that SE is a subject anaphor and English PRO is a null version of it. With qualifications,

119
 ANTONIO FÁBREGAS

Mendikoetxea (2002) also advocates for the idea that impersonal SE is related to PRO, although
she does not place the SE in the spec, TP position (occupied by a PRO, which is coindexed
with SE, in the head position). In the following discussion, with this caveat in mind, we locate
Mendikoetxea (2002) in this family of theories, because the head or specifier status of SE has
no role in her analysis, where the crucial property is the identification of PRO with SE. A close
version of this type of analysis of SE as a pronominal without reference that appears in the
argument position of the subject (for impersonal SE) is Pujalte & Saab (2012) and Saab (2014),
who propose that SE is introduced at PF to fill an unsaturated argument position.
Not all these analysis coming from Oca (1914) argue that SE is equivalent to PRO or should
be seen as an empty filler of an argumental position. Ormazabal & Romero (2019) treat
impersonal SE as a standard pronoun that occupies a subject position, and must receive
nominative case. The configuration where impersonal SE ends up in these analyses is
represented in (327).

(327) TP

SE T

T ...VoiceP

Voice vP

v VP

SE V

V ...

Note, however, that if SE is a subject, one would have to explain through some other means
why SE follows and does not precede negation, unlike other subjects –a property that perhaps
could be made follow from the clitic status of both negation and SE in Spanish–.
In contrast, a second line of approach to this construction, which we could say is currently
predominant, goes back to Belletti (1982) and has Jaeggli (1986), Otero (1986) and, in more
recent times, Suñer (2002), Ordóñez & Treviño (2011) or Ordóñez (2021) as the main
proponents: the idea is that impersonal SE is attached to the head T / Flex, where it has an
impoverishing role –as in many cases of predicate-affecting SE–.

(328) TP

proi T

T ...VoiceP
sei
Voice vP

v VP

pro V

V ...

120
 SE IN SPANISH

Let us schematise the two main approaches, with the two variants of the first approach:

A. SE-as-subject
i. SE is PRO, is coindexed with PRO or fills an unsaturated argument position
ii. SE is a normal subject pronoun
B. SE-as-T
SE impoverishes the T node and is coindexed with an indefinite pro

As we will see, the comparison between Aii and B is more subtle than the comparison
between Ai and B. Thus, we will first concentrate on the approaches Ai and B, and in the next
section we will compare Aii to B.
In the approach Ai, impersonal SE is a version of PRO. One would explain that the inflection
is fixed to 3sg because of the lack of features internal to SE / PRO. Given where it is placed,
as a subject, combined with its lack of features, the existential or generic reading would follow
–pending details that we will get back to–. As the only argument position affected by it is the
subject, we expect precisely that it does not affect the other arguments in the predicate or the
case-assigning capacities of the predicate. As the SE must be coming thematically from the
verbal predicate, and is forced to have a human interpretation, one explains that (i) only verbs
with personal subjects selected thematically can have impersonal SE and (ii) that there are
effects depending on the origin of that SE thematically.
In the approach B, the presence of SE impoverishing T / Flex explains why the inflection is
fixed to 3sg. The existential or generic readings, as well as other interpretations that we will
discuss later, depend on the nature of the pronominal that moves to the subject position from
the verbal complex. The origin of that pronominal is identical to the one that SE has in the
approach of impersonal SE as PRO, which means that the other properties are explained just
as in that theory.
The approach where SE is in the T head and is coindexed with a pronoun has a bit of an
advantage in explaining that the verbs where the subject is a dative cannot allow impersonal
SE: the idea, going back to Belletti (1982), is that impersonal SE has to be controlled by a
nominative argument, or in other terms, that the impoverishing of T due to the presence of SE
cannot license dative or other quirky-marked subjects (see Suñer 1992). If SE is a subject itself,
it becomes surprising that it cannot occupy the place of a dative subject –remember from §3
that SE emerges with datives, and see also §10 below for more relations between SE and
datives–. The approach where SE is in the subject position has to relate the impossibility of
(320b) with the impossibility of having controlled or arbitrary PRO structures with dative
experiencer predicates:

(329) *María prometió PRO gustarle la comida.

María promised like.her the food
Intended: 'María promised to like the food'

In fact, one argument used by the side of the debate where impersonal SE is PRO is the
meaning correspondence between the interpretations of PRO and the meaning of impersonal
SE (Mendikoetxea 2002). For those that propose that impersonal SE is not the subject, the
meaning distinction between the generic and the existential reading has to follow from the
interpretation of the empty pronominal category that is coindexed with SE, and then one has to
assume that the existential reading is not available as an interpretation of pro for internal
arguments to explain that unaccusatives, passives and copulative verbs must have a generic
interpretation. One can assume Diesing (1992) in her proposal that universal readings are

121
 ANTONIO FÁBREGAS

obtained by adding a universal quantifier high in the clause, while the existential reading
requires existential closure at the VP level. Imagine that this restricts the possibility of
assigning an existential reading to the specifier position of VP (modernly, vP or VoiceP); like
this, subjects that are derived from internal argument positions will not have an existential
reading because they are not located in that position, while external arguments, above VP,
would be able to acquire the existential or generic interpretation if the appropriate operator is
introduced. Alternatively, the problem might be related, as Suñer (2002) suggests, with the
licensing of arbitrary internal argument PRO, which is also blocked when the verb appears in
perfective aspect.
Mendikoetxea's (2002) analysis of the impersonal SE as related to PRO also assumes
Diesing (1992) but gives them a different take. The crucial difference between the predicates
that allow the existential reading of PRO / SE and those that don't is which element restricts
the reference of the subject. In her proposal, one has to accept that PRO, lacking enough
interpretation, cannot be used to restrict the reference of the subject, something that is
independently necessary. The verbs that have a thematic and referential external argument –
transitives, unergatives– can project a spatiotemporal argument in the subject position that
restricts the reference of SE. The whole sentence is then interpreted as an assertion that is
predicated from that spatiotemporal argument, and the existential reading is available.
However, the verbs that lack an external argument that is thematic and referential –
unaccusatives, passives, copulative verbs– cannot use this spatiotemporal argument to restrict
the reference of the subject. In order to complete the predication, then, the only option available
in the absence of this referential spatiotemporal argument is to interpret that there are two
predicates holding an inclusion relation (Kanski 1992), a property that forces a generic reading.
Following Kanski (1992), a sentence like (330) is interpretable only if the property of being
human can be said to be included in the property of being ferocious, that is, if humans can be
included in the set denoted by 'being ferocious'.

(330) El hombre es feroz.

the man is ferocious

Given that SE / PRO cannot restrict, (331a) is ungrammatical and a secondary predicate
must be added to the structure to provide the inclusion relation that licenses (330).

(331) a. #Se es feroz.

SE is ferocious
b. Se es feroz cuando hace mucho calor.
SE is ferocious when there.is much heat
'One is ferocious when it is too hot'

Once this secondary predicate is added, (331a) can be the predicate acting as the subject
included in the secondary predicate: being ferocious is part of the situation defined when it is
too hot. Technical details aside –the complexity of the semantic analysis is too high to revise
it fully here– the proposal is that SE, being PRO, cannot restrict the subject enough on its own.
Verbs that have a thematic external argument can choose to introduce a spatiotemporal
argument in the subject position, and obtain an existential reading; verbs that lack that thematic
external argument must complete the predication treating 'SE predicate' as the subject of a
secondary predicate, which includes it, and that forces a generic interpretation.

122
 SE IN SPANISH

7.4. Other facts about the impersonal SE structure: comparison between approaches
Thus, the facts revised so far seem to constitute a tie between the approaches Ai (SE as a
PRO subject) and B (SE as impoverisher of T): the core facts can be accounted by both
approaches, one proposing that SE impoverishes the inflection and is coindexed with a pro
whose referential properties depend on the availability of existential closure / generic readings,
and the other proposing that the 'defective' nature of TP is related to the nature of SE as PRO,
or its compulsory coindexation with PRO. Let us now revise, however, several phenomena that
support one or the other approaches.
We will see that the end result seems to be that the approach where SE is related to PRO or
is treated as an empty filler (Ai) seems to encounter too many empirical problems, in
comparison to the approach where impersonal SE impoverishes T (B). The first part of the
section concentrates on comparing these two approaches. An important reminder at this point
is that the comparison leaves out so far the approaches in Aii, that treat SE as a 'normal' subject
pronoun, specifically Ormazabal & Romero (2019), who do not associate SE to PRO. The
critiques that will be presented against the SE-as-subject approach in the following pages do
not apply to Ormazabal & Romero's (2019) proposal, which we will discuss at the end of this
section.

a) The approach in (328), which claims no connection between PRO and impersonal SE,
has to face a particular challenge when it comes to the impossibility of (332), that is, the
presence of an overt nominal subject is empirically incompatible with impersonal SE:

(332) #La gente se vive bien aquí.

the people SE lives well here

In the approach of SE as PRO, (332) is understandably out because (i) PRO occupies the
subject position and (ii) the referential properties of PRO do not allow it to be substituted by
an overt nominal. In the approach of SE as marking defective T, SE is coindexed with a pro,
which in principle can have its own referential properties, so it should be substitutable by an
overt nominal to the same extent that (333) can alternate.

(333) a. Esas personas hablan mucho.

those people talk much
b. pro hablan mucho.
pro talk much

Suñer (2002) acknowledges this as a problem for the approach, but notes that there is a way
out in the analysis where SE impoverishes T: perhaps the absence of nominal overt subjects is
an effect of the impoverishment of T, which does not allow T to license in its spec overt
subjects. However, if SE has to be associated to a nominative pro, the theory assumes that the
impoverished T assigns case to the subject, so it is unclear why overt nominals should not be
licensed in this context.

b) PRO is generally a category that appears in the subject position of non finite verbal forms
and is traditionally interpreted as not receiving case. If SE is PRO, or alternatively it is
coindexed with PRO, then, it comes as a surprise that impersonal SE cannot appear in a number
of control contexts (data taken from Mendikoetxea 1999):

(334) a. Sería mejor convocar elecciones.

would.be better to.call ellections

123
 ANTONIO FÁBREGAS

b. *Sería mejor convocarse elecciones.

would.be better to.call-SE ellections
Intended: 'It would be better that one calls an ellection'
(335) a. La gente desea trabajar menos.
the people wish to.work less
b. *La gente desea trabajarse menos.
the people wish to.work-SE less
Intended: 'People want that everybody works less'
(336) a. La policía prohibió entrar en el recinto.
the police forbid to.enter in the area
b. *La policía prohibió entrarse en el recinto.
the police forbid to.enter-SE in the area
Intended: 'The police forbid that anyone enters the area'
(337) a. Vimos aclamar a los vencedores.
we.saw to.cheer DOM the winners
b. *Vimos aclamarse a los vencedores.
we.saw to.cheer-SE DOM the winners
Intended: 'We saw everybody cheer the winners'
(338) a. El gobierno hizo construir embalses.
the government made to.build reservoirs
b. *El gobierno hizo construirse embalses.
the government made to.build-SE reservoirs
Intended: 'The goverment made someone build reservoirs'

These are standard control contexts, with arbitrary PRO (334) or controlled PRO
coreferential with the subject (335) or the object of the main clause (336), next to perception
verbs (337) and causative verbs (338). The chances that an approach where SE is PRO has to
explain these facts are not very good: assuming that PRO is present in the grammatical member
of each pair, the question is what is SE adding so that SE cannot appear in the ungrammatical
pairs. The licensing conditions of PRO seem to be in place given the grammatical members,
and SE simply coindexes with PRO (or is a manifestation of PRO itself), so it is unclear what
would block it.
There are two immediate ways out, although both require some extra motivation: (i) ø and
SE are allomorphs of the same category, PRO, in a way that PRO materialises as SE in finite
contexts and as ø in non finite contexts; (ii) SE is PRO but carries additional case licensing
conditions that the non finite context does not provide. In fact, Ormazábal & Romero (2019),
who follow Oca (1914) in the proposal that SE is the subject in these constructions, explicitly
deny any direct connection with PRO, and propose that SE is simply an underspecified
nominative pronoun in these cases. For them, what goes wrong in the non finite contexts is
plainly that infinitives cannot assign nominative to their subjects. Like this, their analysis
accommodates the facts while at the same time proposing that SE occupies the subject position.
From the perspective of the theory where SE impoverishes T –where impoverishing is to
make unavailable features that are otherwise active in T–, the explanation of these facts is
straightforward and was already formulated by Otero (1986). Assume that SE impoverishes T.
In order for SE to appear in T, there should be something that can be impoverished in T. If non
finite T is already impoverished, the addition of SE is unnecessary –and hence impossible–
because there is nothing left to impoverish anymore. This explanation has to assume, crucially,
that temporal and aspectual inflection do not count for the impoverishing role of SE, because
finite contexts with SE have tense and aspect that can be manipulated, while non finite contexts

124
 SE IN SPANISH

only allow some aspectual inflection. If SE had the power to block the inflection of infinitives
as perfect or not (339), then SE could meaningfully combine with a non finite verb.

(339) a. La gente desearía trabajar menos.

the people would.wish to.work less
b. La gente desearía haber trabajado menos.
the people would.wish to.have worked less

However, this conclusion seems to be granted; impersonal SE does not block per se the
temporal or aspectual inflection of the predicate, as in finite forms one can find past, present
and future verbs, as well as perfective, imperfective and perfect aspects. MacDonald &
Melgares (2021), for instance, propose that what impersonal SE does is to absorb a definiteness
feature in T that blocks the presence of definite subjects. Once definiteness is unavailable, the
subject can only be interpreted as indefinite generic or existential.

c) The proposal where SE is PRO finds an additional problem in the interpretation of the
subject once one compares Spanish with other Romance languages. D'Alessandro (2007) notes
that in Italian the existential reading of the unaccusative, passive and copulative verbs is
possible in impersonal SE cases only if a first person plural reading is obtained. Note the plural
inflection in the participles in (340a).

(340) a. Noi, ha detto che non si è stati invitati.

we, has said that not SE has been invited
'We, he said that we have not been invited'
b. Si va!
SE goes
'Let's go'

Spanish is among the languages that reject the interpretation where the 1st person reading
saves the existential interpretation.
It is, to begin with, difficult to understand how this reading can be explained as possible in
Italian but not Spanish in a PRO account (or using the tools of Mendikoetxea 2002, given that
PRO should be universally unable to restrict the subject, by hypothesis given its feature
impoverishment).
In contrast, if impersonal SE is controlled by a pronominal (Ordóñez 2021), the distinction
between Italian and Spanish depends only on the range of pronouns that one can associate to
SE under coindexation in such cases: Italian allows an empty pronominal that includes the
speaker, while Spanish lacks it; the presence of that empty pronominal adds the speaker but SE
has impoverished the T node, and therefore it is not possible to exhibit 1st person plural
agreement in T.

d) PRO is taken to be a universal category: languages that are not prodrop, like English or
French, are assumed to also have PRO, because the behaviour of control verbs in those
languages is not significantly different from how prodrop languages like Spanish or Italian treat
those constructions. From the perspective of the theory where SE is PRO or closely related to
it, then, it comes as a surprise that impersonal SE is not more widely represented. Take French
for instance: in French, the impersonal SE is not used, even if the language has the anticausative
and the passive SE. French is supposed to have PRO, so this lack of impersonal SE comes as a
surprise. In contrast, since Belletti (1982), the approach to impersonal SE where SE is
coindexed with an empty non definite subject predicts precisely that impersonal SE should be

125
 ANTONIO FÁBREGAS

lacking from non prodrop languages, like French (or for that matter, English or German): in
the absence of pro and with SE occupying the position of head in T / Flex, the impersonal SE
construction would not produce as a result any type of well formed clause.

Thus, the conclusion so far seems to be that the approach where impersonal SE is itself PRO
or is closely related to PRO is empirically less straightforward than the approach where SE is
introduced in T to impoverish the node, fix the agreement in 3sg, and block the licensing of
overt nominal subjects.
This, however, does not refute the analysis of SE as a pronoun in the subject position, but
rather the specific analyses of SE where it is equivalent or coindexed to PRO. We still have
approach Aii to compare to approach B. Aii is represented by Ormazabal & Romero (2019),
who keep the SE-as-subject analysis but without a correlation with PRO. Their analysis is very
difficult to disentangle from the analysis of SE as an element in T, because in that second
analysis there is an empty indefinite pronoun that is coindexed with SE, so on the surface
pro+SE might behave exactly as SE in Ormazabal & Romero's (2019) analysis. However, there
are a few tests that in fact can distinguish the two approaches. The arguments that these authors
give in favour of their approach, where SE should be viewed as a normal subject pronoun, can
be summarised as follows, where we add the appropriate comments to compare it to the
approach where SE is in T.

i) Only predicates that have a thematic argument position available allow SE, which makes
sense if SE occupies that argument position and moves from there to become the subject. Note,
however, that in the SE-in-T approach, this property is explained because the pro coindexed
with SE must be initially located in an argument position.
ii) Impersonal SE constructions can control the subject of infinitives just like other subjects
(341), and are subject to the same obviation effects as normal subjects (342). In the same way
as in (342a) the verb intentar 'try' imposes disjoint reference to the subject of the inflected
subordinate clause, in (342b) the indefinite subject must be distinct for each verb.

(341) a. Pedro intentó ir al cine.

Pedro tried to.go to.the movies
'Pedro tried to go to the movies'
b. Se intentó ir al cine.
SE tried to.go to.the movies
'Someone tried to go to the movies'
(342) a. María intentó que enviara un libro.
María tried that sent.sbj a book
'Maria tried for somebody else to send a book'
b. Se intentó que se enviara un libro.
SE tried that SE sent.sbj a book
'Someone tried for someone else to send a book'

Note, however, that the proposal where SE is in T is able to express precisely this fact,
because the indefinite pro that is coindexed with it has to comply with the same kind of disjoint
reference restriction. The data constitute, though, a counterexample to the proposal of
impersonal SE as PRO, which by definition lacks obviation effects.

In these two tests, the approach of SE-as-subject and the approach of SE-in-T seem to tie.
However, once removed the PRO proposal, there are two empirical arguments that give some

126
 SE IN SPANISH

additional support to the SE-as-subject approach, and one empirical argument that constitutes
an initial problem for the SE-as-subject approach.

iii) Impersonal SE cannot appear together with infinitives; this property is explained in
Ormazabal & Romero's (2019) account as a condition on the subject: overt nominative
pronouns are generally not possible in infinitives (cf. 334-338 above). Remember that the
explanation in the SE-in-T approach (B) was that the infinitive T was already impoverished,
so SE could not be introduced there. Ormazabal & Romero (2019), however, show that
impersonal SE can appear with some control infinitives: precisely the infinitives that can
license another nominative pronoun (generally because they are introduced by a preposition,
which might allow T to license the nominative case of the subject even in the infinitive).

(343) a. Al descubrir yo a los culpables...

at.the discover I DOM the culprits
'As I discovered the culprits...'
b. Al descubrir-se a los culpables
at.the discover-SE DOM the culprits
'As someone discovered the culprits...'
(344) a. De haber visto yo a esas personas...
Of having seen I DOM those people
'If I had seen those persons...'
b. De haber-se visto a esas personas...
Of having-SE seen DOM those people
'If someone had seen those persons...'
(345) a. Por acusar yo a esa persona...
by accuse I DOM that person...
'Because I accused that person...'
b. Por acusar-se a esa persona...
by accuse-SE DOM that person
'Because someone accused that person...'

This fact gives a certain edge to the Aii proposal, because it is unclear how the B approach
could explain this: after all, these infinitives are, in T, the same type of infinitives than in the
examples where SE is unavailable, and the difference seems to be with respect to whether
nominative pronouns are allowed. Maintaining the approach initiated by Otero (1986) would
require here a further argumentation; one would have to argue that the presence of the
preposition specifically strengthens the T node enough so that adding SE creates a significant
reduction in the information carried by T. That explanation is not implausible: if the
definiteness feature is what SE impoverishes in T, the grammaticality of specific, definite
nominative pronouns as subjects in the examples (343a, 344a, 345a) can suggest that the
infinitival T has in these cases a definiteness feature which SE could in principle cancel.
However, as far as I know this explanation has not been offered in the literature.

iv) Remember that impersonal SE forces a human reading of the subject, so that verbs with
non human subjects cannot take it and verbs expressing animal actions are interpreted
metaphorically. In the approach B, this has to be introduced as a somewhat idiosyncratic
restriction on the pro that is coindexed with SE or as a parochial effect of the impoverishment
of T, forcing indefinite readings of the subject (Cinque 1988). If SE is a subject pronoun,
however, the human interpretation is simply a case of the general requisite of overt strong
pronominal subjects in Spanish, which force animate interpretations (Cardinaletti & Starke

127
 ANTONIO FÁBREGAS

1999). (346a) would be interpreted as an animate subject for the same reason that (346b) is,
namely the presence of the overt nominative subject.

(346) a. Se ladra mucho.

SE barks a lot
'One barks a lot'
b. Él ladra mucho.
he barks a lot
'He (=human, not a dog) barks a lot'

v) However, the Aii approach treats SE as a normal subject, and then it predicts, against the
facts, that (348) should be grammatical, where in parallel with (347), the subject of the
infinitive raises to the T node of parecer 'seem' in a raising environment. Remember, however,
that we saw above that raising verbs cannot combine with impersonal SE.

(347) a. Parece [estar Juan enfermo].

seem to.be Juan sick
b. Juan parece [estar enfermo].
Juan seems to.be sick
(348) a. Parece [estar-se enfermo].
seem to.be-SE sick
b. *Se parece [estar enfermo].
SE seem to.be sick
Intended: 'Someone seems to be sick'.

Ormazabal & Romero (2019), in fact, document a number of cases in corpora where SE
seems to have raised from the infinitival clause, among those the examples in (349), which in
contrast to (348b) sound perfectly natural and are clear instances of impersonal SE.

(349) a. Se parecía haber ganado en altura.

SE seemed to.have gained in height
'It seemed that someone had got an increase in height'
b. Con ello, se parece abogar por dotar a la percepción de una unidad
with that, SE seems to.advocate by providing the perception of one unity
'With that, one seems to advocate in favour of the perception of unity'
[adapted from CORPES, 17/11/2016]

The SE-as-subject approach perhaps could find some independent reason for the
ungrammaticality of (348b) in contrast to the naturally sounding examples in (349): perhaps
grammatical aspect might play a role in this task, as the impersonal SE is particularly degraded
in raising contexts when the predicate is stative (*Se parece ser bueno 'One seems to be good'
vs. Se parece negar la libertad a los demás 'One seems to deny freedom to others'). Ormazabal
and Romero (c.p.) propose that divergent examples like (348b) might simply depend on SE
having a prefered position within the clause where it originates.
On the other hand, the approach of SE-in-T seems to explain these facts unproblematically:
SE cannot be base generated in the infinitive, and in the inflected verb SE can only be generated
if c-commanded by pro in a thematic position, a property that a raising verb does not satisfy.
Then, that pro could only come from the infinitive, but the pronominal category of the infinitive
would be PRO, which could not move up to the finite verb.

128
 SE IN SPANISH

Thus, to wrap up this section, there is a lot of empirical arguments that pose problems for
the approach of SE as PRO or related to PRO. However, a meaningful debate can be established
between the approach where SE is in the subject position and the approach where SE creates a
defective T and is coindexed with pro.

7.5. Relation between impersonal SE and passive SE: case and the internal argument
So far, we have analysed the properties of the subject and the predicate in the impersonal
SE structure, but we have not said much about the object of the transitive verb in the impersonal
SE construction. This element is relevant for the nature of SE because of two theoretical points:
(i) the relation between impersonal SE and passive SE and (ii) the case relation that is
established between SE, the verb and the possible pronominal that is added to the structure.
When it comes to the relation between impersonal and passive SE, we leave most part of
the empirical discussion about the variation between the two structures and its consequences
for §8.2, where passive SE is discussed. For the time being, however, we want to establish
some generalisations that will be relevant for the characterisation of the case relation that will
be discussed here, and which has led a number of authors to the claim that impersonal SE
should be viewed in essence as a voice phenomenon that is parallel to passive SE. In essence,
these authors argue that impersonal and passive SE can be reduced to basically the same type
of configuration.
With some variation and many more qualifications that will be presented in §8.2 below,
impersonal SE is strongly favoured over passive SE when the direct object is marked by a,
generally taken to represent DOM in Spanish. Verbs that tend to prefer this marking such as
ayudar 'help' (350a,b), and direct objects that trigger this marking given their referential
properties (350c,d; cf. Fábregas 2013) reject the passive SE construal with agreeing verbs, in
most varieties of Spanish.

(350) a. Se ayuda a los niños.

SE helps DOM the children
'One helps the children'
b. *Se ayudan niños.
SE help.3pl children
Intended: 'Children are helped'
c. Se me vio a mí.
SE me saw.3sg DOM me
d. #Me vi yo.
me saw.1sg I
Intended: 'I was seen'

In contrast, although with more variation among speakers than the previous case (as we will
see in §8.2), predicates containing internal arguments that reject DOM tend to favour the
passive SE construal over the impersonal SE:

(351) a. Juan vende pisos.

Juan sells apartments
b. Se venden pisos.
SE sell.3pl apartments
'Apartments are sold'
(352) a. Juan vio niños.
Juan saw children
b. Se vieron niños.

129
 ANTONIO FÁBREGAS

SE saw.3pl children
'Children were seen'

This alternation between impersonal and passive SE is not as clean as these data suggest,
but they are the starting point for the following question: given that the properties of the object
in the active form condition whether impersonal or passive SE can be used, could one propose
that impersonal and passive SE start from the same basic structure, and the nature of the object
–perhaps, the type of case that the object receives– determine the difference between the two
types of SE?

7.5.1. Impersonal SE and passive SE as the same structure

These facts are relevant for the debate about whether impersonal SE absorbs or blocks
accusative case from the verb or not, and secondarily with respect to whether impersonal SE is
related to a nominative subject or other cases are allowed. The second question is more
straightforward to answer, so let us remove it from the table: the evidence in Spanish
(remember the impossibility of using impersonal SE with verbs that assign dative to the
subject) seems to suggest that Spanish impersonal SE is associated only to nominative subjects.
Impersonal SE might be the only case of SE related to nominative –while the other cases that
we have seen so far, and that we will see later, are potentially related to accusative or dative
cases, as we saw above–.
However, it is less clear whether impersonal SE blocks accusative assignment to the direct
object or not. Analyses that propose that it might block case in the direct object are ultimately
attempts to unify impersonal and passive SE in the same group: passive SE clearly absorbs
accusative case from the verb –a property that is perhaps shared by anticausative SE, together
with antipassive SE–, so if impersonal SE can be argued to so something similar, a unification
between the two types of SE is possible.
The discussion goes back to Dobrovie-Sorin (1998). We pointed out in §7.4 above that the
theory that treats SE as the head of T predicts that prodrop languages might have impersonal
SE. Dobrovie-Sorin (1998) notes that this prediction is not correct for Romanian: Romanian is
a prodrop language that lacks impersonal SE. How is this problem solved? The idea put forward
is that being a prodrop language is a necessary but not sufficient condition for having
impersonal SE. In order to develop an impersonal SE use in the language, the language must
allow a nominative SE; languages like French or Romanian only have an accusative SE, and
therefore cannot use SE in a subject position –or coindexed with a nominative subject–. In her
analysis, passive SE absorbs accusative case from the verb, not allowing the licensing of an
accusative object; impersonal SE minimally differs from the passive in that it absorbs
nominative case, but leaves accusative case available. Romanian only has a SE that absorbs
accusative case, and therefore can only make passive SE structures, never impersonal SE
structures. Spanish would have both, according to this analysis.
The next logical step in this discussion is to question whether there are really two SE
structures that differ on which case SE absorbs. Interestingly, Dobrovie-Sorin (2021) proposes
that the passive SE represents a situation where SE is introduced in T and there is no case-
absorption distinction, but rather one in terms of the interpretation of the subject, as we will
see in §8.3 below. This might be one step into the unification of the passive SE to an impersonal
SE.
However, here we are concerned with analyses that do not locate passive SE in T, but rather
that argue that impersonal SE is located lower than T, because what we want to check here is
whether impersonal SE can be assimilated to passive SE, not if passive SE can be treated as an
impersonal SE. The clearest analysis where impersonal SE is argued to be placed in a position
and with a role similar to the one of passive SE is Ordóñez & Treviño (2016), where they argue

130
 SE IN SPANISH

that SE is placed in a verbal head and is related to a structure where SE does not absorb
accusative case and the agreement that is seen in (351, 352) is due to the availability of the
number feature in T.
Ordóñez & Treviño (2016) propose that there are three ways to license the case of the
internal argument in Spanish, which eventually produce the difference between active
sentences, impersonal SE and passive SE:

a) 'Normal' accusative, which is manifested through the clitics lo / la

b) Differential Object Marking accusative, which is assigned by an Agr layer between v and
V
c) Exceptional case assignment through partial agreement with T in number

In this analysis, the presence of DOM does not signal accusative case, but rather an inherent
case that is assigned to human and referential objects by an additional Agr head internal to the
verbal complex. The specific configuration is the one in (353), where movement to spec, AgrP
is not necessary for all objects.

(353) vP

v AgrP

(DP) Agr

Agr VP

V DP (DOM)

AgrP is deployed only if the DP in VP contains some specific referentiality and animacy
properties. Otherwise, AgrP is not deployed, and the object receives 'normal' accusative from
the vP.

(354) vP

v VP

V DP (not DOM)

What Ordóñez & Treviño (2016) propose is that in both impersonal and passive SE
structures, SE marks the v head and is coindexed with a pronoun that only contains person, not
number –from where the indefinite reading follows–. In this version of the analysis, SE
combines with a v head that cannot assign accusative head.

(355) vP

proi v

v ...
sei

131
 ANTONIO FÁBREGAS

The difference between impersonal and passive SE basically depends on the presence vs.
absence of AgrP, and therefore on the possibility of licensing the internal argument as DOM
or not. If the object has the right referentiality / animacy requisites, AgrP is projected and the
derivation is as in (356). Note that these authors assume an arbitrary pro in the relevant spec
position (VoiceP in Ordóñez & Treviño 2011, vP in this version; the distinction is orthogonal
to our purposes) both in the impersonal SE and the passive SE.

(356) vP

proi v

v AgrP
sei
Agr VP

V DP (DOM)

Subsequent movement operations move SE and the pro argument higher in the clause.
Importantly, when pro goes to spec, TP, it checks the person feature of T and forces third person
agreeement.

(357) TP

proi T

T vP
sei +V
pro v

v AgrP
se
Agr VP

V DP (DOM)

The pro that combines with SE in these constructions does not check number in T, so that
feature is left for potential agreement by another DP. In many varieties, the presence of DOM
makes it impossible that the DOM-marked object satisfies the number feature of T (but see
§8.2 below).
In contrast, in the passive SE the object DP does not have the relevant features to deploy
AgrP above, and DOM is not an option.

(358) vP

v VP
se
V DP (not DOM)

When pro moves to spec, TP, it checks the person feature, leaving –as in the case of
impersonal SE– the number feature unchecked. This time, however, in most varieties the object

132
 SE IN SPANISH

DP which did not get DOM, checks the number feature of T, triggering agreement (but, again,
see §8.2 below). This agreement would in Ordóñez & Treviño (2016) mean that the internal
argument gets nominative assigned, but see again §8.2 for a reevaluation of the proposal made
in Ordóñez (2021).

(359) TP

proi T

T vP
sei +V
pro v

v VP
se
V DP (not DOM)

The consequence is that passive SE will be able to display number agreement with the verb,
but not person agreement, a fact that we will return to in §8.3 below.
Thus, the intuition of this analysis is that impersonal SE is simply passive SE with an
additional layer that assigns DOM to the object –or of course, alternatively, passive SE when
there is no internal argument to license through case–. Ormazabal & Romero (2019) also
propose a unification between the two types of SE which is related to DOM, only that in their
system they have a different take on the case assignment possibilities:

i) A-marked objects (DOM), which receive case from vP –not affected by the presence of
SE–.
ii) Non marked objects, which they propose simply do not require case.

The point of the derivation that differentiates impersonal SE from passive SE is represented
in (360), where (360a) is the impersonal SE and (360b) is the passive SE. Note that the only
difference is a-marking, and SE is simply placed in the external argument position and does
not change the case-assigning possibilities of vP.

(360) a. vP

DP v
(DOM)
SE v

v VP

V DP

133
 ANTONIO FÁBREGAS

b. vP

SE v

v VP

V DP

The rest of the derivation is not very different from Ordóñez & Treviño (2016): SE moves
to TP, where it only checks the person feature. The number feature is available to establish a
relation with a DP, but this only happens in the passive SE structure, as that DP has not got
case.

7.5.2. Leísmo in impersonal SE structures

An important ingredient of Ordóñez & Treviño's (2011, 2016) approach –and others that try
to treat impersonal SE as a passive SE and assume that accusative is generally assigned to
objects– is the proposal that the presence of DOM tends to correlate with the non-agreeing
cases which produce the surface impression of impersonal SE. At the same time, we have seen
that Dobrovie-Sorin (1998) argues that impersonal SE absorbs nominative case while passive
SE absorbs accusative case, a point denied by the analyses that unify the two SE-types. From
the perspective of the unification, then, it should be in principle surprising that passive (which
generally is assumed to absorb accusative) does not allow accusative clitics, while impersonals
should allow it.
Ordóñez & Treviño (2016) never claim that SE absorbs accusative, but they do claim (2016:
248) that the case assigned to the internal argument is not accusative: the number agreement
that characterises passive SE structures is in fact agreement with an object which correlates
with a partial nominative. If SE does not absorb accusative case and pro is a subject in TP, for
them it is in principle problematic that the internal argument in passive SE structures cannot be
substituted with accusative clitics (*Se lo venden 'SE it.acc sell.3pl', intended as 'That is sold
to them'). They suggest in that article that the presence of an intervening clitic blocks the
agreement relation between T and the argument, although note that only accusative clitics
should block it (as dative clitics are allowed with SE passives without causing trouble for
agreement). One can argue, however, that the impossibility of using the accusative clitic in fact
comes from the object agreeing partially with the T node, which approaches that clitic to a
nominative. Perhaps the combination of being the object of the verb but still agreeing in part
with T make the use of the clitic impossible.
Seeing now the problem from the opposite perspective, the second side of the problem is
that if passive SE does not allow accusative clitics, one should expect that impersonal SE
should not allow accusative clitics either, if they are both the same type of configuration. What
passive SE structures allow is dative clitics in addition to the internal argument: Se le venden
pisos 'SE him.dat sell.3pl houses, Houses are sold to him'. Thus, what one expects is that
impersonal SE allows dative clitics and rejects accusative clitics. In principle, there are facts
that support this conclusion.
Standard Spanish –whatever that means– is analysed as having a partial case system in
pronominals, with the clitics lo / la marking accusative case, and le marking dative case. The
phenomenon of leísmo describes the situation where speakers use the etymologically dative
clitic in contexts where one expects accusative assignment, as in (361a), where (361b) should
be the normative form.

134
 SE IN SPANISH

(361) a. Le vi.
him.dat saw
'I saw him'
b. Lo vi.
him.acc saw
'I saw him'

Leísmo is in principle geographically and socially conditioned. The relevance of this

phenomenon for impersonal SE is that the presence of this type of SE is known to trigger
'leísmo' in speakers that generally use (361b), not (361a) (Bello 1847, Cuervo 1874, Fernández
Ramírez 1964, Marcos Marín 1978, Martín Zorraquino 1978, Mendikoetxea & Battye 1990,
Fernández-Ordóñez 1999, Ordóñez & Treviño 2016, MacDonald & Melgares 2021).

(362) Se le vio.
SE him.dat saw
'Someone saw him'

Speakers differ with respect to how grammatical it is to use the accusative pronouns in the
presence of impersonal SE. MacDonald & Melgares (2021) report that (343) is ungrammatical
in Honduran Spanish and Ordóñez & Treviño make a similar claim about Mexican Spanish,
but other varieties seem to at least accept it –particularly with feminine objects–.

(363) Se la vio.
SE her.acc saw
'Someone saw her'

The facts in principle support the view where impersonal SE, like passive SE, blocks normal
accusative clitics –assuming of course that DOM is different from the case related to normal
accusative clitics, as in Ordóñez & Treviño's analysis impersonals assign DOM–. Dative case,
as represented in leísmo, would be an effect of accusative clitics not being strictly available in
the impersonal construction; Spanish DOM, where the direct object is marked with the dative
preposition (Fábregas 2013), would partially merge with dative, a fact that is more strongly
made in modern times by Ormazabal & Romero (2013), but that goes back to Cuervo (1874).
There are more facts related to the variation between the number agreeing and non agreeing
construction, but we will revise those in §8.2, as they are informative about further details of
the passive SE construction.
However, the data do not support the conclusion as strongly as it might seem at the
beginning: as we say, several varieties allow (363) and the restriction seems to be stronger for
the masculine form lo. Of course, this might reflect just a fact about the feature specification
of each clitic in the relevant variety –for instance, lo might be marked for accusative case
compulsorily while la might be marked for feminine gender, and can lose the accusative case
specification without being replaced by le–, but any approach based on the feature specification
of the pronouns makes prediction for the further uses of the clitic in other contexts, and it seems
that more empirical research is needed across varieties to confirm or deny this proposal –note,
in fact, that Ordóñez & Treviño (2016) are very careful to propose a specific analysis of
Mexican Spanish, and propose a partial continuum between varieties in terms of –.
However, if the solution to the problem must put into the mix the feature specification of
each individual clitic in order to explain the variation, other options emerge. MacDonald &
Melgares (2021), for instance, accept that impersonal SE imposes a restriction into the available
clitics but do not relate that restriction to case, but rather to a definiteness feature –at least in

135
 ANTONIO FÁBREGAS

the Honduran variety–. The definiteness feature characterises accusative pronouns in this
variety, and in combination with the impersonal clitic SE, under adjacency, the feature
specification of the accusative clitic becomes impoverished as part of a PF rule. See also, for
an alternative interpretation of these facts, Mendikoetxea & Battye (1990), who argue that the
leísmo that emerges in impersonal contexts is due to an intervention effect: the accusative
clitics intervene between SE and the verb and block the formal relation that they need to
establish with each other; leísmo would be the repair strategy to avoid that intervention effect.
Thus, to conclude this section: from the perspective of case, the unification between
impersonal and passive SE encounters one problem, namely that one would expect the same
case assigning possibilities in both impersonal and passive SE. If SE never absorbs accusative
case, the problem is that passive SE structures should be allowed to combine with object clitics.
The problem can be solved, however, if in fact accusative clitics –different from DOM clitics–
are unavailable also in impersonal SE structures, and this can at least be argued with the
available facts.

7.6. Relation between impersonal SE and the predicate-affecting SE uses

As we have seen, impersonal SE stands out within the system of SE uses as being the only
one that is related to a nominative position. Despite the variety of analyses, all of them share
the intuition that SE appears in a subject position or is coindexed with a pronoun that is in a
subject position.
This means a contrast with reflexive and reciprocal uses, where SE is related to accusative
or dative environments; to anticausative uses, where the subject is the entity affected by the
change; to aspectual uses, where it has been argued that SE is either a dative coindexed with
the affected subject or a marker within the internal vP complex; with transitive and antipassive
uses, where SE is analysed plainly as a reflexive object. However, depending on the analysis,
agentive or factitive SE might be similar to impersonal sense in the sense that SE imposes some
type of restriction to the category that plays the role of the subject: in the case of these two SE
uses, the restriction has to do with the thematic interpretation of the subject, while in the case
of impersonal SE it has to do with its reference. However, one could attempt a reduction of
agentive and impersonal SE uses where in both cases SE is coindexed with the subject and the
difference has to do solely with whether the subject is a pro with a particular interpretation or,
alternatively, with a full DP. Assuming that SE impoverishes T in impersonal cases, and that
has something to do in not allowing overt nominal subjects, one could perhaps propose that
agentive SE comes from the same basic source (a configuration like 364) but in the case of the
agentive SE it does not move up to T, therefore not impoverishing it and allowing an overt
subject with definite reference.

(364) vP

DP / pro v

v VP
SE
V ...

136
 SE IN SPANISH

(365) a. TP Impersonal (movement of SE to T, impoverishment)

(no overt subject)
pro T

T vP
se
pro v

v VP
se

b. TP Agentive (no movement of SE to T, no impoverishment)

(possible overt subject)
DP/pro T

T vP

DP/pro v

v VP
se

Let us now move to the second clause-affecting SE use, passive SE.

8. Clause-affecting SE uses (2): passive SE

Passive SE structures, which have already been briefly introduced, are illustrated in (366).
Our starting point for the discussion is the fact, already mentioned in §7.5 above, that passive
SE structures –in contrast to periphrastic passives, as we will see in §8.3– cannot be built from
active sentences where the direct object must receive DOM.

(366) a. Vendí muchas motos.

I.sold many motorcycles
b. Se vendieron muchas motos.
SE sold many motorcycles
(367) a. Vendí a mis hijos.
I.sold DOM my sons
b. *Se vendieron mis hijos.
SE sold.3pl my sons

Personal pronouns of first and second person must always be introduced by DOM when
used as objects. This implies that passive SE structures can never be formed with subjects of
first and second person.

(368) a. Me vio a mí.

me saw DOM me
b. *Se vi.
SE saw.1sg
Intended: 'I was seen by someone'
c. #Me vi.
me saw.1sg

137
 ANTONIO FÁBREGAS

(Only interpretable as a reflexive)

This, just like in the case of impersonal SE, means that the non-paradigmatic nature of
passive SE, which cannot alternate with 1st or 2nd person pronouns (unlike the predicate-
affecting uses), does not need to be postulated, but can actually be derived from a syntactic
restriction that only allows some third person nominals to be the argument of passive SE
constructions.
Passive SE constructions have been examined in many different works, among them
Fernández Ramírez (1957 [1986]), Hatcher (1957), Contreras (1973), Molina Redondo (1974),
Manteca (1976), De Mello (1978), Havertake (1980), Cano Aguilar (1981), Trujillo (1988),
Campos (1989), De Kock & Gómez Molina (1990), Hernández Sacristán (1991), De Miguel
(1992), Gómez Torrego (1992), Devís (1993), Raposo & Uriagereka (1996), Omori (1997),
Ricós (1998), Mendikoetxea (1999, 2008, 2012), Saab (2014), MacDonald & Maddox (2018)
and Dobrovie-Sorin (2021), as well as the authors cited in §7.5 above and who have tried to
unify passive and impersonal SE.
Empirically, there are three main distinctions to make in the domain of passive SE: the
nature of the semantic subject, which contrasts them with anticausative SE structures, the
nature of the agreeing argument, which contrasts them with impersonal SE through complex
patterns of variation, and the comparison with periphrastic passives, which have been crucial
in determining the right structure of passive SE structures. Beyond this, in this overview we
will present the connections that passive SE has with the predicate-affecting SE uses.

8.1. Main properties of passive SE: differences with anticausative SE

Passive SE structures can be characterised, coarsely, by three properties that are however
subject to qualifications. The first one is that passive SE is interpreted as a passive in the sense
that the argument that seems to be a subject is interpreted not as an agent but as an internal
argument that undergoes the event expressed by the verb.

(369) a. Juan encontró muchas cosas en su casa.

Juan found many things at his place
b. Se encontraron muchas cosas en su casa.
SE found.3pl many things at this place
'Many things were found at this place'

Second, the passive SE structure agrees in number (and possibly person, although all passive
SE structures are third person) with the argument interpreted as the object, semantically.

(370) a. Se encontró una cosa.

SE found.3sg one thing
b. Se encontraron muchas cosas.
SE found.3pl many things

Third, passive SE is considered passive also because it allows the introduction of a by-
phrase that expresses the notional subject of the event, the agent.

(371) Se decidió eso por las autoridades.

SE decided that by the authorities
'That was decided by the authorities'

138
 SE IN SPANISH

As we say, all these properties can be questioned, but they are the ones that are mainly taken
to differentiate impersonal SE from passive SE and, to some extent, from anticausative SE.
With respect to the distinction between passive and anticausative SE and how it is empirically
grounded, consider (372).

(372) Se rompieron las ventanas.

SE broke.3pl the windows

In principle, (372) can be interpreted both as an anticausative SE and as a passive SE,

although with a tendency towards the passive interpretation. The anticausative interpretation
would be 'The windows broke', where the event is triggered either by the internal properties of
the windows or by an accident not controlled by any external force; there is no external agent
involved. The passive interpretation is 'The windows were broken', where one interprets that
an external agent that is not specified is responsible for the breaking of the windows.
As seen in (372), the position of the agreeing argument favours one or the other reading.
Passive SE structures tend to be thetic judgements (Ricós 1998) which favour that the subject
appears postverbally. The preverbal position (373) would favour in this case the anticausative
reading, as the anticausative makes an assertion about what happened to an individual.

(373) Las ventanas se rompieron.

the windows SE broke.3pl

As we saw in (374), the anticausative construal can combine with causes, but not with
agents. The passive SE structure allows, in principle, markers that involve the presence of an
agent.

(374) a. Las ventanas se rompieron {por el calor / ??por los niños}.

the windows SE broke by the head / by the children
b. Las ventanas se rompieron a causa del calor.
the windows SE broke because of.the heat
(375) a. Se rompieron las ventanas deliberadamente.
SE broke the windows deliberately
b. Se rompieron las ventanas con alevosía.
SE broke the windows with premeditation

Similarly, the anticausative SE structure is not compatible with markers of the presence of
an agent, but with markers that show that there was no external agent, beyond the entity that
itself undergoes the change (§4.2).

(376) Las ventanas se rompieron {solas / por sí mismas}.

the windows SE broke alone / by themselves

In terms of the syntactic projection of the internal argument, remember that –despite acting
as unaccusatives in their semantic interpretation and in some languages also for the choice of
auxiliary– anticausative SE does not allow bare nominal subjects (§4.1). This is in contrast to
passive SE (Sánchez López 2002).

(377) Se rompieron *(las) ventanas por sí solas.

SE broke the windows by themselves

139
 ANTONIO FÁBREGAS

This fact of anticausative structures is surprising given the general characterisation of

anticausative SE as an unaccusative construction (remember §4.1, where we however pointed
out that it is unclear to what extent anticausative SE structures in Spanish should be related to
unaccusatives, given the readings of absolute participles and the lack of bare nominal subjects).
Remember that unaccusatives normally allow postnominal bare nominals as subjects.

(378) Desaparecieron ventanas.

disappeared windows
'Some windows disappeared'

This is not the only fact that suggests that anticausative SE and passive SE should contrast
in terms of the position of the subject. Anticausative SE and passive SE share the property that
the morphology of the verb is active even if the subject is interpreted as the internal argument
that undergoes the change. This fact is highlighted in Ormazabal & Romero (2019), who notice
that periphrastic passives break the idiomatic interpretation of some fixed expressions, while
passive SE does not.

(379) a. echar margaritas a los cerdos

to.throw daisies to the pigs 'to waste something good with an underserving person'
b. #Las margaritas fueron echadas a los cerdos.
the daisies were thrown to the pigs
c. Se echaron margaritas a los cerdos.
SE threw daisies to the pigs
(380) a. quemar puentes
to.burn bridges 'to break cooperation oportunities'
b. #Los puentes fueron quemados.
the bridges were burnt
c. Se quemaron los puentes.
SE burnt the bridges
'All cooperation oportunities were broken'

This suggests that the passive SE keeps the idiomatic structure, perhaps leaving the internal
argument within the lexical verb, where the domain of idioms is defined (Marantz 1984).
Interestingly the anticausative SE breaks the idiomatic meaning:

(381) #Los puentes se quemaron por el odio.

the bridges SE burned by the hate
Intended: 'The cooperation oportunities broke because of hate'

All these facts are relevant for the question of how the anticausative and the passive SE
structures should differ from each other. If we go back to §4 above, remember that the main
analyses of anticausative structures discussed whether anticausative SE should be a reflexive
(Koontz-Garboden 2009) or it should be a structure where voice is impoverished and SE marks
the impoverished voice head (Schäfer 2008) or is located as an expletive external argument
(Alexiadou, Anagnostopoulou & Schäfer 2015). The analysis that treats anticausatives as
unaccusatives, in principle, supports the analyses where the external argument position is
removed in anticausatives. However, as the characterisation of anticausatives (§4.1) or even
reflexives (§3.1) as unaccusatives is problematic, one seems to require movement of the
internal argument to a specifier position at some point in the derivation. This movement would
have three consequences.

140
 SE IN SPANISH

a) To ban bare nominals, which by hypothesis are only licensed in complement positions
b) To favour an ordering where the subject is preverbal
c) To break the idiomatic meaning of fixed expressions, by removing the internal argument
that is in the complement position from the domain of the lexical verb

Once the internal argument in anticausatives has to move to a specifier position, that
movement is in principle compatible both with the analysis of anticausative SE as reflexive SE,
preserving the external argument, or not –but remember the arguments against the reflexive
analysis of anticausative SE in §4.3–. In contrast, the internal argument in passive SE structures
should stay in the complement position, (a) allowing bare nominals, (b) favouring a thetic
ordering with the subject in postverbal position and (c) preserving idiomatic meaning. How
does that obtain?
If one assumes Schäfer (2008), where Spanish SE is located in the head of a defective Voice
that lacks an external argument, the distinction could be operationalised as follows: SE in
anticausatives and in passives is located in the head position of Voice, in both cases –Ordóñez
& Treviño (2011) locate passive SE in Voice, but remember that in (2016) the head is v–. In
the anticausative construal, the internal argument moves to spec, VoiceP to bind the SE (382a);
in passive SE, the specifier of Voice is filled by a pro argument that gets assigned case (382b).

(382) a. VoiceP

DPi Voice

Voice ...VP
sei
V DP

b. VoiceP

proi Voice

Voice ...VP
sei
V DP

Of course, one would have to retract from the strong position taken by Schäfer (2008) where
SE marks a specific Voice flavour: one would have to treat SE rather as a reflexive pronoun
that can be added to the voice head and must be licensed by a nominal expression that binds it.
That binding can be performed in two ways: either by movement of the internal argument, in
which case one gets an anticausative construal, or by insertion of a pronominal, which results
in the passive construal (and the impersonal construal if DOM is assigned to the internal
argument).
The analyses where anticausative SE is located in the specifier position would be
disfavoured (Alexiadou, Anagnostopoulou & Schäfer 2015) from this perspective, but it is not
impossible to account for the difference with them: the internal argument in the unaccusative
construal would have to move to a higher specifier position; the biggest complication of the
approach is that the pro that corresponds to the agent would have no obvious insertion point,
which makes the unification between the passive and the anticausative derivations less obvious.

141
 ANTONIO FÁBREGAS

Similarly, this disfavours Koontz-Garboden's (2009) reflexive analysis, for the same reasons:
the presence vs. absence of an agent is not clearly accounted for here.
Note that a solution to passive SE along the lines of Ormazabal and Romero (2019) is also
able to explain the distribution of bare nominals. In their analysis of passive (and impersonal)
SE, the SE element is a subject pronoun that occupies the position of spec, vP –with case
determining the difference between the impersonal and the passive–. Starting from a derivation
like (383), one could in principle propose that passive SE is located in the specifier of Voice,
making the internal argument remain in place (383a); from there SE moves to the subject
position in spec, TP. In the anticausative SE construction one would require the internal
argument to move to a specifier position, perhaps the subject position (383b).

(383) a. TP

SE T

T vP

SE v

v VP

V DP
b. TP

DP T

T vP

SE v

v VP

V DP

The problem that (383b) encounters is what makes it possible for the DP to cross over SE
to land in TP: in principle, if SE is a normal subject pronoun, SE should block movement of
the internal argument to TP in the same way that the external argument of any transitive verb
blocks movement of the internal argument to the subject position. Note that it is not
straightforward to claim that movement of the DP is necessary to bind SE, because in principle
nobody is binding SE in (383a). One minimal alternative could be to propose that in
anticausatives, SE is in the head position and does not block movement, but then the parallelism
between passive and anticausative dissolves.
Thus, the facts related to the distribution of bare nominals constitute an argument that
favours approaches where SE is located in the head of vP or VoiceP. Crucially, from this
perspective (382) the difference between the anticausative and the passive is that the passive
introduces a pro category that is interpreted as the indefinite subject.
The approach that we have just presented is an attempt to unify anticausative and passive
SE structures using the same type of SE in both cases –the next subsection will revisit the
unification of passive and impersonal SE from the perspective of the agent and the variation
present in the agreement patterns with the internal argument–. However, there are other

142
 SE IN SPANISH

accounts of the distinction between anticausative and passive SE that try to use the same
ingredients.
Kempchinsky (2004, 2006) present a solution based on her proposal that SE is an aspectual
element that introduces, saturates or links to a subevent of the verb (remember §5.2 above). In
her analysis, passive SE structures involve the configuration in (384; Kempchinsky 2006).

(384) vP

se v

v AspP

Asp VP

V DP

In this approach, passive SE is located in a specifier position of vP. This projection

introduces the first subevent, related to the agent; SE links to that subevent in this position, not
removing the causation component –thus, leaving the interpretation that there is an agent, but
not letting it be projected as a normal nominal expression–. In contrast, the anticausative SE,
which Kempchinsky (2004) labels inchoative / ergative, corresponds to the structure in (385).

(385) VoiceP

Voice AspP
se
DP Asp

Asp VP
se
V DP

Note that vP is missing from (385), and therefore also the subevent related to the causation
of the event. SE is projected as the head of Asp, where it introduces a change-of-state subevent
as the definition of the second subevent of the event. The absence of vP is explained as follows:
SE introduces a subevent that already licenses the eventuality and defines an argument linked
to it, through movement. SE then carries all the necessary event information, and movement to
VoiceP defines a complete event without the need to introduce vP. Thus, Kempchinsky's
account can also explain why the anticausative internal argument appears in a specifier
position, while the passive one remains in place, and the distinction between anticausative and
passive is explained by the absence or presence of a causation subevent introduced by vP.
A third account of the distinction is Dobrovie-Sorin (2021). In her account, the difference
between anticausative and passive SE is double. (386) presents the structure for the active,
where the author assumes that T licenses the external argument (introduced in spec, VoiceP)
and Voice licenses the internal argument.

143
 ANTONIO FÁBREGAS

(386) TP

T VoiceP

DP Voice

Voice VP

V DP

Anticausatives and passives share the property that Voice is devoid of unvalued features,
which make it inert: it does not introduce an external argument and it does not license the
internal argument. In the case of Romance SE, SE is introduced as a clitic directly in T, where
it probes the internal argument. The difference between anticausative and passive SE depends
on the phi feature structure of SE. In the anticausative, a plain version of SE is introduced:
nothing in (387) defines a subject.

(387) TP

T VoiceP
se
Voice{ø} VP

V DP

In contrast, in the passive version, a SE containing an arbitrary interpretation is introduced.

Dobrovie-Sorin (2021) proposes that the initiator / agent of an event can also be introduced in
TP, and claims that because of this technical innovation, the arbitrary interpretation of SE
allows to infer the presence of an external agent, although with an indefinite reference.

(388) TP

T VoiceP
se-arb
Voice{ø} VP

V DP

Once treated like this, the analysis of anticausative and passive SE becomes close to the
analysis of impersonal SE in the approaches where the impersonal is located in T (Otero 1986).
The distinction between the passive and the impersonal could derive from a case factor, as in
Ordóñez & Treviño (2011, 2016) or Ormazabal & Romero (2019). The complication that
Dobrovie-Sorin (2021) has to encounter, however, is that she needs to make available a second
position in the tree for the agent interpretation, TP, a position that is generally not related to
this interpretation to the extent that other thematic interpretations can be located in the subject
position. Note, interestingly, that in this approach SE does not mark the impoverishment of the
vP or VoiceP layer.
This takes us back to the problem of how the passive and the impersonal SE structures relate
to each other. The next section takes up this question again.

144
 SE IN SPANISH

8.2. More on passive SE and impersonal SE: more details about the distribution of the internal
argument
In §7.5 above we already discussed some approaches that argued that passive and
impersonal SE structures share a basic configuration, treating their difference as based on the
case assigned to the internal argument. There are, however, three empirical factors that should
be discussed from the perspective of passive SE and that might complicate the unification:

i) The position of the clitic SE within periphrastic structures

ii) The problem of the external argument: if passive and impersonal SE share a
configuration, why can one add an agent complement in the passive but not in the impersonal?
iii) The problem of agreement: the unification proposes that agreement with the internal
argument in passive SE structures correlates with configurations where the internal argument
did not receive case from another head. However, the empirical picture is more complex than
this initially suggests.

Let us start with the problem of the position of the clitic SE in periphrastic constructions.
This is probably the most problematic property from the perspective of the unification of
impersonal and passive SE structures. The observation is the following: in combination with
auxiliaries, passive SE can remain together with the infinitive or gerund (389a, 389d), while
impersonal SE must appear before the auxiliary (389b, 389e)

(389) a. No pueden hacerse estas cosas.

not can to.do-SE these things
'These things cannot be done'
b. ??No puede invitarse a estas personas.
not can to.invite DOM these people
c. No se puede invitar a estas personas.
not SE can to.invite DOM these people
'One cannot invite such people'
d. No están haciéndo-se estas cosas.
not are doing-SE these things
'These things are not being done'
e. *No está invitándo-se a estas personas.
not is inviting-SE DOM these people
f. No se está invitando a estas personas.
not SE is inviting DOM these people
'One is not inviting such people'

Note that there is no contrast between impersonal and passive SE when it comes to the
distribution in control infinitives: just as impersonal SE cannot appear in these contexts
(remember §7.4), passive structures cannot with SE or not: *Juan prometió hacerse esas cosas
'Juan promised to do-SE those things' ~ *Juan prometió ser hechas esas cosas 'Juan promised
to.be done these things'. The contrasts in (389) suggest that impersonal SE has to raise to a
nominative position: assuming that the auxiliary and the non finite form are within one single
clausal context, and that the T node is either above or at the same height as the auxiliary,
impersonal SE has to move to that node while passive SE can remain low, potentially within
the verbal complex. The control infinitive data seem to suggest that both types of SE require a
nominative position, on the other hand.
One possible way out would be to propose that passive SE does not need to raise to T, while
impersonal SE has to move to that position, even if both of them end up being linked to a

145
 ANTONIO FÁBREGAS

pronoun (or a position) where nominative case has to be assigned. In that sense, the
impoverishing function related to T that SE is associated to in passives would be an effect of
the impoverishing of vP / VoiceP, which forces the presence of a pro with properties that, when
moved to T, trigger the partial agreement.
Let us now move to the second issue, the agent complement of passive SE structures. In
principle, being a passive structure, SE should be compatible with an agent complement –as
we saw in some examples above–, but this is shocking from the perspective of the analyses that
treat impersonal and passive SE as the same type of structure. Passive SE displays verbs in
active forms, not passive forms; by hypothesis passive SE shares the same configuration with
impersonal SE, which does not allow agent complements.

(389) Se vive bien aquí (*por todo el mundo).

SE lives well here by all the world
Intended: 'Everybody lives well here'

In fact, the empirical evidence suggests that the claim that passive SE allows agent
complements (that goes back to Gili Gaya 1943) is not empirically correct. Gili Gaya's (1943)
famous example (390) is not deemed acceptable by most contemporary speakers, who strongly
prefer a periphrastic passive structure to license the agent complement (391).

(390) ??Se firmó la paz por los embajadores.

SE signed the peace by the ambassadors
(391) La paz fue firmada por los embajadores.
the peace was signed by the ambassadors

There are three facts that cast doubt on the claim that passive SE structures combine with
real agent complements, which is precisely what the unification between impersonal and
passive SE expects.

a) In contrast to the agent complement of periphrastic passives, por 'by' –when possible–
can be substituted by por parte de 'on the part of', which in fact sounds more natural than the
short version with these passives (Pujalte 2013, Saab 2014). This expression has a broader
distribution than verbal passives, and is in fact prefered with nominalisations.

(392) a. Se quemaron muchos contenedores por ??(parte de) los manifestantes.

SE burnt.3pl many containers by part of the demonstrators
b. Muchos contenedores fueron quemados por (*parte de) los manifestantes.
many containers were burnt by part of the demonstrators

b) The referential properties of the alleged agent complement are also restricted.
Expressions involving singular referential individuals are generally not natural (393), and
(according to RAE & ASALE 2009: §41.11h-k) the nominal expressions that are more natural
with passive SE structures have a generic interpretation, particularly in plural and expressing
abstract administrative entities which strictly cannot be characterised as animate nouns –unless
personified– (394).

(393) a. Se afirman esas cosas (*por Juan).

SE claim those things by Juan
b. Se propuso hacerlo (*por el anterior director).
SE proposed to.do-it by the former director

146
 SE IN SPANISH

c. Se escribieron muchas normas (*por un administrativo).

SE wrote many rules by an administrator
(394) a. Se afirman esas cosas (?por la opinión pública).
SE claim those things by the public opinion
b. Se propuso hacerlo (?por las autoridades).
SE proposed to.do-it by the authorities
c. Se escribieron muchas normas (?por la nueva administración).
SE wrote many rules by the new administration

The distribution of these nominal expressions is reminiscent to the strict distributional

restrictions about referentiality in other allegedly passive environments which have been
analysed as not involving real passivisation, such as adjectival passives and modal adjectives
(cf. Gehrke 2013, Oltra-Massuet 2014). This casts doubt on the claim that these are real agent
complements; they might be expressions linked to the arbitrary pro subject that characterises
these structures, which (remember §7.5) is a generic or existential pronoun which cannot
denote definite individuals.

c) It is in fact more frequent to have the intended agent of the passive expressed with other
prepositions, among which in the current oral language desde 'from' is the most natural one.
(394) is still degraded, and speakers prefer (395).

(395) a. Se afirman esas cosas (desde la opinión pública).

SE claim those things from the public opinion
b. Se propuso hacerlo (desde las autoridades).
SE proposed to.do-it from the authorities
c. Se escribieron muchas normas (desde la nueva administración).
SE wrote many rules from the new administration

The conclusion seems to be that one cannot treat the PPs above as agent complements in the
same sense as they are treated in periphrastic passives. It might be empirically better, given
their properties, to treat them as adjuncts that semantically specify further a property that is
already satisfied in the verbal structure: the presence of an agent introduced in a thematic
position. The preposition used to introduce them, then, is not an agent one.
What makes passive SE structures special in contrast to other active configurations is that
in them the agent is satisfied by an indefinite expression, a pro or an arbitrary interpretation
introduced by SE; this restricts the types of nominals that sound natural as specification of the
semantic content of that indefinite pronoun to nominals that are also indefinite, generic,
abstract, etc.
Thus, when one considers the details, the properties of the apparent agent complements in
passive SE sentences in fact support the analysis where they are unified with impersonal SE
structures, while the contrasts in the position of the clitic SE with auxiliaries are a problem for
the unification. How about the agreement with the internal argument, is it a property that would
force us to treat impersonal SE and passive SE as two distinct structures or not?
In §7.5 above, we noted that the more typical pattern is between the configurations in (396):
a non-agreeing verb combined with a DOM-marked object, and a number-agreeing verb
combined with an internal argument that would not have received DOM anyways.

(396) a. Se vio a los niños.

SE saw DOM the children
'Someone saw the children'

147
 ANTONIO FÁBREGAS

b. Se vieron niños.
SE saw.3pl children
'Children were seen'

From the perspective of case, one expects that any internal argument marked with a
preposition, not just DOM, will pattern with the configuration in (396a) and only produce what
on the surface looks like an impersonal SE sentence. This is unsurprising too from the
perspective of a more traditional account of passive SE, as in Spanish passives cannot be built
from predicates whose internal argument is a PP. In general, the pattern in (396) is also
followed by the prepositional verbs (397).

(397) Se confió en ellos.

SE trusted in them
'Someone trusted them'

However, the patterns are not so clear once variation is put into the mix. The pattern is
violated in two senses:

i) Non-agreeing verbs with nominals that do not get prepositional marking

ii) Number-agreeing verbs with nominals that carry prepositional marking

The first case is illustrated by (398; Otero 1966, De Mello 1995, Sánchez López 2002,
Ormazabal & Romero 2021):

(398) a. Se vende casas.

SE sells houses
'Houses are sold'
b. Se necesita dentistas.
SE needs dentists
'Dentists are needed'

Though not predominant, the non-agreeing version has been documented in particular in
some American varieties (De Mello 1995; remember Bello 1847: §792). The non-agreeing
version is more natural for speakers that tend to agree when the verb appears in combination
with modal auxiliaries:

(399) a. Se puede vender casas.

SE can sell houses
b. Se pueden vender casas.
SE can.3pl sell houses

There is one way of viewing these facts which does not affect the analysis of SE –in
particular, it does not affect whether impersonal and passive SE can be unified– but that we
would like to point out. The reader might have noticed that all the cases cited in the literature
where there is lack of agreement with an argument that would not receive DOM involve bare
nominals, never nominals combined with quantifiers or determiners. Some impressionistic
evidence suggests that the non-agreeing pattern is much more restricted when the non DOM
argument is not a bare noun; that is, that Se vende casas 'SE sells houses' is accepted to a much
broader extent than Se vende muchas casas 'SE sells many houses' (as Bosque 1999 proposes
in the introduction to the first exercise in his book). If confirmed, this might mean that the non-

148
 SE IN SPANISH

agreeing variety in the absence of DOM involves incorporation or reanalysis of the argument
with the verb, instead of using some other head to check case on the nominal or absence of
case checking in the non agreeing version. However, as we say, this alternative does not make
the unification less likely from the perspective of SE.
Note, in any case, that the pattern in (399) does not alter all the factors of the general
distribution that was presented in §7.5. What (398, 399) show is that objects that in principle
do not get DOM can allow for agreeing and non-agreeing constructions. The opposite situation,
objects that should receive DOM and display agreement, is not encountered as easily. Ordóñez
(2021) shows that sentences like (400) are impossible in the passive interpretation.

(400) #Se vieron los niños.

SE saw the children
Intended: 'The children were seen'

Only in one particular configuration and with one particular interpretation can internal
arguments that should receive DOM appear in agreement configurations (Sánchez López
2002): in preverbal position and receiving a kind interpretation where the noun denotes a class
of entities, not specific individuals, and contrasts that class with other possible classes.

(401) Los dentistas se necesitan, no los políticos.

the dentists SE need.3pl, not the politicians
'Dentists are needed, not politicians'

One could speculate that (401) does not deny the generalisation shown by (400) if one
assumes one of the following two things: (i) DOM does not need to be assigned compulsorily
when the noun denotes a kind –in fact, (402a) is marginally more acceptable than (402b), where
the second expresses individuals– or (ii) within the vP, the internal argument is in fact a bare
nominal, but movement to a preverbal position and focalisation force insertion of an expletive
article that makes it look as expressing a definite expression. In any case, sentences like (401)
have not been treated in the formal literature.

(402) a. ??Necesitamos los dentistas más que los políticos.

we.need the dentists more than the politicians
'We need dentists more than politicians'
b. *Necesitamos los dentistas de ayer más que los dentistas que hay hoy.
we.need the dentists from yesterday more than the dentists that there.are today
'We need the dentists that came yesterday more than the dentists that are here today'

The interpretation of the pattern in (398) leads to a reevaluation of the case approach, or at
least a partial reevaluation of its main claims. In the approaches revised in §7.5, the passive SE
structure emerges when T licenses the case of the internal argument. Otero (1966) claimed
already that the existence of a non-agreeing pattern in fact tells us that the internal argument,
also in the passive SE, carries accusative case. The agreeing cases are instances of T agreeing
with an accusative object. One could think, in fact, that nominative case requires total
agreement with T, in both person and number. In the passive SE structure, there is only partial
agreement with T, in number, and that might not assign nominative case to the subject. Otero
(2002) emphasises this idea, and proposes that the agreeing pattern in the passive SE structure
is simply agreement with an accusative argument. In essence, and going back to his original
ideas from (1966), this means that passive SE does not exist at all: contra Ordóñez & Treviño
(2016) SE does not eliminate accusative case. The basic configuration is the impersonal SE

149
 ANTONIO FÁBREGAS

structure, and the appearance that there is a passive SE structure comes from the cases in which
speakers check the unvalued number feature of T with the accusative argument.
Ordóñez (2021), taking into consideration the facts in (398) but leaving aside the fact in
(401), explains these contrasts within the framework of his previous analyses with Treviño.
While in Ordóñez & Treviño (2016: 248) they explicitly say that the internal argument receives
nominative case from T in passive SE structures, in this revision he argues, following also
Otero (1966), Rigau (1997) and Ormazabal & Romero (2019), that agreement with the internal
argument in the apparent passive SE cases are instances of agreement with an object, and this
does not make the internal argument get nominative case assigned by T. Nominative is assigned
systematically to the pro category in the specifier of TP, and the internal argument still gets the
case of an object assigned. Evidence of this is (i) the impossibility of having nominative
pronouns in passive SE –see §8.3 also for this– and (ii) the availability of the object clitics in
the SE configuration.

(403) a. *Se vimos nosotros.

SE saw.1pl we
Intended: 'We were seen'
b. Se nos vio.
SE us saw
'Someone saw us'

Remember, however, the leísmo facts in §7.5.2, which suggest that one cannot simply claim
that the case assigned to the object in such configurations is plainly accusative, unless one takes
number agreement with the object to block presence of the accusative clitics.
This basically means that what makes passive SE special is not that accusative case is
absorbed by SE, but rather that the pronoun that receives nominative case in TP (pro with the
indefinite interpretation) only checks person, and makes number available for the agreement
with the object, which might already have got case. The agreeing pattern and the non-agreeing
patterns involve simply two possibilities available: to value number in T with an argument that
has number or to leave number unvalued (perhaps à la Preminger 2014). The 'standard' variety
tends to follow closely the two rules in (404).

(404) a. DOM objects never agree with T

b. Accusative objects always agree with T

From this perspective, SE passives involve number agreement with the object, and this
possibility is simply optionally available: it is equally convergent to satisfy the number feature
of T with the object or to leave it unvalued. The increased acceptability of the non-agreeing
cases with auxiliaries might mean that the choice is sensitive to the distance between T and the
object, so that when there is more material between them non-valuation becomes a stronger
option.
Varieties that allow (398) are varieties where (404b) is not part of the grammar: the number
feature of T is left unvalued without consequences.
It seems that there are also varieties where (404a) is relaxed. Ordóñez & Treviño (2016)
document cases like (405) in Mexican and Colombian Spanish; see also Ormazabal & Romero
(2021).

(405) No se pueden nombrar a los embajadores.

not SE can appoint DOM the ambassadors
'The ambassadors cannot be appointed'

150
 SE IN SPANISH

This phenomenon has been noted in American varieties since very early (see Cartagena
1972; Martín Zorraquino 1979). In the Mexican variety, the agreement does not take place with
pronominal objects or when the object is preposed (which requires a clitic). Agreement is
restricted in this variety to DOM objects.

(406) a. *Se les nombraron a ellos.

SE appointed DOM them
b. *A los embajadores se les nombraron.
DOM the ambassadors SE them appointed
c. *Se dieron a ellos.
SE gave to them
Intended: 'Someone gave (something) to them'

In Ordóñez & Treviño (2016) the proposal is the following: the absence of agreement for
the speakers that follow the rules in (404) is motivated by the presence of Agr assigning DOM
in the verbal configuration (cf. §7.5.1); DOM creates a closed domain that does not allow that
T agrees with the DOM object. In the varieties that allow (405), Agr does not close a domain;
provided that there is no clitic intervening between T and the object, agreement can take place.
Movement of the object to a preverbal position signals that Agr has closed the domain, so again
the agreement relation is broken. Dative prepositions (and other kinds of prepositions) are
heads that also close the domain, breaking the possibility of establishing agreement between T
and the object. The facts in (406) follow if one assumes that the presence of clitics also signals
that a syntactic domain has been closed, again making it impossible to extend agreement.
Another proposal, perhaps more elaborated than the previous one, is presented in Ormazabal
& Romero (2021). For them, the problem of the presence or absence of agreement in examples
like (405) cannot be amenable to a syntactic principle –or a morphological one–, because the
plural agreement can be triggered by any linearly adjacent postverbal nominal, even adjuncts
(cf. their documented example Se trabajan los fines de semana, 'SE work.3pl the ends of week',
'One works the weekends'). In a nutshell, their proposal is that these facts should be treated as
processing problems, perhaps related to agreement attraction through linear adjacency: under
certain conditions, syntax and morphology leave one feature unvalued, in this case number,
and whether that value is satisfied with a default or by attraction of some other element is
irrelevant for the grammatical operations, but might trigger effects under processing.
Ormazabal & Romero (2021) also address the apparent complementary distribution between
the presence of so-called object clitics and agreement, not only in marked examples as (405)
but more in general in impersonal vs. passive SE structures. Within their proposal that the same
structure underlies both types of SE, these authors propose that the so-called object clitic and
the verbal agreement are two ways of expressing number features in the predicate. The core of
their proposal is that the clitics that are traditionally called 'accusative clitics' manifest a number
feature within the system of Spanish. When they appear in the configuration that superficially
is called impersonal SE, they are the exponents of the number feature of the verb. When the
verb agrees in number in the so-called passive SE structure, the number feature is present in
the verb. Ormazabal & Romero (2021) propose, then, that there is a 'clitic mutation' operation
that can turn the accusative clitic (an exponent for number) into another exponent for number,
this one one that linearises as a suffix to the verb. In their view, thus, presence or absence of
agreement with the verb cannot be taken to be a property that differentiates impersonal SE and
passive SE, as in both cases we have the same features involved and what changes is just their
surface manifestation.

151
 ANTONIO FÁBREGAS

With respect to the fact in (405) –which one sometimes can see extended in the oral language
to constructions with prepositional complements, although in such cases speakers tend to
perceive them as lapsus linguae– there are not many analyses that are developed enough to
allow a serious evaluation, beyond the ones just presented. Ordóñez' (2021) treatment of the
passive SE as agreement between T and an accusative object –contra Ordóñez & Treviño
(2016)– might simply adapt this view where agreement of T-number with the object is available
provided nothing blocks it, and accept the distinction between varieties where the head that
assigns DOM blocks the agreement relation and the varieties where it is transparent for that
relation. Ultimately, the proposal in Ormazabal & Romero (2021) involves treating the
variation that one finds in impersonal / passive SE structures with respect to number agreement
in the verb as an extragrammatical operation related to language processing.
To conclude, the variation phenomena that have been examined here in fact support a
unification between the passive and the impersonal SE: once one accepts that DOM is assigned
as a case different from accusative and that DOM cannot be dropped if the internal argument
meets the right requisites, agreement between number in T and number in the internal argument
is just a case of agreement with the object, as Otero (1966) originally proposed. The only
contrast refers to the position of impersonal SE within periphrases, which contrasts with the
options displayed by passive SE, although one could imagine that impersonal SE must move
to a position that is higher –perhaps T itself– than passive SE.
With respect to accounts where the internal argument in passive SE receives nominative,
both impersonal and passive SE assign nominative to the external argument, which is an
implicit indefinite pronoun, and the agreement possibilities are modulated only by (i) whether
in the particular variety there is a head that blocks agreement, as might be the case with the
head that assigns DOM and (ii) whether the internal argument occupies a position that allows
the agreement or not.

8.3. Distinction between passive SE structures and periphrastic passives

In the analyses above, where impersonal and passive SE share one structure, passive SE is
not passive. The consequence of this is that the only passive structure in Spanish is the
periphrastic passive in (407).

(407) Los libros fueron vendidos en esta tienda.

the books were sold in this shop
'The books were sold in this shop'

Consequently, one expects to find sharp differences between the periphrastic passive and
the SE passive, which in fact are found. We have already mentioned a few of them: (i)
periphrastic passives break idioms, but SE passives don't (Ormazabal & Romero 2019); (ii) the
internal argument in SE passives can be a bare nominal, but this is less frequent in the
periphrastic passive, although also grammatical (408, Sánchez López 2002); (iii) periphrastic
passives allow agent complements and SE passives do not combine with real agent
complements (Pujalte 2013); (iv) SE passives tend to be thetic statements where the internal
argument is postverbal, while periphrastic passives are cases where the internal argument is the
subject of the predication, hence preverbal (409).

(408) a. Se vendieron pisos.

SE sell.3pl houses
b. ?Fueron vendidos pisos.
were sold houses
'Houses were sold'

152
 SE IN SPANISH

(409) a. ??Los pisos se venden.

the houses SE sell.3pl
Intended: 'The houses are sold' (but easier to interpret as a middle, cf. §9)
b. Los pisos fueron vendidos.
the houses were sold

All these facts can be accounted for in a system where impersonal SE has a subject pro
which the periphrastic passive lacks. That subject pro allows the internal argument to stay
within the verbal complex, as we overviewed in §8.1 above in the comparison with
anticausatives. Periphrastic passives would then involve structures where the internal argument
might occupy the subject position and leave the vP; the external argument position is not
occupied by a pro and therefore can be linked to a real agent complement. The idiomatic
meaning might break if (as some have argued, cf. Dobrovie-Sorin 2021 as an example) the vP
layer is missing from periphrastic passives, thus making it impossible to store it as a vP idiom.
There are however three other differences that we have not mentioned so far and that require
further discussion. Let us start with the fact already mentioned in passing that passive SE
structures cannot allow personal subjects of 1st or 2nd person; periphrastic passives can do it.

(410) a. #Se / Me vi yo.

SE me saw.1sg I
Intended: 'I was seen'
b. Yo fui visto.
I was.1sg seen
(411) a. #Se / Os visteis vosotros.
SE you saw.2pl you
Intended: 'You all were seen'
b. Vosotros fuisteis vistos.
you were.2pl seen
'You all were seen'

This restriction has been interpreted in two ways: (i) the already presented explanation of
DOM: the personal pronouns of 1st and 2nd person force DOM, and then one gets an
impersonal SE construction (Ordóñez & Treviño 2011, 2016; Ormazabal & Romero 2019,
Ordóñez 2021); (ii) an instance of a Person Case Constraint, that is, cases where the presence
of a particular nominal pronoun in one function prevent the appearance of a person-marked
pronoun in another function. In this case, the characterisation of (410, 411) would be that the
presence of the SE structure in fact inhabilitates a person marked pronoun from being licensed
in the clause. Dobrovie-Sorin (2021) has argued for this type of approach.
Remember from §8.1 above that in Dobrovie-Sorin (2021) (412) corresponds to the
structure of a passive SE structure.

(412) TP

T VoiceP
se-arb
Voice{ø} VP

V DP

153
 ANTONIO FÁBREGAS

In (412), the presence of SE in T, with its arbitrary specification, values the person feature
in T and makes it unavailable for other agreement relations. Therefore, if a person-marked
pronoun has to be licensed by T, it better not carry person features, because they would not be
licensed by the T head. (Note that this same explanation, only that proposing the proarb located
in spec, TP values the person feature, could be assumed in Ordóñez & Treviño 2016).
In contrast, (413) represents the structure of a periphrastic passive according to Dobrovie-
Sorin (2021). Note that the arbitrary interpretation is now restricted to Voice. This has the effect
that the external argument of Voice is interpreted as indefinite.

(413) TP

T beP

be VoiceP

DP Voice

Voice VP
[arb]
V DP

The definition of Voice as arbitrary makes it impossible to check features with the internal
argument DP, which moves to the specifier position of VoiceP. From that position, T, which
has not lost any of its features, agrees with it. Crucially, the person feature in T is available for
agreement with the internal argument, so person-marked pronouns are allowed.
The analysis assumes that SE only appears in T, so that the presence of an [arb] specification
in Voice does not trigger insertion of SE. The analysis does not explain fully why the internal
argument has to move to spec, VoiceP, but note that in fact considering the Spanish facts we
probably want to make sure that the object DP can stay in the complement position of VP, in
order to account for the cases where the passive subject can be a bare nominal. Note that in
Dobrovie-Sorin (2021) there is no vP layer here or in the passive version: the loss of idiomatic
meanings does not follow from the absence of a vP layer, but it might follow from the different
specification of the Voice head, which –being specified as [arb] in the case of the periphrastic
passive– would differ from the stored idiomatic construction.
A second further difference between periphrastic passives and SE passives is the restriction
on the types of predicates that allow passive construals. The SE passive allows any predicate
that has an internal argument and an external argument, crucially including the following
classes (De Miguel 1992, Mendikoetxea 1999, Sánchez López 2002): (i) stative transitive verbs
(414); (ii) light verb constructions (415); (iii) intransitive verbs with cognate objects (416), and
(iv) verbs with non-referential direct objects that express quantities (417).

(414) a. Se saben muchas cosas.

SE know.3pl many things
b. ??Muchas cosas son sabidas.
many things are known
c. Se odian estas cosas.
SE hate.3pl these things
d. ??Estas cosas son odiadas.
these things are hated
(415) a. Se tienen muchos problemas.

154
 SE IN SPANISH

SE have.3pl many problems

b. ??Muchos problemas son tenidos.
many problems are had
c. Se toman varios medios de transporte.
SE take.3pl several means of transportation
d. *Varios medios de transporte son tomados.
several means of transportation are taken
(416) a. Se vivieron muchas aventuras.
SE lived many adventures
b. ??Muchas aventuras fueron vividas.
many adventures were lived
c. Se durmieron varias siestas.
SE slept.3pl several siestas
d. *Varias siestas fueron dormidas.
several siestas were slept
(417) a. Se nadaron tres piscinas.
SE swam.3pl three swimming-pools
b. *Tres piscinas fueron nadadas.
three swimming-pools were swum
c. Se corrieron tres kilómetros.
SE ran.3pl three kilometers
d. *Tres kilómetros fueron corridos.
three kilometers were run

Part of the literature claims that two verb classes disallow periphrastic passives, but we do
not share the judgements: ditransitive verbs and verbs that take clausal objects.

(418) a. Los presos fueron entregados a las autoridades.

the convicts were delivered to the authorities
b. El dinero fue donado a la caridad.
the money was donated to the charity
(419) a. Que era inaceptable fue indicado por el profesor.
that was unacceptable was indicated by the teacher
'It was said by the teacher that that was unacceptable'
b. Que habría huelga fue planteado en la reunión.
that would.be strike was proposed in the meeting
'It was proposed in the meeting that there would be a strike'

What seems to be at play here, however, is that some verbs that have a very light meaning,
both in the group of ditransitive and saying or thinking verbs, seem to accept the periphrastic
passive worse, but this might be another instantiation of periphrastic passives being bad with
light verbs.

(420) a. Las flores fueron {entregadas / ??dadas} a Juan.

the flowers were {delivered / given} to Juan
b. La respuesta fue {presentada / ??dicha} por el estudiante.
the answer was presented / said by the student
c. La solución fue {concebida / ??pensada} por el estudiante.
the solution was conceived / thought by the student

155
 ANTONIO FÁBREGAS

In sum, what seems to be defining these contrasts is one of the following two factors, or
their combination: (i) the object that becomes the subject in the periphrastic passive does not
introduce a clearly independent referent (416, 417) or is part of the predicate rather than an
argument (415), or (ii) the verb does not denote an event that can be conceptualised as telic or
completed (414). This is what one might expect if the periphrastic passive implies a categorial
structure where the grammatical subject is a theme, while in the SE passive the internal
argument stays in place and is part of what is predicated from a generic subject. Moreover, the
periphrastic passive is restricted to some specific temporoaspectual interpretations that do not
restrict the SE passive, which is the third relevant difference between the two structures.
Periphrastic passives tend to be ungrammatical with imperfective aspect, unless interpreted as
habitual statements, and prefer perfective or perfect meanings (see Brucart 2012 and Camacho
2012, where the participle has a telicity-inducing feature that the copula cannot license without
combining with perfective or perfect aspectual forms; note, in fact, that the presence of estar
in the progressive periphrasis saves the grammaticality in imperfective, as in 422c):

(421) a. Se venden muchos pisos.

SE sell.3pl many houses
b. ??Muchos pisos son vendidos.
many houses are sold
(422) a. Se leían muchos libros.
SE read.3pl many books
b. ??Muchos libros eran leídos.
many books were read
c. Muchos libros estaban siendo leídos en aquellos momentos.
many books were being read in those moments

The difficulty of accepting the periphrastic passive in imperfective constructions might

relate to the difficulty of forming them from stative predicates or from atelic predicates (Marín
2011). (423) tends to be interpreted as (424a), not as (424b), as the periphrastic passive tends
to be related to a completion meaning which involves telicity (Fábregas & Putnam 2020).

(423) El carro fue empujado.

the cart was pushed
(424) a. Empujaron el carro de golpe.
pushed the cart at once
'They gave a push to the cart'
b. Empujaron el carro durante un rato.
pushed the cart for a while
'They pushed the cart for a while'

This might relate to the possible absence of vP from the periphrastic passive, which
highlights a change of state that must be telic –alternatively, the participial form could define
a completed event that, therefore, must be interpreted as telic; see Fábregas & Putnam (2020)
for the analysis where periphrastic passives involve turning the VP layer into the figure of the
event, highlighting the change of state component that is associated to the event–. In general,
atelic verbs have difficulty forming periphrastic passives, as noted in Marín (2011).

(425) a. Luis condujo el coche.

Luis drove the car
b. ??El coche fue conducido.

156
 SE IN SPANISH

the car was driven

c. Se condujo el coche.
SE drove the car

Before ending this section, note that the fact that passive SE structures have a broader
distribution than periphrastic passives does not mean that each single verb that can be transitive
can have a passive SE version. As expected from a theory where passive SE structures project
an indefinite external argument interpreted as human, verbs that do not have a thematic position
for the external argument reject the passive SE structure (Ormazabal & Romero 2019). Thus,
(426) contrasts with (416) –the problem is not the cognate object, but the fact that the
meteorological verb does not have a theta position for a human subject–.

(426) *Se llueven piedras.

SE rain.3pl stones

Let us now leave passive and impersonal SE uses and move to middle SE uses.

9. Clause-affecting SE uses (3): middle SE

Middle SE constructions (427) –also known as 'middle-passive' SE or 'generic passive' SE,
particularly in those theories that use the term 'middle' for anticausative structures– are
generally related to passive SE constructions, and in many analyses have been treated as
passive SE structures where a particular operator has been added in order to make the predicate
stative and merely descriptive of the internal properties of the subject. However, as we will see,
their surface properties relate them closer to anticausatives than to passives.

(427) a. Estas camisas se planchan con facilidad

these shirts SE iron with ease
'These shirts, by virtue of their properties, can be easily ironed'
b. Este libro se lee bien.
this book SE reads well
'This book reads well'
c. Este material se rompe fácilmente con las manos.
this material SE breaks easily with the hands
'This material is easy to break with the hands'

The properties of middle statements in Spanish have been analysed, among others, by
Martín Zorraquino (1979), Omori (1997), Mendikoetxea (1999, 2000, 2012), Sánchez López
(2002), García Negroni (2002), Lekakou (2005), Kempchinsky (2006) and Fábregas & Putnam
(2020). The literature is not very abundant, in particular when one compares it with the one on
middles in Germanic languages or other Romance languages –which are also treated in some
of the works cited above– (Keyser y Roeper (1984), Roberts(1987), Fellbaum (1986) y Hale y
Keyser (1986, 1987), Condoravdi (1988), Fagan (1988, 1992), Stroik (1992, 1995), Hoeckstra
y Roberts (1993), Zribi-Hertz (1993) or Ackema & Schoorlemmer (1995), among many others.

9.1. Main properties: the grammatical subject

Let us start presenting the properties of middle SE structures. There are four crucial aspects
of their syntax and semantics that should be highlighted.

i) The position and referential properties of the subject

ii) The aspectual properties of the predicate

157
 ANTONIO FÁBREGAS

ii) The possibility of introducing a syntactically-active agent

iv) The presence of a verbal modifier

Let us start with the properties of the grammatical subject. Like in passive SE structures, the
grammatical subject agrees in number with the predicate, and is interpreted semantically as the
entity that would be affected in the event described.

(428) a. Este camión se conduce bien.

this truck SE drives well
b. Estos camiones se conducen bien.
these trucks SE drive.3pl well
'These trucks are easy to drive'

Like passive SE structures, personal pronouns are not allowed.

(429) a. #Yo me detengo con facilidad.

I me arrest with ease
Intended: 'I am easy to arrest'
b. #Tú te engañas con facilidad.
you you cheat with ease
Intended: 'You are easy to cheat'

The contrast with passive SE structures comes from two sources: first of all, unlike passive
SE and like anticausative SE, the subject appears almost compulsorily in preverbal position.
This establishes a similarity between middle SE structures and other stative predication
structures, where the subject has a strong tendency to be in preverbal position (Mendikoetxea
1999).

(430) a. ??Se lavan estas camisas con agua fría.

SE wash these shirts with water cold
b. ??Es Juan muy inteligente.
is Juan very intelligent

Secondly, and also unlike passive SE and like anticausative SE, the grammatical subject
cannot be a bare nominal.

(431) #Se lavan camisas con agua fría.

SE wash shirts with cold water
Intended: 'These shirts should be washed with cold water'

Both properties have been treated from a semantic perspective: unlike SE passives, which
are thetic statements where the whole clause is rhematic, middle SE structures involve stative
predication (Lekakou 2005): the middle SE structure does not express participation of the
grammatical subject in an event, but predicates from the grammatical subject a set of properties
by virtue of which the subject has a predisposition to potentially participate in that event, or
the participation in such event is expected, necessary, etc. Middle statements, then, are non-
episodic: they do not express specific instantiations of the event described by the predicate in
concrete time periods within the real world. They simply describe the subject: in (428) one
predicates from the truck (an individual representing a class of objects) that, by virtue of its
internal properties, it can participate easily in an event of being driven. However, one does not

158
 SE IN SPANISH

need to have driven that truck to know that driving it would be easy, and the truck might have
never been driven before: one simply deduces from its known properties that it should be easy
to drive it.
Now, from this perspective, the preverbal position of the subject follows: the truck is the
subject of a stative predication where a set of properties is ascribed to it.
In correlation with this, there is a strong tendency to interpret the subject as denoting a class
of entities. The denotation of a specific individual is only possible to the extent that the
individual is an exemplar of an established class. Note that middle SE structures combine
naturally with subjects that are marked explicitly as denoting a kind:

(432) a. Esta clase de libros se lee fácilmente.

this class of books SE read easily
b. Este tipo de camisas se lavan con agua fría.
this type of shirts SE wash with cold water
c. Libros como este se leen con facilidad.
books like this-one SE read with ease

The tendency of the subjects of middle SE structures to carry definite determiners (433)
follow from the kind denotation: they either express kinds that are taken as unique or they act
as exemplars taken from a unique class (García Negroni 2002). Note, however, that the relevant
generalisation seems to be that the subject should be related to a kind as an exemplar or as
denoting the class itself, given the grammaticality of (432c). In this sense, note that (433) can
be interpreted as 'some types of shirt', but not as 'some exemplars of shirt'.

(433) #Algunas camisas se lavan con agua fría.

some shirts SE wash with water cold
'Some types of shirt are supposed to be washed in cold water' (not 'Some shirts of this
only type are supposed to be washed in cold water')

Again, (433) improves if one makes sure that a kind denotation is available, as in (434),
where one interprets the indefinites as denoting two different kinds of shirts.

(434) Unas camisas se lavan con agua fría, otras con agua caliente.
some shirts SE wash with water cold, others with water warm

9.2. The aspectual properties of middle SE

With respect to the aspectual properties of the predicate, the intuition is that middle SE
stativises a predicate that is in principle eventive, making it denote a stative predication that
ascribes to the subject the necessity or tendency to participate in an event. As a result of this,
the predication is non-episodic, and relates to individual level predicates (Carlson 1977). This
makes middle SE structures be grammatical only with imperfective aspect, a property that
passive SE structures do not have. (435) loses the middle reading and is only possible in a
passive construal.

(435) a. #Estas camisas se han lavado con agua fría.

these shirts SE have washed with water cold
b. #Estas camisas se lavaron con agua fría.
these shirts SE washed.pfcv with water cold

159
 ANTONIO FÁBREGAS

Correlatively, the kind interpretation of the subject is lost, and in (435) the tendency is to
interpret that we talk about specific shirts, not a general class of shirts, or individuals that
simply represent a class. Interestingly, the generic reading of nominal subjects is typical with
individual level predicates (436, Carlson 1977), which together with the restriction on the
subject position noted in (430), supports the idea that middle SE structures are stative
Individual Level Predicates.

(436) Firefighters are brave.

The aspectual restrictions of middles has been argued to be reflected also in the type of
lexical aspect that the predicates that allow it must have, although with some controversy.
There is agreement that stative predicates cannot allow middle SE (437), unless they are
reinterpreted as telic processes –one can know Spanish history completely, after learning it step
by step, but one cannot know how to drive a truck completely (Sánchez López 2002, footnote
68)–.

(437) a. *Conducir este camión se sabe bien si se estudia lo suficiente.

to.drive this truch SE knows well if SE studies enough
b. La historia de España se sabe bien si se estudia lo suficiente.
the history of Spain SE knows well if SE studies enough

The ban on middles constructed from predicates that are already stative follows if somehow
the middle reading involves introducing some modal operator that binds the event expressed
by the predicate –roughly, instead of denoting an event, the middle denotes the tendency or
necessity that the event happens–. Stative predicates that are Individual level lack an event
variable that can be bound by the operator, so they are expected to be out unless recategorisable
as processes.
In contrast, it is more debated whether any event predicate can have a middle. Mendikoetxea
(2000) notes that only accomplishments trigger naturally the middle SE reading, and proposes
that activities and achievements do not really trigger middle SE interpretations (see Sánchez
López 2002 for a partial critique of this approach).

(438) a. Este camión se conduce con facilidad. Activity

this truck SE drives with ease
b. Estas setas se encuentran con facilidad. Achievement
these mushrooms SE find with ease

These sentences should be passive SE cases according to Mendikoetxea (2002). Without

denying that accomplishments are the class that most productively produces middle SE
structures, it is unclear that (438) are not cases of middle SE structures. Note that they are not
natural in a perfective SE, in the same interpretation where the subject denotes a kind.

(439) a. Este camión se condujo con facilidad.

this truck SE drove with ease
b. Estas setas se encontraron con facilidad.
these mushrooms SE found with ease

(439) is not interpreted as meaning that the truck was driven easily due to the internal
properties of the truck: it might be that the truck is difficult to drive, but that the driver was so
skilfull that the event was performed easily; similarly the mushrooms could have been found

160
 SE IN SPANISH

easily by chance or because the searcher was cunning, not because of some property of the
mushrooms. This shows that (438) has a special reading where the eventuality is necessary
easy to perform because of the internal properties of the subject, which is lost in the perfective
aspect interpretation. Thus, it might be that (438) are real cases of middle SE, and (439) are
passive SE structures.
What seems to be a generalisation is that the internal argument must be involved enough in
the event so that one can meaningfully predicate from it that its internal properties define a
tendency, propensity or necessity to participate in the event. This might be what explains that
most cases of middle SE structures involve accomplishments, as transitive accomplishments
tend to express changes of state with incremental themes, and an incremental theme measures
the event in a way that one can expect that its internal properties count in order to make the
event necessary or define a tendency towards it. In the case of (439a), driving a vehicle can be
conceived as dependent on the properties of the vehicle, and in the case of (439b), finding
something can be influenced by the abundance or frequency of the examplars representing this
class. It is true that most activities do not have this property, which explains the
ungrammaticality of (440).

(440) #Estas setas se buscan con facilidad.

these mushrooms SE search with ease

Searching something is not influenced by the internal properties of that something. With
respect to achievements, most achievements are intransitive, so this explains why they do not
appear naturally with middles, as they are 'passive' in the sense that the grammatical subject is
semantically an object.

9.3. Is there an agent present?

Let us now move to the semantic subject, which is one of the most controversial aspects of
the syntax of middle structures. The main question (discussed extensively in Stroik 1992, 1995)
is whether middles license overt agents.
That middles, like passive SE and unlike anticausative SE, license the reading of an external
agent is clear from sentences like (441), where PRO subjects are interpreted in the control
infinitive that are related to the notional agent of the middle.

(441) Esta clase de medicamentos se absorben fácilmente después de PRO comer.

this class of medicines SE absorb easily after of PRO to.eat
'This class of medicines are easily absorbed after eating'

Stroik (1992) also points out that in principle one can find anaphors that should be licensed
by the subject.

(442) Los libros sobre uno mismo se leen con gusto.

the books about one self SE read with pleasure

However, it is not so clear that these facts mean necessarily that there is an agent expressed
in the clause. (442) might involve a pro that is introduced within the structure of the NP itself,
given that the same anaphor is fine in (443), where one does not expect to find a thematic
position for an agent.

(443) Un libro sobre uno mismo es siempre divertido.

a book about one self is always fun

161
 ANTONIO FÁBREGAS

With respect to (441), the problem is that the interpretation of the agent in the infinitive is
anyways human and generic, a property that is general of PROs that are not controlled –the
problem being that by hypothesis the agent of a middle, if there is one, should also be
indefinite–.
In fact, there is more evidence that the middle in Spanish does not introduce an agent
thematically. Let us revise that evidence.

a) Middle SE structures, unlike passive SE structures, do not seem to license final clauses.
It is true that this property might follow from the stativity of the predicate, but remember that
middle SE structures start from eventive structures, and the stativity comes from an operator
that is merged higher in the tree: at the level at which the final clause is introduced, the operator
might not have been introduced yet. Also, note that some stative verbs license final clauses.

(444) Este libro se lee fácilmente (??para entender mejor el tema).

this book SE reads easily in.order.to understand better the topic
(445) Juan quiere a María para que hable bien de él.
Juan loves DOM María so that speaks well of him

b) Middle SE structures cannot introduce easily an agentive complement with the

preposition that introduces agent complements. In this sense, remember §8.2 above: passive
SE cannot introduce agent complements either, but these are marginally allowed with generic
or indefinite referents. Middle SE seems to be even pickier in this respect.

(446) a. Este libro se lee bien (?? por todo el mundo).

this book SE reads well by everybody
b. Este libro se lee bien (??por parte de niños y mayores).
this book SE reads well by part of children and elders

c) In contrast to passive SE structures, as we saw, the internal argument cannot be a bare

nominal. This is unexpected if, like in passive SE structures, the verb defines an external
argument as a pro and the internal argument can stay VP internally, in a complement position.
However, this matches the description of the anticausative, where the internal argument moves
to an external position.
d) Middle SE structures reject adverbs that are directly oriented towards agents, such as
queriendo 'willingly', deliberadamente 'deliberately' or a posta 'on purpose': *Estas camisas se
lavan a posta 'These shirts SE wash on purpose' is not grammatical. This goes along the lines
of anticausative structures too.

The question relates also to whether middle SE structures are unaccusative or not. From the
perspective of Spanish, the unaccusative tests either cannot be applied or give a negative result.
Internal arguments cannot be bare nominals, unlike in proper unaccusatives, but the absolute
participle test cannot be applied because it presupposes episodicity and middles are generics.

(447) #lavadas fácilmente estas camisas...

washed easily these shirts...

(447) is interpreted as a causative event: the shirts are not denoting a kind, and the ease does
not express the predisposition to participate in an event, but how easy the event was in fact
performed.

162
 SE IN SPANISH

Once the unaccusative analysis is discarded, there are two options to account for the position
of the internal argument: either it moves, as proposed for anticausatives in §8.1 above, or it is
base-merged in a specifier position.
The second side of the issue is the interpretation of the agent: why do we interpret an external
agent, unlike in the case of anticausatives, if that agent is not syntactically active? Lekakou
(2005) proposes that the reason is the lexical semantics of the verb, which unlike verbs that
allow an anticausative pair (§4.1) have a meaning that is incompatible with the interpretation
that the event is triggered exclusively by the internal argument, without participation of
external agents. Lekakou (2005) proposes that the almost compulsory presence of a modifier
in the predicate is related to this property: the modifier allows the semantic licensing of the
agent that is needed for the lexical verb, but remains unprojected in the syntax. Then let's
examine the modifier that is part of the middle structure.

9.4. Modifiers in middle SE structures and the interpretations of the middle

The modifiers that typically appear together with middle SE structures denote manners
(448a), instruments (448b), means (448c) or general situations that involve an agent (448d); in
all these cases, one deduces the presence of an agent that performs the event in one manner,
manipulates the instrument or uses the means.

(448) a. Estas camisas se lavan con dificultad.

these shirts SE wash with difficulty
b. Estas camisas se lavan con este jabón.
these shirts SE wash with this soap
c. Estas camisas se lavan en lavadoras como esta.
these shirts SE wash in washing-machines like this
d. Estas camisas se lavan si uno tiene que ir elegante.
these shirts SE wash if one has to dress elegantly

In fact, Spanish allows constructions where the modifier is missing. This takes us to the
question of how many non-episodic readings are allowed with middle SE. One can distinguish
three interpretations.

a) The necessity interpretation emerges typically without modifiers. The predicate says that,
by virtue of the properties of the internal argument, one is expected to perform a particular
event on them. The notion of agent is in such cases licensed by the necessity reading, that
requires some agent to perform the action.

(449) Estas camisas se planchan.

these shirts SE iron
'One must iron this type of shirt'

Note that contra García Negroni (2002) it is not true that the absence of a modifier forces
an exemplar reading of the subject. The necessity reading also emerges with some modifiers.

(450) Estas camisas se lavan con agua fría.

these shirts SE wash with water cold
'These shirts must be washed in cold water'

Next to the necessity reading, there are two more interpretations: the reading that also
involves the agent is interpreted as the subject having internal properties that facilitate or make

163
 ANTONIO FÁBREGAS

the event more difficult in combination with the means or instruments expressed by the
modifier. The reading is facilitated when there is a manner adverb.

(451) Estas camisas se lavan bien con agua fría.

these shirts SE wash well with water cold

The tendency reading means that by virtue of its internal properties the kind has a propensity
to participate in the event (452). In this reading, the agent is not involved.

(452) Este material se rompe (fácilmente) cuando hace calor.

this material SE breaks (easily) where there.is heat

This last reading takes us back to the question of whether middle SE structures should be
considered as derived from the passive SE or as derived from the anticausative SE.
Interestingly, one can divide the readings that involve the agent and those that simply express
the propensity to participate in an event according to whether the verbs involved allow an
anticausative or not (§4.1).

(453) a. Este material se rompe cuando hace calor. Propensity reading

this material SE breaks when there.is heat
b. Esta clase de puertas se abre cuando hace viento. Propensity reading
this class of doors SE open when there.is wind
(454) a. Estos libros se leen bien. Facilitation reading
these books SE read well
b. Estas camisas se planchan fácilmente. Facilitation reading
these shirts SE iron easily

This opens up a possible analysis of middle SE structures: they are anticausative structures
where one simply adds an operator that triggers the non-episodic reading of the event and the
kind interpretation of the internal argument. As in the case of the anticausative, the internal
argument has to move to a specifier position. There is no agent projected in the structure: verbs
that lexically involve an agent and do not allow the anticausative reading require, because of
that, a modifier that licenses the agent interpretation that is required by the predicate but is
missing from the syntactic structure: the necessity or the facilitation reading emerges.
Predicates that allow the anticausative reading without qualification give rise to the propensity
reading in combination with the operator.

9.5. Middles as anticausative SE, not passive SE

If we continue the reasoning presented in §8.1, where we argued that the configuration in
(455) characterises the anticausative construal, we could propose that (456) stands for a middle
SE: an operator expressing genericity (Condoravdi 1989, Steinbach 2002, Marelj 2004) or
ascriptional properties (Lekakou 2005) is added to the structure, triggering the interpretation
of both the event and the internal argument as generic.

164
 SE IN SPANISH

(455) VoiceP

DPi Voice

Voice vP
sei
v VP

V DP
(456) XP

Op X

X VoiceP

DPi Voice

Voice vP
sei
v VP

V DP

The structure does not project any agent syntactically, but agentivity might be expressed in
the meaning of vP, which perhaps could be treated as absent in anticausatives. Importantly, just
as in the case of anticausatives, locating the internal argument in spec, VoiceP triggers a reading
where the event is related to the internal properties of the internal argument: in this case, those
internal properties require that an event is performed, facilitate an event or define a propensity
to participate in such event.
Verbs that allow the anticausative reading are directly semantically licensed in (456), while
verbs that reject the internal-causation of the event require a modifier that satisfies the semantic
condition that there is an agent.
Once the propensity and the facilitation / necessity reading are differentiated by the type of
verb, we can revisit the only important difference that on the surface anticausative SE and
middle SE have: in (429) above we said that, like passive SE, middle SE does not allow
personal pronouns as subject. This ban, however, can be derived from two sources that might
overlap: (i) the subject of a middle statement is either a kind or an exemplar from a kind, and
personal pronouns do not denote any kind because they lack descriptive properties; (ii)
introducing a personal pronoun in the middle structure in (455) would force a reading where
the personal pronoun, compulsorily human, would be interpreted as the agent of the structure,
contra the required interpretation of the middle.
However, before going any further let us revise other accounts of the middle SE structure.

9.5.1. Other syntactic accounts of the middle

There are three types of analysis of middle SE in the market.

i. Middles are structures that don't project the agent, where the internal argument moves
ii. Middles are structures that project the agent and where the internal argument moves
iii. Middles are structures where the syntactic subject is projected in the external argument
position

165
 ANTONIO FÁBREGAS

The first one is illustrated by Kempchinsky (2006), which is the one closest to the one we
have just sketched: the structure of the middle does not involve any syntactically-projected
subject and there is movement of the internal argument. In accordance with Kempchimsky's
treatment of SE as an aspectual element, the role of SE is to cancel the eventivity of the
predicate, in combination with the adverbial modifier. She proposes that the modifier occupies
the position of complement of VP (Stroik 1992), forcing the internal argument to project as a
specifier. The stativity of the middle statement follows from a double operation that reduces
the subevents: the modifier, not the internal argument, moves to spec, AspP, blocking the
possibility that there is a telic change of state component that is identified by the internal
argument. Secondly, SE is introduced in the head v, where it blocks the causation subevent that
defines the initial stage of the event. As a result, the event does not have an initial or a final
point: it now denotes a state.

(457) vP

DPi v

v AspP
sei
PP/AdvP Asp

Asp VP

DP V

V PP/AdvP

Inserting SE in v blocks the possible introduction of an agent, producing a monadic predicate

that she argues would naturally trigger a generic reading: a monadic predicate that has
cancelled all its eventive properties would be interpreted as a generic statement about the
properties of the subject. In their semantic representation, middle SE structures involve a high
generic operator which can only be restricted by the only overt argument, the internal argument.
This forces the subject to move to a high position to restrict the operator, which produces the
compulsory preverbal position of the subject.
As can be seen in (457), the internal argument is always in a specifier position, accounting
for why it cannot be a bare nominal. The structure differs from passive SE in this account (cf.
§8.1) because in this case SE is not the specifier of vP, so it does not block movement of the
internal argument to that position. The difference with anticausative SE is that the anticausative
involves an operation in the head Asp which makes it possible not to project vP: anticausatives
would lack vP, and middles and passives would have SE. However, both anticausative and
middle SE would be structures where the agent is not syntactically projected. Thus, the analysis
relates the middle SE with the passive, not with the anticausative.
The main problem of this account is that it presupposes that the modifier should be
compulsory in Spanish middles, and moreover that it should be a 'small' constituent; we have
seen above that the modifier can be missing or can even be expressed as a clausal constituent
with a conditional or a temporal meaning, which is less likely to be introduced as an internal
argument of the verb and then displace to the specifier of AspP.
The second type of analysis proposes that the syntax of middles is basically like the one of
passives because the agent is projected, although tacit, and the internal argument moves to the

166
 SE IN SPANISH

subject position. Hoekstra & Roberts (1993) or Stroik (1995) propose this, with the only
difference that the former propose that the agent is located in the cannonical position and the
latter adds it to an adjunct position.

(458) TP

DP T

T vP

pro v

v VP

AdvP VP

V DP

As in Kempchinsky (2006), a modal operator could be added to the structure at a high

position, which would determine the kind interpretation of the subject and the generic reading.
Presumably this analysis would place SE in the point where the external argument pro is
introduced. The problem of this account is the body of evidence that we have already seen,
which suggest that, unlike passive SE structures, middle SE structures do not have a
syntactically active agent. Moreover, this analysis treats middle structures as unaccusatives,
which we have seen is not warranted in Spanish.
The third approach to middle SE is the one that Ackema & Schoorlemmer (1995) advocate
for (see also Lekakou 2005): the grammatical subject is base-generated in an external argument
position. The authors assume that middles are created lexically through a pre-syntactic
operation that alters the aspect and the valency of the transitive verb in order to obtain an
unergative intransitive verb with a stative meaning. Being unergative, the subject must be
preverbal and cannot be a bare nominal; the ascriptional meaning derives from the
underspecification of the argument that corresponds to the agent, which is assumed to be
defined as arbitrary by the lexical operation. Presumably, in this analysis SE would mark at the
VP level the operation that alters the aspect and argument structure of the verb, reducing it to
stative and monoargumental.
The problems with this approach can be summarised in two points: (i) the authors are forced
to divorce the position of the argument from its thematic interpretation. An argument located
in the external argument position should not receive a patient reading; the lexical operation that
they propose tries to clarify this issue by idiosyncratically proposing a change in the position
where theta roles are assigned, but beyond the technical solution the problem persists; (ii) the
analysis cannot explain the correlation between the kind reading of the subject, stativity in the
predicate and the reading where the internal properties of the internal argument define the
propensity or facilitation of the event. In contrast, when the analysis associates middles with
anticausatives, the relevance of the internal properties of the object to define the event become
linked to one and only one configuration where the internal argument moves to spec, VoiceP
to bind the reflexive.

9.5.2. Relation between middle SE and predicate-affecting SE uses

The syntactic evidence suggests then that middle SE could be merged with anticausative SE
with only two differences:

167
 ANTONIO FÁBREGAS

i) In middle SE there is an operator with a generic / ascriptional meaning that is absent from
anticausative or passive SE
ii) In middle SE verbs that require an agent for semantic reasons can also participate,
provided that there is a semantic component that licenses that agent reading

The point in (ii) essentially means that, as proposed in other accounts, verbs do not have to
fill syntactic positions to discharge their theta-grid (Ramchand 2008), so that presence or
absence of an argument is determined by the syntactic structure, and the theta roles are read
from the syntactic configurations. Thus, a verb that cannot be semantically anticausative in
principle can anyways appear in a syntactically anticausative construction provided that the
mismatch between the type of event that the verbal stem expresses and the lack of an agent in
the syntax is somehow solved. For instance, one can accept something like (459), even though
the verb does not allow an anticausative:

(459) Esta clase de políticos se asesinan con facilidad.

this class of politicians SE murder with ease

Modification seems to be the way to solve this issue, when the verb cannot receive an
anticausative reading.
A second partial correlation between middle SE and some types of predicate-affecting SE
uses, that is advanced in Sánchez López (2002) is the correlation with aspectual SE (§5). The
core of the connection would be based on the observation relating to examples like (437) above
–repeated here with its original numbering for convenience–.

(437) a. *Conducir este camión se sabe bien si se estudia lo suficiente.

to.drive this truch SE knows well if SE studies enough
b. La historia de España se sabe bien si se estudia lo suficiente.
the history of Spain SE knows well if SE studies enough

The intuition is that the verb saber 'know' is stative in its prototypical uses, but can be
reconceptualised as implying some type of process (437b) when the direct object can be
conceptualised as an entity with different internal parts that can define some kind of process.
This would approach these predicates to the transitive incremental theme verbs that generally
combine with aspectual SE (§5.3). The problem, however, has two sides. The first side is that
it is unclear in what sense one can claim that (437) has some notion of process, as the verb
cannot accept the progressive periphrasis.

(460) *Estoy sabiendo la historia de España.

am knowing the history of Spain

Perhaps one could claim that (437b) defines a state with internal parts, in such a way that
each portion of that state can be related to a portion of the object that is known, and –despite
the absence of any dynamic component– one can still treat the object as a path that measures
the extension of the knowledge. In this sense, one could perhaps relate these predicates with
the typical aspectual SE predicates, and claim that the restriction that makes aspectual SE also
partially applies to middle SE.
The second part of the issue is that one could consider saber 'know' an auxiliary in
combination with infinitives, as Bosque (2000) does, while it is a real predicate in combination
with nominal objects. Like that, then, the contrast would not refer to an aspectual distinction,

168
 SE IN SPANISH

but rather to a distinction between different kinds of verbs (although, note, that still does not
solve the problem of how saber would be compatible with an aspectual SE, assuming that
aspectual SE telicises the event).
It is difficult to extend the aspectual account to the rest of data where stative verbs seem to
combine with aspectual SE (MacDonald 2017; §5.3 above). MacDonald (2017) noted that there
are at least three verbs that are stative and combine with a SE that behaves as aspectual in the
sense that (i) it varies with the subject; (ii) it does not detransitivise the verb; (iii) it forces the
direct object to be bounded (461, repeating 194 above).

(461) a. Juan se cree lo que has dicho.

Juan SE believes that what have said
'Juan believes what you have said'
b. Juan se sabe la lección.
Juan SE knows the lesson
c. Juan se conoce muy bien a María.
Juan se knows very well DOM María

Interestingly, also here one could claim that these stative verbs involve a 'completely'
meaning with SE, but here the objects do not always clearly correspond to entities that can be
argued to be divided in different parts that match different mereological parts of the state. One
could decompose 'what you have said' into different statements, and 'the lesson' in different
ideas, but it is more difficult to do the same with an individual object like 'María'. In fact, non-
mereological objects, consisting on only one idea or concept, can be combined with these
verbs:

(462) a. Juan se cree que has venido.

Juan SE believes that you.have come
'Juan believes that you have come'
b. Juan se sabe que no puede salir.
Juan SE knows that not can go.out
'Juan knows that he cannot go out'

That is why MacDonald (2017) proposes that these verbs should, as other aspectual SE uses,
be analysed as dative reflexives which add the meaning that the subject is somehow involved
and affected in the belief –the subject is more implied in that knowledge, and one assumes that
the knowledge might affect him somehow–. Then, going back to the example in (459b), it
seems that what might be at play here is that the internal properties of an object like the history
of Spain can be conceptualised as influencing how easily the event of knowing it can happen.
(459b) involves knowing facts or dates, while (459a) involves knowing how to do something,
and this might influence the distinction.
Thus, to conclude, middle SE seems to establish only a relation with anticausative SE, not
with aspectual SE. Let us now finally move to the last type of SE structure, spurious SE.

10. Clause-affecting SE uses (4): spurious SE

Concluding this overview with the case of the spurious SE is a good way to remind ourselves
that SE is, after all, a member within the clitic system of Spanish, despite the preponderance of
analyses that we have seen above and that treat SE as the spell out of a verbal head. Spurious
SE is illustrated in (463): SE seems to replace the dative clitic le / les, irrespective of number,
when an accusative clitic appears in the same verb.

169
 ANTONIO FÁBREGAS

(463) a. Juan le dio una manzana a Pedro.

Juan him.dat gave an apple to Pedro
b. *Juan le la dio.
Juan him.dat her.acc gave
c. Juan se la dio.
Juan SE her.acc gave

Although of course noted as a phenomenon earlier, spurious SE has been analysed in a

number of works starting with Perlmutter (1971): Bonet (1991, 1995), Ordóñez (2002, 2012),
Heap (1998), Heggie & Ordóñez (2005), Nevins (2007), Walkow (2012, 2013), Cuervo (2013),
Alcaraz (2017) or Fábregas & Cabré (2020, 2021), among others.
First we will provide the main empirical properties of the spurious SE in order to later
discuss the two main types of analyses in the market: (i) morphological and (ii) syntactic.

10.1. Main properties

The distribution of spurious SE is dependent on two factors: (i) the availability of a third
person dative clitic, which on the surface SE substitutes; (ii) the presence of an accusative third
person clitic. (463) above corresponds to this description. When one of the two conditions is
not met, spurious SE is unavailable.
On the one hand, person-marked clitics are not replaced by SE, even in the presence of a
third person accusative clitic:

(464) a. Me lo dan.
me it give
'They give it to me'
b. Os lo dan.
you it give
'They give it to you'

On the other hand, when the direct object is not cliticised, spurious SE does not appear.

(465) a. Le entregamos el regalo a María.

her gave the present to Maria
b. Le dimos la respuesta a ella.
her gave the answer to her
c. Les ofrecimos ayuda a todos.
them offered help to all
'We offered help to all of them'

Spurious SE is not sensitive to the number of the dative pronoun; remember that SE as a
clitic does not have number variation:

(466) a. Les dimos el libro.

them gave the book
b. Se lo dimos.
SE it gave
'We gave it to him'
(467) a. Le dimos el libro.
her gave the book
b. Se lo dimos.

170
 SE IN SPANISH

SE it gave
'We gave it to her'

Spurious SE is not sensitive either to the syntactic function or semantic interpretation of the
dative; so-called ethical datives expressing just the emotional involvement of a participant in
the event is replaced by SE (468), just like indirect object datives expressing goal (469), source
(470), benefactive (471), possessor (472) or experiencer (473):

(468) a. El niño no le come las verduras.

the boy not him eats the vegetables
'The boy does not eat his vegatables, and that affects her'
b. A ella, el niño no se las come.
to her the boy not SE them eat
'The boy does not eat them, and that affects her'
(469) a. Les repartimos los caramelos.
them distributed the candy
b. Se los repartimos.
SE them distributed
'We distributed the candy to them'
(470) a. Le robaron la moto.
him stole the motorcycle
b. Se la robaron.
SE it stole
'They stole the motorcycle from him'
(471) a. Le compusieron una canción.
her composed a song
b. Se la compusieron.
SE it composed
'They composed a song for her'
(472) a. Le operaron la mano.
her operated the hand
b. Se la operaron.
SE it operated
(473) a. A María le da asco eso.
to María her gives disgust that
b. A María se lo da eso.
to María SE it gives that

Generally, the spurious SE is exclusively associated to dative pronouns in the literature.

There is, however, one case where SE seems to be replacing an accusative pronoun, noted by
Gómez Torrego (1992: 6):

(474) a. Llaman tonta a Juana.

call.3pl stupid DOM Juana
'They call Juana stupid'
b. La llaman tonta.
her.acc call.3pl stupid
'They call her stupid'
c. Se lo llaman.
SE that call

171
 ANTONIO FÁBREGAS

'They call her that'

The example (474c) involves a rare construction where one could have two clitics of third
person, one for the predicative adjective and one for the direct object that is the subject of that
predicate. As in the cases of dative replacement (463a), when the predicative adjective is not
substituted by a clitic the replacement does not happen.

(475) La llaman así.

her.acc call.3pl like.that
'They call her like that'

Thus, contra the general claim in the literature, that treats spurious SE as a strategy to replace
the dative clitic, an example like (474) suggests that one has to accept the principle in (476):

(476) SE is the clitic used to substitute direct or indirect third person clitics when there is
another third person clitic in the structure

The problem would be that contexts where two accusative pronouns could in principle
appear are rare, so this is an almost unique example. There are three empirical points that should
be made with respect to (474). The first one is that the replacement of a predicative adjective
by an accusative clitic is rare in Spanish. While general with copulative verbs (477), most
predicative adjectives –including selected predicative complements, as those in (478)– cannot
be pronominalised by an accusative; substitution by other non-clitic proforms is preferred
(479).

(477) a. Juan es alto.

Juan is tall
b. Juan lo es.
Juan it is
'Juan is like that'
(478) a. Consideran a Juan alto.
consider.3pl DOM Juan tall
b. *(Se) lo consideran (a Juan).
SE it consider to Juan
Intended: 'They consider Juan like that'
(479) Consideran a Juan así.
consider DOM Juan like this

The case of (464) is almost unique. Other verbs that, if only marginally, allow the
pronominalisation of the predicative complement are virtual synonym of llamar 'to call':

(480) a. ?Se lo denominan.

SE it denominate
'They call her that'
b. ?Se lo califican.
SE it qualify
'They qualify her as that'
c. ??Se lo designan.
SE it designate
'They designate her in that way'

172
 SE IN SPANISH

What makes llamar special so that the predicative adjective can be pronominalised is
unclear, but might perhaps be related to a property of this verb in some grammatical traditions:
its anticausative version, llamarse 'to be called' is put in the same group as copulative verbs for
instance in the Norwegian grammatical tradition because the predicative complement is
compulsory.

(481) Mi hijo se llama *(Pedro).

my son SE calls Pedro

Other selected predicative elements, when removed from the overt expression in the
predicate, at least allow a generic reading:

(482) El jefe no considera a Juan.

the boss not considers DOM Juan
'His boss does not consider Juan (in a good way)'

This might mean that the predicative complement is, in a verb like this, a real argument of
the predicate, and not just part of the predicate.
The third empirical side of the issue is that the direct object of the verb llamar 'call' receives
DOM even when it is an object.

(483) En esta región llaman 'papa' a la patata.

In this region call.3pl 'papa' DOM the potato
'In this region, the potato is called 'papa''

This might be an extreme manifestation of the fact noted in Ormazabal & Romero (2013),
namely that in contexts where there is a predicative complement DOM can be added to objects
that generally are not DOM:

(484) a. Dejaron atrás (*a) los árboles.

left.3pl behind DOM the trees
'They left the trees behind'
b. Dejaron (a) los árboles sin hojas.
left.3pl DOM the trees without leaves
'They left the trees without leaves'
(485) a. Vimos (*a) un avión.
saw.1pl DOM a plane
b. Vimos (a) un avión despegar.
saw.1pl DOM a plane take.off
'We saw a plane taking off'

In correlation with this, the dative pronoun seems to be available for non-leísta speakers to
substitute the direct object with llamar (but see 486c, the title of a famous movie), but this is
more difficult with feminine accusatives (although see 487c, which is the title of a famous song
by José Luis Perales).

(486) a. Llamaban tonto a Juan.

called.3pl stupid DOM Juan
'They called him stupid'

173
 ANTONIO FÁBREGAS

b. Le llamaban tonto.
him.dat called.3pl stupid
b. Lo llamaban Trinidad
him.acc called.3pl Trinidad
'They called him Trinidad'
(487) a. Llamaban loca a Juana.
called.3pl crazy DOM Juana.
b. La llamaban loca.
her.acc called crazy
c. Le llamaban loca.
her.dat called crazy

Note, in any case, that if periphrastic passives show that an argument is marked as accusative
and not dative in Spanish, the direct object is clearly accusative, given the possibility of
passivisation.

(488) Era llamada loca.

was called crazy
'She was called crazy'

Be it as it may, empirically the principle in (476) seems to hold. Let us now move to the
main analyses of spurious SE that are in the market.

10.2. Morphological analyses

There are two main types of analyses of spurious SE which, of course, concentrate on the
dative cases, which are the more general ones. One first, predominant, analysis treats the
spurious SE as a morphophonological phenomenon, that is, as a replacement that happens after
the syntactic structure has been computed and which affects only the morphological
manifestation of the clitic without interfering with the syntactic or semantic interpretation. This
type of analysis is performed, with technical details, in Perlmutter (1971), Bonet (1991, 1995),
Harris (1995), Heap (1998), Nevins (2007; see also 2012), Pescarini (2010), among many
others.
The core intuition of this type of analysis is the following:

i) What goes wrong in a combination *le lo is not a syntactic problem

ii) The problem is that the feature specification of the two clitic pronouns are too similar to
each other and the clitics are adjacent in the morphological structure
iii) This similarity should be taken to be a manifestation of a morphological Obligatory
Contour Principle (see Nevins 2012), a principle originally coming from phonology which bans
sequences of adjacent segments whose feature specification is too similar. Similarly,
morphology would ban combinations of adjacent clitics that are too similar.
iv) The solution to avoid the near-identity between the clitics is to erase (part of) the feature
endowment of the clitic
v) As SE is the least specified clitic within the system of Spanish clitics, the removal of
features from the dative produces SE as a result.

Although with a clear (and acknowledged by the author) antecedent in Perlmutter (1971),
the analysis is most clearly represented by Bonet (1991), where she proposes the following
morphological feature specification for SE (489a), the dative clitic (489b) and the accusative
clitic (489c).

174
 SE IN SPANISH

(489) a. CL (cl = clitic)

Arg (Arg = argumental)

b. CL

Arg 3rd

([pl]) ([fem])

c. CL

Arg 3rd

([pl]) Obl (Obl = oblique)

(489b) and (489c) are too similar to be adjacent. Bonet (1991) assumes the existence of a
template in Spanish that determines the order of clitics when spelled out, and which
unfortunately forces that the 3rd person dative and the 3rd person accusative are adjacent.
The post-syntactic component, in the morphophonological structure, reduces the similarity
by removing features from (489b) as follows:

(490) CL / CL ---> CL / CL

Arg 3rd Arg 3rd Arg Arg 3rd

([pl]) Obl ([pl])

The resulting specification of the dative clitic after the reduction is almost identical to the
one corresponding to SE, despite the possible plural feature, which remains and might explain
that in some varieties (491a) becomes (491b) (Harris 1994, 1995, Grimshaw 1997).

(491) a. Les di el libro.

them gave the book
'I gave them the book'
b. Se lo+s di.
SE it+pl. gave
'I gave it to them'

This analysis is virtually unquestioned in the existing literature, at least in its core
assumptions. The debate reduce to three points that can be considered as relatively minor: (i)
whether one needs to postulate a template to obtain the order or not (Harris 1995, for instance);
(ii) whether the impoverishment operation is necessary or the default clitic can emerge through
the conspiracy of other principles that also refer to the similarity in features (Grimshaw 1997,
for instance) and (iii) whether the result of impoverishing the dative clitic produces a form that
is identical to SE or just similar enough to it so that the same SE clitic is used, with correlative
alternative proposals about the feature geometry that characterises the clitics (cf. Cuervo 2013,
for instance). All these approaches, however, share the intuition that this replacement is a

175
 ANTONIO FÁBREGAS

morphophonological effect, which does not affect the syntax or the semantics of the
construction.
Note also that the intuitions also can account for the rare case in (474), where an accusative
clitic is replaced by SE in the presence of another clitic: like in (490), there we have a sequence
of two adjacent 3rd person clitics, which under the same premises as the dative case would lead
to the impoverishment of the features of the first clitic.

10.3. Problems for the morphological analyses

However, there are several aspects that cast doubt on the general morphological analysis, as
put forth by Alcaraz (2017) and Fábregas & Cabré (2021).

i) Is the rule proposed enough to account for all instances of spurious SE?
ii) Is it really the case that spurious SE is an effect that only takes place in the
morphophonological component, not affecting syntax or its interpretation? Note that this would
make spurious SE the only instance of the use of SE that is vacuous for syntax or the semantic
component.

Let us start with the first point. Alcaraz (2017), who in fact argues that spurious SE is
nothing but another normal instance of SE in syntax, shows that there are contexts where the
spurious SE can appear but where there is no initial dative clitic to replace. He notes, from
Fernández Soriano (1989), that dative clitics cannot double bare nominal indirect objects.

(492) (*Les) entregaron los regalos a niños sin padres.

them delivered the presents to children without parents

However, spurious SE is possible in this context.

(493) Se los entregaron a niños sin padres.

SE them delivered to children without parents

What Alcaraz (2017) suggests, although not giving many details, is that SE is necessary to
license two third person objects –note that his approach manages to predict that there should
be cases like (474 above, given that here we would also have two third person clitics). We will
get back to this idea, but first let us revise the evidence about the second problem: spurious SE
does not behave like a dative in the syntax.

a) LE can solve Weak Crossover (WCO) Effects, configuration where after A-bar
movement of a constituent correferential with another constituent, coreferentiality is
impossible. (494a) is an instance of WCO, and (494b), with the dative clitic, does not trigger
the ungrammaticalituy.

(494) a. ¿A quiéni dio un tortazo la chica que estuvo bailando con él*i?
to whom gave a slap the girl that was dancing with him?
b. ¿A quiéni lei dio un tortazo la chica que estuvo bailando con éli?
to whom him gave a slap the girl that was dancing with him?
'Who received a slap from the girl that had been dancing with him?'

SE does not trigger the same improvement of the WCO effect.

(495) ¿A quiéni sei lo dio la chica que estuvo bailando con él*i?

176
 SE IN SPANISH

to whom him it gave the girl that was dancing with him?
'Who received (a slap) from the girl that had been dancing with him?'

b) In contrast, LE produces Principle C violations when a proper name appears in postverbal

position, as seen in (496a) –remember that Principle C states that a referential expression
cannot be coreferential with a pronoun that precedes it within its domain–. Spurious SE does
not produce a Principle C violation.

(496) a. Lei entregará el premio el padre de Messi*i.

him will.deliver the prize the father of Messi
Intended: 'Messi's father will give the prize to Messi'
b. Sei lo entregará el padre de Messii.
SE it will.deliver the father of Messi
'Messi's father will give it to Messi'

c) Spanish sendos is a dual distributive quantifier ('one for one individual, another one for
another individual') which requires a plural antecedent in its domain (Bosque 1992). The plural
dative clitic can act as that antecedent (497a), but spurious SE cannot even if it is supposed not
to affect the plurality in the syntactic component (497b).

(497) a. Lesi entregué la carta en sendosi sobres.

them delivered the letter in SENDOS envelopes
'I delivered them the letter in a different envelope for each one of them'
b. *Se la entregué en sendos sobres.
SE it delivered in SENDOS envelopes
Intended: 'I delivered it to them in a different envelope for each one of them'

Next to this, Alcaraz (2017) shows evidence that spurious SE behaves like a regular
reflexive clitic.

a) In ellipsis contexts, Vehicle Change (Fiengo & May 1994) saves (498a), in combination
with the dative clitic, does not save the derivation: one either violates Princple C (498b) or
Principle B (498c).

(498) a. *Pedro le echó la culpa a Maríai, y ellai también.

Pedro her put the blame on María, and she too
b. *Pedro put the blame on María and she put the blame on María too (Principle C)
c. *Pedro put the blame on María and she put the blame on her too (Principle B)
Spurious SE, however, avoids the infraction of Principle B, because SE is a reflexive
anaphor.

(499) a. Pedro se la echó a Maríai, y ellai también.

Pedro SE it put on María, and she too
b. *Pedro put it on María, and she put the blame on María too (Principle C)
c. Pedro put it on María, and she put the blame on her too (No Principle B)

Following Torrego (1995), strong pronouns doubled by the reflexive SE do not constitute
infractions of Principle B, which is what happens in (499c).

177
 ANTONIO FÁBREGAS

(500) Juan se criticó a él.

Juan SE criticises DOM him
'Juan criticised himself'

All these effects are unexpected if the spurious SE emerges from a normal syntactic dative
pronoun through operations that affect only morphophonology: the presence of SE has
syntactic and semantic effects that suggest that spurious SE is indeed the same type of reflexive
pronoun that one has in the other cases.

10.4. Possible syntactic analyses

Alcaraz' (2017) suggestion was that SE is a strategy to license two third person clitics within
the same syntactic domain; while he does not develop the idea, here we will allow ourselves to
speculate about its possible implementation, proposing two alternative views:

(i) SE is introduced instead of LE to allow another element agree with a functional head.
(ii) SE is introduced as a functional head in order to license the agreement of that element.

Although he deals with Person Case Constraint effects, Walkow (2012, 2013) is, to the best
of our knowledge, the author that has provided the most explicit implementation of clitic
restrictions in Romance. The core of his proposal is the following:

i) Clitic incompatibilities are situations where there are two elements that need to be licensed
by agreement and only one head to agree with.
ii) There is a functional head –call it v– that is the sole source of agreement with the objects
iii) This head has a limited number of features; it will agree first with the closest element
within its complement, and then if there are enough features left it will agree with the second
closest element in its domain
iv) The clitic that gets spelled out corresponds to the features that have been licensed by
agree –Walkow's (2012, 2013) view of clitics corresponds to the theories that in §11 below we
will describe as 'clitics-as-agreement'–.

Walkow's analysis of the prototypical PCC effect –the one where a dative clitic blocks the
presence of a person marked clitic– would start from a configuration like the one that follows:

(501) vP

v HP

DO H
person
H VP

V ApplP

IO Appl
part
Appl DO

Assume that (i) v has unvalued features for person and participant, (ii) the case of the direct
object (DO) and indirect object (IO) have already been checked, respectively, by H and ApplP

178
 SE IN SPANISH

and (iii) the direct object moves above the indirect object for case assigning purposes. In this
configuration, v agrees first with the DO, which is closer in the constituent contained in its
complement. If the DO does not have a participant feature –that is, is not specified as 1st person
or 2nd person, which are the two participants of the event–, the person feature is used to check
the DO and the participant feature is left for the indirect object, which is in this case a 1st or
2nd person dative. This generates (502). Note that the ordering of clitics reproduces the
ordering of the agreement operations: the closest clitic is the one corresponding to the DO,
because this was the one that agreed first with v.

(502) a. Me lo dio.
me it gave
'He gave it to me'

The situation where the DO contains participant has the problem that the DO checks with v
both the person and the participant feature, not letting any additional argument for checking by
v.

(503) vP

v HP

DO H
person
part H VP

V ApplP

IO Appl

Appl DO

As the IO cannot agree with v, there cannot be a dative clitic, banning (504), which is the
standard PCC effect.

(504) *Me le presentó.

me.acc him.dat presented
Intended: 'He introduced me to him'

What one can generate is (505), without a clitic, to save the PCC.

(505) Me presentó a él.

me.acc introduced to him
'He introduced me to him'

In the case of the spurious SE, the proposal is that the DO checks all features from v, with
the exception of one feature that characterises datives in Spanish: a [loc] (local) feature, that
can be used as a feature in v when v carries no participant feature.

179
 ANTONIO FÁBREGAS

(506) vP

v HP

DO H
person
H VP

V ApplP

IO Appl
local
Appl DO

Direct objects lack [loc], so that feature is always left for the IO to check. If [loc] is the
feature endowment of SE, this produces a sequence of clitics where the closest clitic is an
accusative and the more external clitic is SE:

(507) Se lo di.
SE it gave

(508) is of course banned because v only has one person feature, which is used by the DO,
and the dative clitic would require agreement in the person feature too.

(508) *Le lo di.

him it gave

Walkow's approach has two problems, from the perspective of the facts above. The first one
is that it is unclear how the approach would be extended to examples where SE substitutes an
accusative pronoun, as (474) above: even if one can assume the configuration in (509), where
the predicative is lower than the DO and introduced by some sort of applicative head ('to give
a name to someone'), one should not be able to use the specification [loc] for the direct object,
precisely because using that feature with some DOs would fail to explain the effect for datives.

(509) * vP

v HP

(DO) H

H VP

V ApplP

PRED Appl

Appl DO

To account for (474), no matter which configuration is used –with DO above or below
PRED–, the DO should have [loc], in order to emerge as SE under this analysis, and then it

180
 SE IN SPANISH

should actually be a dative, something that is not crystal clear in the pattern (474), given the
choice of clitics –at least in the varieties that do not use le for it–.
The second problem is probably more serious: Walkow's analysis treats the effect as
syntactic only to the extent that the syntactic derivation derives the impoverishment of the
clitic, but syntactically the IO is still an IO and SE emerges at spell out, as a
morphophonological reflex that the IO has not been licensed in full. This means that spurious
SE should still behave in syntax and semantics as a dative, not as a reflexive, against the
evidence that Alcaraz (2017) invokes.
Walkow's system, however, opens two possibilities that –as we will see again in §11– relate
to two alternative views of what SE is next to the verbal marker: (i) SE is a second agreeing
head that introduces uninterpretable features that licenses the dative when the accusative takes
the inflection –SE-as-verbal-head– (ii) SE is an argument that is introduced in the derivation
instead of the dative so that the only agreeing head has enough features left for the accusative
–SE-as-argument–. Let us start with the first.
The first approach would relate, as we will see, with Reuland's (2011) proposal that reflexive
elements are in fact heads containing unvalued features (see §12.4 below); specifically, we
could assume unvalued number and person features, without participant features. Take the
configuration in (506) as a starting point, without positing the feature [loc].

(510) vP

v HP

DO H
person
H VP

V ApplP

IO Appl

Appl DO

In this configuration, the indirect object is not licensed by the verb, but the direct object is.
Imagine now that we introduce SE in a higher position, as in (511), and that it contains also
unvalued features. In (511), as the DO is already licensed, the DO does not intervene and SE
can agree with the IO.

181
 ANTONIO FÁBREGAS

(511) SEP

SE vP

v HP

DO H

H VP

V ApplP

IO Appl

Appl DO

This generates (512). Thus, SE would be SE already in syntax.

(512) Se lo di a mis hermanas.

SE it gave to my sisters

The example (474) above would require a bit more of technology, one where crucially
pronominalisation of the predicative argument is possible only when it moves above the DO,
but once that movement is granted, the configuration established would not differ from (511):
v would license the predicative argument and the SE projection would license the direct object.

(513) SEP

SE ...

... vP

v XP

PRED X

X HP

DO H

H ...

Note that this explanation could keep the point of view that clitics are agreement markers in
the verb provided that SE is not an agreement marker (cf. §11.4 below) but a head that agrees.
The second alternative building from Walcow (2012, 2013) would correspond to an analysis
where SE is an argument, not a head that agrees (§11.1 and §11.2 below). The use of SE, which
like the other clitics could not be considered an agreement marker in this analysis, would be a
strategy to avoid that the features that a head agrees with are not eaten up by the dative.

182
 SE IN SPANISH

Using a structure like Walkow's (2012), the crucial change in his analysis would be that the
DO does not cross over the indirect object –against the surface position, which suggests that
such movement takes place (514)–.

(514) a. Juan le dio el libro al chico. Neutral ordering

Juan him gave the book to.the child
b. #Juan le dio al chico el libro. Focalisation
Juan him gave to.the boy the book

(515) vP

v HP

H VP

V ApplP

SE Appl

Appl DO

However, there are other possibilities sharing the same intuition, as the one in Fábregas &
Cabré (2021), which we will sketch here. If we go back to §2.2-§2.4, where we presented a
Kaynean structure for the clitic area where each clitic is an argument of a designated head, the
game could be as follows –see §12 below for more details developing this type of approach–:
assume one head to license the accusative 3rd person clitics, containing person, number and
gender, and one head to license the dative 3rd person clitics, containing only a number feature,
as in (516).

(516) DatP

Dat

Dat AccP
[num]
Acc

Acc vP
[gend, num]
v ...ApplP

IO Appl

Appl DO

From (516), the dative would move first because it is closest to the clitic area. It would move
to spec, AccP, also because this is the lowest projection within the area. In spec, AccP it checks
its number feature –dative clitics do not have gender–. The problem is that, then, when the
accusative clitic moves to the clitic area, it cannot move to spec, AccP, because the number

183
 ANTONIO FÁBREGAS

feature has already been checked. It cannot move to spec, DatP, because this head lacks gender
features, and the accusative clitic does have gender features. The result would be a non
convergent derivation where the clitics have not been licensed.

(517) DatP

* Dat

Dat AccP
[num]
le Acc

Acc vP
[gend, num]
v ...ApplP

le Appl

Appl la

The solution could be to introduce SE instead of the dative clitic in the derivation. SE lacks
features, so it will not intervene between the DO and AccP.

(517) DatP

Dat

Dat AccP
[num]
la Acc

Acc vP
[gend, num]
v ...ApplP

SE Appl

Appl la

Naturally, these two approaches to spurious SE will have to be checked against the rest of
SE uses that we have revised in the previous sections, a task that we will perform specifically
in §12 and §13 below. Let us now close the discussion of spurious SE and move to the global
theoretical discussion about the nature of SE in the very different analyses that we have revised
until now. But before doing that we will summarise the main facts about clause-affecting SE
uses, as an interim summary.

10.3. Wrap-up: clause-affecting uses of SE

As expected, clause-affecting SE uses are not as messy as the predicate-affecting ones,
although they are also complex. There is a core function of SE where it is related with

184
 SE IN SPANISH

suppresing the overt expression of one of the arguments of the verb (the external argument),
which is shared by the impersonal, the passive and the middle SE. In the first of these, SE
appears apparently in place of the subject, while in the other two uses it relates to construals
where the internal argument gets subject properties, at least with respect to agreement.
The evidence examined in the previous sections suggest that impersonal SE is likely to be a
head that is controlled by an indefinite pronoun which eventually becomes the subject of the
clause; the passive SE shares a lot of those properties, and –with the exception of the available
positions for the clitic in auxiliary sequences– the behaviour of the two types of SE is parallel.
When it comes to middle SE, the evidence suggests that rather than being a modalised version
of the passive SE, it should be viewed as a modalised version of anticausative SE.
Spurious SE stands out within the clause-affecting SE uses in that it does not seem to
suppress an argument of the verb, or fill a syntactic function that gets blocked. Spurious SE
seems to exist because of the impoverished nature of the SE clitic, which almost lacks entirely
any feature content, which allows it to save constructions where two third person elements
would compete for licensing, either at a morphophonological or a syntactic level.
Once we compare clause-affecting SE uses with predicate-affecting SE uses, we can see that
clause-affecting uses do not alter the lexical meaning of the predicate. We only had to talk
about lexical meaning restrictions in two senses:

i) thematic external arguments: passive and impersonal SE need to be built from predicates
that have thematic positions for the external argument
ii) affected patients: middle structures require internal arguments of predicates such that the
internal properties of the internal argument can be meaningfully invoked to define facilitation,
necessity or propensity towards the event.

But largely the effects of clause-affecting SE uses are not lexical, but involve the
grammatical functions of the clause as a whole (subject, object, indirect object). In correlation
with this, none of the 4 clause-affecting SE uses can be claimed to be obligatory in the same
sense as the inherently reflexive verbs in §6.3 above are.
Interestingly, many of the analyses of impersonal and passive SE uses place SE in a position
shared with some of the predicate-affecting uses: Voice or v. This creates a problem, in
principle: if the clitic is introduced within the verbal complex, why isn't the lexical meaning of
the predicate affected, as we frequently saw in the case of the paradigmatic SE cases? There
are three ways out of this problem:

i) Clause-affecting SE corresponds to the uses of SE which are never lower than VoiceP,
which makes any analysis treating them in a vP position or lower incorrect. The domain of the
lexical verb, where idiosyncratic meaning can be affected, does not include VoiceP.
ii) Clause-affecting SE uses can appear within the material of the lexical verb, without
semantic effects. Note that the fact that lexical meaning can be affected by material introduced
within the syntactic domain of the verb does not mean that all material introduced in that
position has to affect it. In other words: even if the idiomatic meaning can be built with the
verb and the object (Marantz 1984), this does not mean that any combination of verb and object
must have an idiomatic meaning.
iii) Clause-affecting SE uses and predicate-affecting SE uses are not differentiated by the
position occupied by the elements in the structure, but by the feature endowment of SE, which
only in the latter case is able to affect lexical meaning.

We will get back to this problem in §12 below.

185
 ANTONIO FÁBREGAS

Another important aspect of clause-affecting SE uses is that the SE found there is always
non-paradigmantic, that is, fixed in a 3rd person form. In sections §7-§10 we have seen that
this fixation on the third form is plausibly epiphenomenal, and does not grant positing two
types of SE, one that is inherently paradigmatic and one that is inherently non paradigmatic.

i) Impersonal SE is third person because it is bound by a pronominal indefinite external

argument which must be third person, given its indefiniteness
ii) The same explanation can be extended to passive SE
iii) Middle SE is fixed to a third person because the subject must denote a class or be taken
as an exemplar of a class, and that bans personal pronouns of 1st or 2nd person as potential
binders of SE
iv) Spurious SE involves a clash between two third person pronouns competing for the same
licensing, so it will only emerge in 3rd person contexts.

With the possible exception of the spurious SE –which can be viewed as having the
strengthening function of licensing a second third person argument–, the clause-affecting uses
of SE also have an impoverishing function, which always affects the external argument. Here
we have to distinguish between three options: (i) possibility of expressing syntactically the
agent as a DP; (ii) possibility of expressing the agent prepositionally and (iii) possibility of
modifying or rescuing the agent in the semantic interpretation. The second dimension is case
marking in the internal argument. Finally, in the case of the spurious SE, there is a third
dimension which addresses its possible impoverishing function: reducing the expression of the
indirect object and, following Alcaraz (2017), solving WCO effects and being able to bind an
anaphor. The following table summarises these properties.

Table 6. Impoverishing function of clause-affecting SE

Type of SE Agent as DP Agent as PP Semantic Indirect Binder of
agent object anaphors and
expressed WCO effects
fully
impersonal No No Yes Yes Unaffected
SE
passive SE No Marginally, Yes Yes Unaffected
with
indefinite
and generic
elements
middle SE No No Yes Yes Unaffected
spurious SE Yes No (spurious Yes No (in the No
SE requires clitic; the PP
the presence is
of an unaffected)
accusative
clitic)

Finally, in the clause affecting uses of SE we have not seen the kind of messy discussion
between authors to determine which uses belong to one or the other class or which uses should
be collapsed together, with the exception of the proposals that want to reduce impersonal and
passive SE to one macroclass of SE structures where differences follow from case assignment.
Perhaps the two more controversial aspects are whether middle SE structures should be

186
 SE IN SPANISH

considered unaccusatives or not and whether they should be derived from anticausative or
passive structures. In the case of spurious SE, there is agreement that the problem that SE solves
is due to the similarity of two elements competing for the same 'slot' or licenser, and the
discussion is in essence whether the effect is morphological or syntactic.
As a summary of the empirical patterns, the following table summarises the effect of SE and
the main differential properties of the type of SE in all the constructions that we have analysed.

Table 7. Differential properties of SE types

Type of SE I II III IV V VI VII

1. Reflexive Yes Subject and Yes, with Yes Yes Yes No
(§3.1) DO / IO are a sí mismo
coreferential (but
remember
naturally
reflexive
verbs)
2. Reciprocal Yes Subject and Yes, with Yes Yes Yes No
(§3.2) DO / IO are uno P otro
coreferential
3. Anticausative Yes Absence of No Yes No No No
(§4) an external
agent
4. Dative Yes Affectedness, No Yes Yes Yes No
benefactive involuntary
(§4.4, §5) action
5. Aspectual (§5) Yes Telic No Yes Yes Yes No
component;
incremental
theme;
transition
6. Agent Yes Verb can No Yes Yes Yes Yes
reflexive only take
(§5.2) agents
7. Transitive Yes Verb has a No Yes Yes Yes No
(§5.2) different
meaning
8. Factitive Yes Subject is No Yes No No No
(§6.1) instigator
9. Antipassive Yes Internal No Yes Yes Yes No
(§6.2) argument
expressed as
PP
10. Inherent (§6.3) No -- No Depends Dep. Dep. Dep.
11. Figure Yes Agent ends No Yes Yes Yes Yes
reflexive up in a
(§6.4) location
12. Impersonal Yes Agent is No No No Yes No
(§7) indefinite

187
 ANTONIO FÁBREGAS

13. Passive (§8) Yes Subject is a No Yes, in No Yes Yes

patient number
14. Middle (§9) Yes Subject is No Yes, in No Yes No
patient; number
modal
reading
15. Spurious (§10) Yes A second 3rd Yes Yes Yes Yes Yes
person clitic
is present
I. Is there an alternating version without SE; II. What effect does SE have?; III. Can SE be doubled?; IV. Can
the verb agree?; V. Is expression of the agent syntactically possible?; VI. Is the agent accessible semantically?;
VII. Can some overt argument(s) be bare nominals?

At this moment we have finished the empirical discussion of SE and we are ready to move
to the broader questions affecting SE: its nature as an element (§11), the position or positions
that it should occupy (§12) and what assumptions should be made in order to provide a unified
analysis of all these structures (§13). Let us then begin the final tracks of this chapter.

11. Advantages and shortcomings of existing analyses (1): the nature of SE

The possibilties about how SE should be analysed as a linguistic object, and which are
presupposed by the different analyses that have been reviewed in §3-§10, are to some extent
limited by the analytical options available when analysing clitics. There are two main
possibilities in the market, starting from the standard assumption that SE is basically a clitic
(see §13 below too for this point):

i) SE, as an anaphoric reflexive, belongs to the system of clitics

ii) SE should not be fully integrated in the system of clitics in Spanish: it is a phonological
clitic, but its syntactic status is different from other clitics.

The option (i) can be further subdivided in two families of theories, that we will revise here.

ia) SE and the other clitics are arguments of verbs

ib) SE and the other clitics are agreement markers associated to verbal heads

These two families of approaches, as we will see in §11.1, allow for some mixture:

ic) Some clitics are arguments, and some clitics are agreement markers
id) SE can be alternatively part of a verbal head, as an agreement marker or not, and an
argument of a verbal projection.

§11.1 starts by showing the main syntactic theories about clitics in Spanish; §11.2 and §11.3
overview the theories that propose that SE is mainly an argument, details its strong points and
discusses its shortcomings given the previous discussion. §11.4 does the same with theories
where SE is part of a verbal head or even an agreement marker, including relevantly cases
where it should be considered a set of unvalued phi features.
The debates on the nature of SE as an element have to take into account one fact that, as we
anticipated in §3 above, is assumed to be true despite some apparent complications: reflexive
SE is the basic use of SE. This almost forces an analysis where SE is integrated with other
clitics, at least personal clitics which can alternate with SE depending on the properties of the
subject; the status of SE as an agreement marker or as a pronoun, then, depends on what status
is given to the other elements.

188
 SE IN SPANISH

11.1. Main theories about the nature of clitics: agreement or arguments

Since Kayne (1969), one of the most active debates in Romance linguistics is what status
object clitics should have. There are two different hypotheses:

i) Object clitics are agreement markers that combine with the verb (Strozer 1976; Aoun
1981; Jaeggli 1982, 1986; Borer 1984; Saltarelli 1994; Fernández Soriano 1989; Franco 1993,
2000; Sportiche 1996; Fontana 1993; Landa 1995; Barbosa 2000; Anderson 2005)
ii) Object clitics are arguments that need to cliticise to the verb (Kayne 1969/1975, 1989,
1994; Rizzi 1986; Uriagereka 1988, 1995; Roca 1992; Torrego 1995, 1998; Cardinaletti 2000;
Cardinaletti & Starke 1999; Roberts 2010; Nevins 2011; Ordóñez 2012)

As in other cases, the problem is complicated by the double factor that clitics do not behave
exactly in the same way in all Spanish varieties and that it is possible that the system of clitics
is mixed and contains members whose analysis should be (i) or (ii) (see for instance Bleam
1999, Ormazabal & Romero 2013, Alcaraz 2021).
The vision of clitics as agreement markers treats clitics as introduced to mark agree relations
in functional verbal heads (v, Voice, a more generic FP...), in parallel to how subject agreement
is introduced as part of the marking of the T head. Clitics would then play the same role as
agreement in (518).

(518) a. Nosotros vini-mos a la fiesta.

we came-1pl to the party
b. Nos vio a nosotros.
us saw DOM us

In the same way that -mos is compulsory when there is a subject, nos is compulsory when
there is a 1pl object. Cases where the clitic is present but there is no double can be treated as
instances of an empty pronominal category (519).

(519) La vi pro
her saw.1sg pro
'I saw her'

Thus, la in (519) like nos in (518) would be a morpheme that marks agreement with the
direct object. In English, where there are no clitic pronouns, her corresponds to the empty
pronominal and the clitic which should correspond to agreement is not present, in correlation
with the general typological property of English, a language with reduced inflection.
The prediction that this approach makes is double: (i) being agreement markers, the clitic
would not introduce referential properties of its own and (ii) doubling will always be possible
with clitics. The first property is supported by the behaviour of the person marked pronouns,
which do not produce Principle B infractions in the context of sentences like (520) (see §10.3
above, and Alcaraz 2021 for a detailed study).

(520) a. Yo me vi a mí mismo.
I me saw myself
b. Yo me tomé un café.
I me had a coffee

189
 ANTONIO FÁBREGAS

Remember that Principle B states that pronouns –thus, elements with referential features–
cannot have an antecedent that is also coreferential within their syntactic domain, the clause
for our purposes. As a personal pronoun is also referential, (520) should constitute infractions
to Principle B if the clitic had referential features because it is also a pronoun; as (520) are
grammatical, they cannot be infractions to Principle B and therefore these pronouns do not
have referential features. Thus, these clitics should be agreement markers, which lack
referential features of their own: there is no Principle B violation in (520) for the same reason
that there is no infraction in (518a).
With respect to doubling, if a clitic is an agreement marker, it should double another
argument compulsorily. This is what we see for personal pronouns in (521).

(521) a. *(Me) vio a mí.

me saw DOM me
b. *(Te) vio a ti.
you saw DOM you

The alternative theory proposes that clitics are in fact pronouns with referential features.
According to this theory, the clitic in (522) is in fact the argument that corresponds to the direct
object, and which has been incorporated or cliticised to the verb.

(522) Me vio.

The non doubling cases are explained directly: the object is present in the structure as the
clitic, and from that position, for either syntactic or morphophonological reasons, it cliticises
to the verbal head. Cases with doubling (523) require a technical innovation to be accounted
for:

(523) Me vio a mí.

As the clitic is not an agreement marker, (523) produces the problem that the clitic should
be coming from the same argumental position as the strong pronoun with a PP that appears in
postverbal position. In order to accommodate the two elements, Uriagereka (1995) famously
proposed the existence of big-DP structures, such as the ones in (524), where the element that
ends up being the clitic in fact originates from a head position of a DP that contains the double
in its specifier.

(524) DP1

DP2 D

D NP D ( XP...)
el niño le

Thus, a case of doubling like (525) would be an instance where the structure in (524) is
located in the indirect object position, where it gets dative case. The head D is spelled out as a
clitic le, and the double, the specifier of that clitic, also gets dative marking, through the same
operation. The double stays in place and the head D is moved to cliticise to the verb either in
syntax or in phonology.

(525) Le di un libro al niño.

190
 SE IN SPANISH

him gave a book to-the boy

The two predictions made by this approach are the opposite of the approach where clitics
are agreement markers: the pronoun carries referential features and the doubling is not
compulsory. Just like (524) is possible, a small DP could be present where the specifier is not
occupied and therefore there is no double.
The behaviour of third person clitics supports the view of clitics as pronouns coming from
argumental positions. Note that (526) is an infraction of Principle B, as expected if the clitic
carries referential features.

(526) *Juani loi ve.

Note also that doubling is not compulsory in (527); in the case of accusative pronouns, in
fact, doubling is impossible in general Spanish when the overt argument is a noun phrase, with
the exception of some varieties in Argentina or the Andinean region.

(527) a. (Le) entregué un libro a Pedro.

him delivered a book to Pedro
b. (*La) vi a la niña.
her saw DOM the girl

As we can see above, the evidence is not conclusive, and the predictions of one theory seem
to be better fit for some clitics, while other clitics support the alternative theory more strongly.
In general, the clitic-as-agreement theory finds stronger support in the behaviour of personal
pronouns and SE: if these pronouns are argument D heads, it comes as a surprise that they do
not produce Principle B violations. The clitic-as-pronoun theory would have to posit that these
clitics are in fact anaphors, coming also from argument positions, which lack referential
features. While this would not be surprising for SE, the person marked pronouns would also
have to correspond to this type of approach, where a clitic like me would have to be an
anaphoric element that is controlled by a higher first person pronoun in each case, also where
there does not seem to be one, as in (528).

(528) Juan me vio con María.

Juan me saw with María

This is not impossible, of course: theories like Sigurdsson (2000) have argued that person
features corresponding to the speaker and the addressee must be introduced at a very high
position, as discourse referents, in the CP area. However, some kind of technical tweak is
necessary to account for the clitic-as-argument theory, which ultimately will require retracting
from the position that all clitics carry referential features.
On the other hand, the behaviour of third person pronouns, specially the accusative clitics,
seems to support strongly the clitic-as-argument approach and posit initial complications for
the clitic-as-agreement approach: they produce Principle B infractions. The clitic-as-agreement
side of the debate will have to propose that the Principle B infraction is caused by the presence
of a pro argument that cannot be coreferential with the subject. However, this is complicated
by another fact, the difficulty of doubling in most Spanish varieties when it is a nominal and
the obligatoriness of doubling when it is a pronoun.

(529) a. (%Lo) vi a Juan.

him saw DOM Juan

191
 ANTONIO FÁBREGAS

b. *(Lo) vi a él.
him saw to him

(529a) is unexpected from the clitic-as-agreement side, which has to somehow determine
that the argument must emerge as pro and not as an overt syntactic constituent when in the
direct object position, triggering accusative agreement; (529b) is on the other hand unexpected
in the clitic-as-argument approach. To be fair, the clitic-as-argument side should also explain
why a big-DP with an overt nominal specifier is also banned in their approach for these contexts
(529a), and why the double must compulsorily appear when it is a pronoun (529b), so one
could say that the doubling cases are problematic for both theories in principle –although the
big-DP approach could argue that there is a size difference between pronouns and nominal
doubles that makes the first necessary and the second impossible with the clitic–.
As can be seen, then, the system of clitics in Spanish does not seem to be homogeneous for
the purposes of doubling or Principle B. This has triggered hybrid approaches where some
clitics in Spanish are considered to be agreement markers and other clitics are considered to be
pronouns corresponding to potential big-DP structures. Let us mention two recent cases of this
hybrid analysis.
Ormazabal & Romero (2013) propose that the accusative clitics are arguments, while the
dative third person clitics and presumably also the person marked clitics are agreement
markers. Not assuming a bid-DP as an option (at least in the varieties of Spanish that cannot
double) automatically explains why the presence of the accusative clitic prevents any other
overt constituent that corresponds to the direct object. Ormazabal & Romero (2013) further
note that if the accusative clitics were agreement markers of the verb, that would be the only
case in Spanish in which a finite verb would agree in gender with an argument: remember that
the accusative pronouns contrast in gender (§2.2 above). In contrast, dative clitics, as well as
the person marked clitics and of course SE, have at most person and number features, which
correspond to the morphological properties that verbs agree with in the case of subjects. In this
analysis, dative clitics are agreement markers that signal that the verb has assigned object case
to an argument, while accusative clitics are incorporated as arguments to the verb –
consequently, they essentially don't need case to be assigned to them, a point made by
Ormazabal & Romero in other articles–.
Alcaraz (2021) has a slightly different take on the issue: he strenghthens the correlation
between agreement marker and not having referential properties. He notes, contra Ormazabal
& Romero (2013) that the dative clitic is subject to Principle B:

(530) *Juani lei dio un libro.

Juan him gave a book

Therefore, all third person clitics would be pronouns, and the agreement markers would only
be the person-marked clitics and SE. An approach that treats the dative clitic as an agreement
marker must treat (530) as a problem that comes from the presence of a pro that triggers the
Principle B infraction, roughly as (531).

(531) *Juani le dio proi un libro.

However, if that pro is present, (531) should have the same type of interpretations as a
construction like (532): pro equals the overt pronoun introduced by the PP, and the clitic does
not appear because given the presence of a preposition there is no agreement with the verb.

(532) Juanj confía en éli.

192
 SE IN SPANISH

Juan trusts in him

'Juan trusts him'

(532), however, allows coreference between the subject and the pronoun if the subject is
focalised (Büring 2005).

(533) Solo JUANi confía en éli.

only Juan trusts in him
'Only Juan trusts Juan'

In contrast to this, a construction like (530) cannot become coreferential through focalisation
of the subject.

(534) *Solo JUANi lei dio un libro.

only Juan him gave a book
Intended:'Only Juan gave a book to Juan'

Thus, (530) does not contain a pro, because then we would expect the same type of binding
with focus as in (532). One has to conclude that the dative clitic is referential by itself, and
derive the contrast from –presumably– the syntactic position where the clitic ends in contrast
to the position where the pronoun embedded in a PP is located (although see Alcaraz 2021: 85
and folls. for a semantic proposal).
These are the two main options with respect to any clitic in Spanish then, as an argument
and as an agreement marker. In the next sections we will explore what these options mean for
the analysis of the SE constructions noted above.

11.2. SE as an argument (1): overview and integration within the system of clitics
In sections §3-10 we have seen several examples of analyses where SE has been treated as
an element introduced in argument position, and which we will highlight here. The SE-as-
argument analysis has been proposed in constructions that have at least one of the following
properties:

a) One argument of the verb has been removed

b) An argument that without SE can occupy a position X cannot use that position in the
presence of SE
c) The case-assignment capabilities of the verb –accusative– have become inactive in the
structure.

The constructions that support this most strongly are those where all these properties
combine at the same time, as the following:

a) Antipassive SE (§6.1), where Masullo (1992) claims that SE occupies an argument

position that forces PP projection of what otherwise would be the argument:

(535) a. Aprovechar algo.

to profit something
b. Aprovecharse de algo.
to profit-SE from something

193
 ANTONIO FÁBREGAS

b) Passive and to some extent anticausative SE structures –thus, also middles–, where the
external argument is not projected. We have not seen many examples of approaches where SE
there occupies an argument position –most analyses presented in §4 and §8-9 above argue in
favour of a projection of SE in the head of a verbal projection. However, one could in principle
argue for a derivation for passives where SE, like the impersonal SE analysis in Ormazabal &
Romero (2019) presented in §7.5, occupies the position of the external argument (536).

(536) Voice / vP

SE Voice / vP

Voice / v ...

Possibly, the reason why this approach is not adopted generally is that from this position
one does not expect SE to absorb accusative case, as it does not do it in the case of the
impersonal; one has to combine this approach with a proposal where passive and impersonal
SE vary independently on the option of assigning DOM case to the object, which is what these
authors propose anyways.
In a more standard account of case, one could perhaps propose that SE is located in the
specifier position but forces accusative case absorption, blocking the internal argument from
getting standard accusative case (only DOM can be assigned as a case distinct from nominative,
as in Ordóñez & Treviño 2016).
In the case of anticausatives, where there is no inference of the existence of an external
argument, one cannot propose literally the derivation in (536) above, if that one conditions the
interpretation that there is an external argument position which is however not referentially
independent. There are three options that could be available:

(i) À la Schäfer (2008) and Alexiadou, Schäfer & Anagnostopoulou (2015), the head
involved in anticausatives and passives differ on whether an agent theta role is assigned to the
specifier. One would have to assume, though and contra these authors, that SE despite being
referentially null could satisfy the agent theta role for the passive.
(ii) Voice and v are distinct, with Voice used to define the nature of the external argument
(or its absence), and little v used to define the type of event. In this theory, the difference
emerges from the distinction between the type of vP that is merged below Voice: a vP
corresponding to CAUSE or DO triggers the passive construal, and a vP corresponding to
BECOME triggers the anticausative construal. In an approach like Ramchand (2008) this could
correspond to having InitP as the complement of Voice or having Proc as its complement.
(iii) The distinction is purely semantic, follows from the interpretation of the verbal stem
and whether it can be conceptualised as internally caused or not, and the structure above works
for both cases.

c) The third construction that could be argued to show a SE occupying an argument position
is the impersonal, which Ormazabal & Romero (2019) argue to be an argument that necessarily
ends up in a subject position. Remember the discussion in §7.5 above, where we noted that this
approach is successfull in accounting for the human interpretation of the subject and the
distribution of the SE clitic across infinitives, including periphrastic structures, but where it
wrongly predicts that SE should rise in the context of parecer 'seem'.

d) Part of aspectual SE, specifically the cases of intransitive verbs where Sánchez López
(2002) and García Fernández (2015) have argued that SE turns the verb from unaccusative to

194
 SE IN SPANISH

unergative (remember 537) could be cases of SE introduced as an argument in the internal

position, forcing the overt nominal subject to be merged in an external argument position:

(537) a. Sale humo.

comes.out smoke
b. Se sale *(el) humo.
SE comes.out the smoke

(538) a. vP (possible unaccusative construal)

v VP

V NP / DP

b. vP (possible unergative construal)

DP v

v VP

V SE

e) Transitive SE (Armstrong 2013), where SE has the effect of forcing the projection of the
internal argument as a DP and potentially alters the argument structure of the predicate, could
be treated as an argument that occupies the complement position and forces the internal
argument to be a specifier.

(539) a. La mesa se come la habitación.

the table SE eats the room
'The table occupies the whole room'
b. *La mesa se come habitaciones.
the table SE eats rooms
Intended: 'The table occupies any room, no matter which one'

(540) a. vP (version without SE)

Juan v

v VP

V NP / DP

195
 ANTONIO FÁBREGAS

b. vP (transitive SE version)

Juan v

v VP
[num]
DP V

V SE

vi) In general, SE would be in an argument position in all analyses where SE fills a

participant position created by extra structure, which then has to be coreferential with the
subject or another argument, such as several analyses of aspectual SE which treat it as related
to extra verbal heads, or the analysis of figure SE.

In contrast, the types of SE that do not alter case and / or argument structure would be the
ones that seem less compatible with an analysis of SE as occupying an argument position: the
aspectual SE with transitive predicates is problematic for this approach unless one adopts a
version of MacDonald (2017) where aspectual SE has the same effect as (540b) in forcing the
internal argument to be in a specifier position, forcing it to be introduced as a DP –remember
that MacDonald (2017) locates SE in a Voice position, though–. Agentive SE (§5.3) cannot
have any obvious effect in the interpretation of the external argument from a low position, and
it cannot be introduced in the position of the external argument because it does not block its
presence. The same goes for factitive SE and spurious SE.
One of the most controversial constructions from the perspective of whether SE can occupy
an argument position is reflexive and reciprocal SE. The problem is the following: on the one
hand reflexive SE blocks the introduction of an argument that is independently referential,
which suggests that SE might occupy the position of that argument (541), but on the other hand
many reflexives can be doubled, which means that if that double is possible, SE cannot be
occupying that position. (542) is thus not compatible with (541) unless other assumptions are
made.

(541) vP

Juan v

v VP

V SE

(542) Juan se vio a sí mismo.

Juan SE saw DOM himself

Remember from our discussion in §3 that this creates two main analytical possibilities for
SE, both of them treating SE as occupying a verbal head position: (i) SE marks a type of
predicate that is monoargumental and removes the external argument, so the internal argument
moves to bind SE (Embick 2004) and (ii) SE marks a monoargumental predicate that leaves
one internal argument position unsaturated where the double can be introduced as an expletive,
non referential element (Labelle 2008). The problem that (i) encounters, remember, is that there
is no syntactic position for the double.

196
 SE IN SPANISH

The analysis of reflexive SE has to account for the fact that reflexive SE is only possible for
dative and accusative elements in Spanish. Labelle (2008) proposed that this is because SE can
absorb accusative or dative, but not other cases. An alternative approach to this problem would
be that prepositionally marked reflexives are arguments of the preposition, not the verb, which
makes it impossible that SE marks in the verb a property that does not refer to its argument
structure.
There is a developed analysis of reflexive SE, which we sketched in §5.2, where reflexive
SE occupies an argument position: Kempchinsky (2004). In her approach one has to
differentiate between doubling and non-doubling reflexive structures.

(543) a. Juan se lava.

Juan SE washes
'Juan saw himself'
b. Juan se vio a sí mismo.
Juan SE saw DOM himself

In the first case, SE is a simple DP which is introduced in the spec, AspP position. It does
not get case, but it intervenes between the internal argument and the verb, not letting it get case
from the verb, which means that those arguments cannot be introduced inside the VP: if present
together with SE, they would not get any case. Note that this analysis has to assume that dative
case is assigned by vP, not by a preposition or another head, like an applicative.

(544) vP

v AspP

SE Asp

Asp VP

DP V

V DP

Note that this analysis encounters problems in explaining a sentence like (545), where the
indirect object is coreferential with the subject but there is an overt direct object: who assigns
case to the direct object?

(545) Juan se mandó una postal.

Juan SE sent a postcard

It is true that (545) is perhaps marginally worse than the structure with doubling (546). Thus,
Kempchinsky (2004) restricts the configuration in (544) to cases of naturally reflexive
predicates where doubling is not possible, and proposes that the cases with doubling, like (546)
involve a different structure.

(546) Juan se mandó una postal a sí mismo.

Juan SE sent a postcard to himself

197
 ANTONIO FÁBREGAS

Kempchinsky (2004), in fact, proposes that the structure with doubling is different. In those
cases, SE is not base merged in spec, AspP, but is part of a big-DP structure that is introduced
in the corresponding argument position (see also Torrego 1995).

(547) DP

(a) sí D

D AgrP
SE
mismo Agr

Agr XP

sí mismo

A non-naturally reflexive complement without the double would still correspond to (547),
only that the complement of Agr would be pro. The complex structure in (547) could be
generated in a VP position and move from there, if necessary, to the position where internal
arguments get case from vP.
One could propose a parallel derivation for reciprocal uno P otro where the elements also
originate low and move up passing through an agreement projection.
To the extent that we understand this approach, there are two problems with it: the first is
that naturally reflexive verbs can also have direct objects, which cause the problem that, if they
correspond to (544), it should not be possible to assign case to the object –note that any account
that saves (548) by licensing case of the object cannot explain why the object cannot be overt
in (543a).

(548) Juan se lava las manos.

Juan SE washes the hands

The second problem is how the approach restricts SE structures to cases where the internal
argument is not a direct or indirect object. Of course, trivially one could say that SE only
associates to dative, accusative and perhaps nominative, but this brute force procedure is
somehow unsatisfactory. In principle nothing should block a structure like (549). In the case
of the doubling structure in (547), note that it could be possible to argue that when this structure
is embedded under a PP, SE cannot cliticise to the verb, and this creates a problem at the
phonological surface structure.

(549) a. vP

v AspP

SE Asp

Asp VP

V PP

P DP

198
 SE IN SPANISH

b. *Juan se confía en él.

Juan SE trusts in him

(550) vP Structure where SE cannot cliticise to v

v AspP

Asp

Asp VP

V PP

P SE

Thus, there are empirical problems for the approach where reflexive SE occupies an
argument position. In this account note that we have not mentioned the problem of the
referentiality of SE which we introduced in §11.1 in the context of the discussion of what
Spanish clitics are. Remember that SE does not introduce referential features of its own. The
next section will discuss a possible way out from the perspective of the SE-as-argument
approach to this problem.

11.3. SE as an argument (2): if argument, what are its features?

Remember that SE does not produce Principle B infractions in the relevant contexts,
something that supports for some authors that SE is an agreement marker and does not occupy
an argument position.

(551) a. Juani sei lava.

Juan SE washes
b. Juan se vio a él.
Juan SE saw DOM him

The traditional account of this problem is that SE is an anaphoric element which must
comply to Principle A, that is, SE does not introduce referential properties of its own and for
this reason it never triggers Principle B infractions. The crucial idea, then, is that SE does not
trigger the effect not because it is an agreement marker vs. an argument, but because its feature
endowment is different from the other pronouns. Thus, it could be located in an argument
position without having effects for reference.
Leaving the problems noted in §11.2 above for approaches where reflexive SE occupies an
argument position, one way of solving the problem would be to treat Spanish anaphoras as sets
of nominal features lacking a DP layer that introduces reference. Déchaine & Wiltschko (2020)
propose a formal typology of reflexives that uses this type of account. In their analysis, they
assume the existence of (at least) three areas in the structure of nominals, which in (552) are
represented as single heads although in their theory these could be expanded in richer
sequences of heads.

199
 ANTONIO FÁBREGAS

(552) DP

D PhiP

Phi NP

English reflexives would be DPs that occupy argument positions. French reflexives, as a
representative of Romance SE, would in contrast be sets of phi features, and would not
introduce any referential properties.
We must however emphasise that if a set of phi features can occupy an argument position
depends on further assumptions about argument structure. Déchaine & Wiltschko (2020) in
fact propose that French SE occupies a head position (Voice), following Labelle (2008),
assuming that only referential elements can saturate argument positions. In contrast, if one
assumes (for instance, following Chung & Ladusaw 2003) that non-D arguments are possible
and can restrict the predicate, one could still maintain that Romance SE is in an argument
position that cannot be saturated. That said, let us examine the tests offered by these authors to
claim that SE lacks D properties:

a) Equative copulative sentences in English, where the anaphora is DP, can be built with
reflexives, but not in Romance, here illustrated by Spanish.

(553) a. He will be himself.

b. *Él será se. // *El se será.
he will.be SE he SE will.be

This follows if the equative requires a DP (554) and the reflexive cannot be a DP in Spanish.
Note that the strong pronoun can do it.

(554) a. *He will be boss.

b. Él será él mismo.
he will.be he self

b) In English, anaphors are built with possessive morphology, which follows if possessors
are built as DPs; this is not the case in Romance.

(555) myself / my book

Once established that SE is not a projection of DP the problem of referentiality dissolves for
the theories that wish to introduce it in argument positions. However, there are two additional
questions here:

i) Are those phi features valued or unvalued?

ii) Which nominal features does SE have?

In principle, we would naturally expect that SE occupies an argument position if its features
are interpretable. If the features are uninterpretable, then, we would be retracting to the position
where SE is an agreement head, perhaps as the other clitics or as a subset of the other clitics.
The idea would be then whether reflexive SE has features that are valued or unvalued.
If we go back to §2.2 above, the evidence tells us that SE is an impoverished element that
does not contrast in gender or number:

200
 SE IN SPANISH

(556) a. Él se vio.
he SE saw
b. Ella se vio.
she SE saw
c. Ellos se vieron.
they SE saw
d. Ellas se vieron.
they SE saw

In terms of person, paradigmatic SE does contrast with personal pronouns, but that person
specification depends on the features of the subject, which makes it likely that they are non
interpretable.

(557) a. Yo me veo.
I ME see
b. Tú te ves.
you YOU see
c. Él se ve.
He SE sees

SE and personal pronouns can co-occur, which shows that as a clitic it does not even specify
a person feature in itself. This is the case for instance in anticausative structures involving a
causer dative (§4.2).

(558) Se te rompió.
SE you broke
'You broke it by accident'

Now, this suggests that SE is not able to have any interpretable feature, not even one for
third person. This coincides with Reuland's (2011) proposal that (in general, not proposing
typological differences) treats anaphors as bundles of uninterpretable nominal features. The
evidence suggests that Reuland is right at least for the case of Spanish SE, and the anaphor has
to be a bundle of non interpretable phi features. This casts serious doubt on any analysis that
treats SE as occupying an argument position.

11.4. SE as (part of) a verbal head (3): overview and integration within the system of clitics
Thus, let us now revise how well theories that treat SE as a projection of a verbal head fare
with respect to the different types of SE structures.
These theories have in their favour a very basic and solid fact about SE, that we particularly
focused on in §3 and §4: SE does not appear in any reflexive predicate, but only in verbal
predicates that are reflexive (if other conditions are met); SE does not appear in any
anticausative, but only on the verbal ones; similarly, the other uses of SE only appear in verbal
structures. This follows if SE is the manifestation of a verbal head, but it is less clearly
accounted for if SE is an argument, as other grammatical categories can take arguments. An
approach that treats SE as an argument and wants to block that it may appear with non verbal
categories has to posit a second type of restriction, for instance that SE as a clitic needs a verbal
host.
The cases that are most amenable to the type of analysis where SE is a verbal head are those
that have the following properties:

201
 ANTONIO FÁBREGAS

a) SE restricts the interpretation of arguments, not suppressing any of them

b) SE is related to a structure where the verb or tense are interpreted as carrying a meaning
or function that is somehow impoverished

This makes agentive and factitive verbs more amenable to the SE-as-verbal head analysis,
where the role that it plays seems to be to impose a particular interpretation to an argument that
is not suppressed. For different reasons, the impersonal SE would fit –inflection is
impoverished, and inflection is assumed to be related to a functional head–. If one takes SE to
be an impoverisher, the constructions where the external argument is removed syntactically
and / or semantically would also fall here: passive, anticausative and middle. The question for
these approaches would be which property is the verbal head marked with SE sensitive to. The
main options here are the following:

a) SE marks a reduced verbal head in the sense that the external argument is either not
present or restricted to indefinite semantics
b) SE marks a reduced verbal head in the sense that accusative assignment is not possible
c) SE marks a verbal head that is sensitive to the scalar properties of the predicate

Similarly, the need to mark the verbal head with SE would be influenced by the lexical
specification of the predicate in a case like the anticausative: some verbs would need SE to
impoverish or mark the impoverishment of the verbal head that the verb carries, and some
would not because that head is already impoverished enough in the SE-less version of the verb
–the approach can also be extended to TP, following Otero (1986), where infinitives are already
impoverished–. Alternatively, the presence or absence of SE marking would be related to other
properties, but the logic would be the same: which verbs already express that meaning alone
and which verbs lack it and require insertion of SE to mark the meaning.
When it comes to reflexive SE above, an unaccusative analysis would relate them to passive
SE structures, but even in an analysis like Legendre (2008), where the reflexive is unergative,
the SE would have an impoverishing function in the verbal structure, because it would leave
an internal argument position unsaturated. Note that this move, that reflexive is a particular
specification of a verbal head, is necessary in the approach that treats SE as not being a
pronoun: there is overwhelming evidence that the reflexive is historically the origin of the other
SE uses (see §12 below), so if the reflexive would involve a SE in argument structure it would
be a surprise that the other uses have necessarily grammaticalised it as a verbal head.
The most problematic cases are, of course, those that involve the apparent addition of an
argument that has the effect of occupying an argument position of the predicate. This includes
the antipassive SE, the transitive SE and the apparently aspectual uses of SE. In the first case,
the agent is not affected –just like in the case of agentive SE– and no internal argument is
removed; in the case of the aspectual SE there is no apparent impoverishing effect, and the
result is just that an aspectual reading is imposed.
In general, the problem for analyses where SE occupies the position of a verbal head, as
agreement or not, is that the effect of SE is not always the same in these analyses, and
consequently that the role that SE plays can in fact be considered as contradictory:

i) In anticausative, middle, impersonal and passive SE the role seems to be to restrict the
reference of the external argument or to remove it
ii) In anticausative, middle, passive and antipassive SE the role affects the case assignment
iii) In agentive, factitive and aspectual SE the role is to affect the meaning of already existing
elements, such as the external argument or the verbal event

202
 SE IN SPANISH

iv) In spurious SE the role is to reduce the expression of a dative

More to the point even, in some aspectual SE uses –García Fernández (2015)– the role of
SE would be to move from an unaccusative to an unergative construal, marking in fact a verbal
head as introducing an external argument (perhaps like agentive SE), while in passive SE the
role would produce a predicate whose distribution is closer to an unaccusative verb –see also
the anticausative SE–. Thus, the main challenge for this approach is precisely that the roles that
SE would produce in those verbal heads would be too varied to grant a unification between all
SE uses.
At the same time, proposing that SE is in a head of the set that builds a verb does not always
equal the claim that it is an agreement head or is manifesting agreement. In the case of spurious
SE (§10 above) we showed that an analysis where SE acts as an agreement head is possible,
and would solve the problem of how two 3rd person arguments are licensed as verbal
arguments, but in other cases the role as agreement is less clear. Perhaps one could treat
reflexive SE as the manifestation of a head that agrees compulsorily with the external argument
provided that such agreement does not license it with respect to case, or as agreement of the
verb with a pronominal that is introduced in argument structure. However, the same type of
approach is less likely for the verbal constructions where an argument is at least removed as an
overt nominal, such as anticausative, impersonal and antipassive ones, where the agreement in
the relevant verbal construction does not preclude the subject from having to agree or move to
a TP layer in the clause. The role of SE as a merely agreeing head –a head introduced with the
sole purpose of agreeing– is also less likely in structures where SE does not introduce or reduce
arguments, where again it is unclear why extra agreement would be necessary in the
configuration, with respect to the version without SE. The only case, then, where it is clear that
SE is used as a probe that licenses through agreement something that would not be possible to
license otherwise is spurious SE, then.
Thus, to conclude this section, the problem for analyses that treat SE as the manifestation of
a verbal head is the extreme diversity of roles that SE would mark in such heads, which
sometimes involve contradictory requisites. There is only one case where clearly SE is
introduced to add additional agreement. SE, from this perspective, should not be equated with
an agreeing head used as a pure probe that always licenses an element that otherwise could not
be licensed, but rather with an element that might be introduced in the derivation for other
reasons and that happens to trigger agreement. There seem to be cases where SE needs to be
taken as an argument and cases where SE needs to be taken as a verbal head, and each of the
two approaches is able to explain part of the data: maybe one has to assume the existence of
two SEs, one that is a PhiP argument with perhaps valued features and one that is a head that
contains unvalued features.
An additional problem that is present in this section is whether SE can be integrated within
the system of clitics in Spanish as an element on a par with le, me or lo, or whether it is an
element of a different nature that happens to be cliticised to the verb and display a surface
position that we expect of clitics. An approach that treats SE as a pronoun can integrate it with
other clitics, if these are arguments, and an approach that treats it as an agreement marker can
integrate it within the system, assuming that clitics are also agreement markers. However, we
have seen here two options that are not necessarily advocating an integration of SE within the
system of clitics.

a) SE is a marker of verbal heads

b) SE is a probe that introduces an additional head that can license an argument

203
 ANTONIO FÁBREGAS

The first option would not treat SE on a par with other clitics unless it is the manifestation
of agreement in the verbal head –a position that is problematic– or unless it is treated as a
pronoun that satisfies a property of the head. The second option would in principle be
compatible with the approach of clitics as agreement heads, but note that in this case the object
agreement would not be located in the verb, as standardly assumed in these approaches, and
SE would introduce its own agreement layer independently of the verb.
All in all, the approach where SE is the manifestation of a verbal head seems to be more
supported by the empirical facts –lack of reference being the crucial component–, but it has the
problem that the functions of SE are too diverse. One possible way out in the approach that
tries to unify SE as an element projected within a verbal head is to play around with the position
of SE in the structure. However, this carries its own problems, which we will revise in the next
section.

12. Advantages and shortcomings of existing analyses (2): configurations of SE

If in the previous version we concentrated in the problem of whether one can provide a
unified characterisation of SE from the perspective of its nature –argument or verbal head–,
and once done so if the roles played by it can be unified, in this section we will address a
different problem, namely how one can express the fact that SE is always linearised in the same
position within the clause if SE builds different types of configurations.
Remember the facts, that were introduced in §2.4: all SE types linearise as the first clitic in
a sequence, irrespective of the type of SE that they express.

(559) a. Se le entregó un libro. Passive

SE him delivered a book
b. Se le presentó a los invitados. Impersonal
SE him introduced DOM the guests
c. Se le rompió el libro. Anticausative
SE him broke the book
d. Se le cayó encima. Aspectual
SE him fell on.top

Now, how can an approach account for this fact? We will see that standard analyses are
required to always violate a well-established assumption in linguistics to provide a unified
account of SE despite the difference in the configurational and semantic properties of the uses
of SE. The problem is complicated by one single case where one type of SE has a different
position from the others, which is the impersonal SE in combination with auxiliary verbs.

(560) a. Se pueden encontrar estos pisos. Passive

SE can.3pl find these houses
b. Se puede encontrar a esas personas. Impersonal
SE can find DOM these people
(561) a. Pueden encontrar-se estos pisos. Passive
can.3pl find-SE these houses
b. *Puede encontrar-se a estas personas. Impersonal
can find-SE DOM these people

However, this affects the ordering within the verbal group, not the ordering with respect to
other clitics.

204
 SE IN SPANISH

(562) a. Se le pueden encontrar estos defectos. Passive

SE him can find these defects
b. *Se pueden encontrar-le estos defectos.
SE can.3pl find-him these defects
c. Pueden encontrar-se-le estos defectos.
can.3pl find-SE-him these defects
(563) a. Se le puede encontrar a estas personas. Impersonal
SE him can find DOM these people
'One can find these people for him'
b. *Se puede encontrar-le a estas personas.
SE can find-him DOM these people
c. *Puede encontrár-se-le a estas personas.
can find-SE-him DOM these people

The approaches that attempt a unification explicitly can take as a starting point the proposal
that there is a configurational ambiguity that is possible due to SE's impoverished structure
(Otero 2002): given the small feature endowment that SE has, it is an element that is ambiguous
between a maximal and a minimal projection. Not taking any compulsory complement, in such
approaches, SE is at the same time the maximal and the minimal element in its projection,
following Bare Phrase Structure (Chomsky 1995).

(564) XP (=X)

X
se

Thus, from this perspective across the derivation SE can act as an XP that satisfies a non-
head position and displace as an XP, or act as a head that attaches to another head or projects,
and moves like an X. This is in fact what Kempchinsky's theory of SE proposes (2004, 2006).
Moreover, the theories can also take advantage of the small feature inventory of SE to allow it
to appear in a variety of syntactic contexts making a minimal contribution, which makes SE be
flexible in the types of configurations where it can appear.
Interestingly, none of the existing theories where the unification is attempted explicitly take
as a starting point the grammaticalisation path that was sketched in §3 above, and which we
repeat here (565).

(565) reflexive > anticausative > passive > impersonal

The reason is that these theories are concerned with the synchronic representation of the
element in contemporary Romance; note that it is almost inescapable to accept that, irrespective
of the historical evolution, the representation of SE in the mental grammar of the speaker is not
reflexive –thus explaining why SE is not a necessary nor sufficient condition for reflexivity in
contemporary Spanish–. The alternative would have to be to accept the existence of
homophonous SE elements, a possibility that is not completely impossible (see Otero 2002,
where he notes that the phonological structure of SE is the simplest one, one single CV syllable
with the most underspecified consonant /s/ and the most underspecified vowel /e/), but that is
in principle dispreferred due to theoretical considerations.

205
 ANTONIO FÁBREGAS

12.2. Kempchinsky's unified theory of SE

Kempchinsky's analysis of SE treats it in fact as an aspectual element as its basic use. As a
unit, it always has the same function: to link, suspend or introduce a subevent.
Following Déchaine & Wiltschko (2002), this author arrives to the same conclusion in
(2004, 2006) than Déchaine & Wiltschko (2020) obtain for Romance SE: it is a bundle of phi
features without any DP layer. The proposal that SE is a bundle of phi features allows
Kempchinsky to assign to it two potential syntactic roles that it could not have had otherwise:
it can be in an argument position, where presumably these features are already valued, and it
can be in a predicate position, where presumably the features would be unvalued. When SE is
an argument, it links to a subevent; when it is a head, it suspends or introduces a subevent.
This, combined with the structural ambiguity that SE exhibits given its impoverished nature,
as simultaneously XP and X, allows Kempchinsky (2004, 2006) to treat SE as alternatively
specifier or head. Here is her own proposal about the different uses.

(566) vP

SE v
(passive)
v AspP
SE
(middle) SE Asp
(natural reflexive)
Asp VP
SE
(aspectual) SE V
(double reflexive)
V ...

Note first that there are some uses that are not taken into account here: impersonal, which
Kempchinsky does not discuss but which presumably could be seen as a version of passive SE,
with differences coming from case assignment, as the theories in §7.5 do; spurious, which she
does not treat at all, and some of the minor uses (agentive, factitive, antipassive...) which –
although she does not say– she presumably views as not real classes which are subcases of the
other structures.
The double reflexive SE is a case where SE introduces a big-DP that is merged in VP and
which moves to spec, AspP to link to the process subevent of the verb. The single reflexive is
base merged in that position, and therefore blocks insertion of the internal argument in VP –
with the problems noted in §11 above–. The aspectual SE introduces a telic change of state
component and (remember §5.2 above) can trigger an anticausative or an aspectual construal
depending on whether vP is projected. Middle SE is introduced as the head v, where it cancels
the causation subevent, preventing the presence of an agent, and where the modifier occupies
the position of spec, AspP, triggering a stative reading. Finally, passive SE is itself the agent,
and links to the causation subevent.
This approach is probably the only explicit one that tries to unify all SE uses in Spanish
under one single element, playing around with the ambiguity of SE between XP / argument
and X / predicate. The problem that it produces, in fact, has to do with the position of SE in the
sequence of clitics, which has to be the same in all cases. Let us see why.
The cases where SE is in an argument position require necessarily that SE moves to a
position that is higher than the rest of clitics, given the facts in §2.4, where SE is always the
highest element in the group.

206
 SE IN SPANISH

Note that this movement has to be directly to the highest projection within the clitic area if
clitics are generated as arguments, according to the clitic-as-argument theories revised in §11.1.
The evidence in Spanish seems to suggest that at least the third person accusative clitics are
such arguments. The fact that accusative clitics are more internal to the verb than dative clitcs,
personal clitics or SE needs an ordering of elements where the accusative clitics are merged
(or move) lower than SE. Remember the structure presented as (49) in §2.3 above, represented
as if each clitic was an argument.

(567) ... X'''P

se X

X''' X''P

me X''

X'' X'P

le X'

X' XP

lo X

X ...vP

In a structure like (567), the verb would move to X and the clitic will attach to it, or
alternatively through a roll-up movement that keeps the clitic in the left margin through
phonological conditions; a subsequent movement of the verb to X'' would cyclically attach the
person clitic externally, and then SE after presumable movement of the verbal head to X''',
generating (568).

(568) Se te lo dijo.
SE you it said
'Someone said it to you'
(569) a. [XP lo X+v [dijo] ...]
b. [X''P me X''+X+v [lo + dijo] ...]
c. [X'''P se X'''+X''+X+v [me + lo + dijo]]

This means, for a theory like Kempchinsky, that SE, even when it is base generated in a
high position and combined with a proper direct object (570), cannot prevent movement of the
direct object clitic to XP and will move directly to X'''P despite the presence of a lower head
that in principle could license it.

207
 ANTONIO FÁBREGAS

(570) X'''P

SE X'''

X''' XP

lo X

X vP

SE v

v AspP

lo Asp

Asp ...

The possible solution would have to claim that all X heads except X''' attract elements that
are DP, with SE being the only element that is not a DP; in this way, the accusative clitic will
not have SE as an intervener and X''' would be the only head that can attract SE.

(571) X'''P

SE X'''

X''' XP
[uphi]
lo X

X vP
[uD]
SE v
[phi]
v AspP

lo Asp
[D]
Asp ...

Note that the problem would not be directly solved if Kempchinsky treats clitics as
agreement markers, unless that is complemented with the assumption that SE is different from
the other clitics. If we assume, with Walkow (2012, 2013), that clitics reflect the ordering of
the agree operations with vP, SE is still closer to vP than the accusative argument, so SE should
linearise more internal than the accusative clitic.
Once the assumption in (571) is made, and SE is unlike the other clitics, the facts in (560)-
(561) can easily be explained if SE in impersonal contexts must move to spec, TP while SE in
passive contexts can stay in X'''P because a nominal subject moves to spec, TP or agreement

208
 SE IN SPANISH

of T with the nominal argument makes that movement unnecessary. Either way, nominative
SE will end up in a TP position and passive SE could stay lower.
The problem for Kempchinsky's approach is stronger from the perspective of the uses of SE
that are part of the predicate head, where the ordering facts are identical to the argument facts.
This is because the analysis of SE in the argument cases requires treating SE as a nominal
element that moves to a high position and this movement is not expected if SE is a predicate.
For such cases one would have to propose a different view of the position of the clitic SE: if
SE is in v, one expects that a cyclic derivation would give an ordering lo + se, against the facts.
The cyclic derivation would need to locate SE is a high position, as an agreement head, a
position like (567) above. Secondly, one would have to propose that SE has a feature that no
other clitic in the system has –call it [uSE] to highlight that this element would be the only one
that has this feature, and that the feature is not really identified independently–. This [SE]
feature might be the result of the grammaticalisation of the reflexive pronoun from Latin, and
the elements that carry this feature cannot introduce referents independently: either they are
coreferential with some other element, as the standard reflexive anaphor, or they mark a head
where one of the arguments must be coreferential with another. In any case, the presence of
this [uSE] feature is necessary to license predicative heads that are defective in the sense just
presented –they introduce argument structures where one of the positions cannot be associated
to a definite referent–. One can assume that such heads have the [SE] feature and need to be
licensed at a later stage by the projection spelled out as SE:

(572) X'''P

X'''

X''' XP
SE
[uSE] lo X

X vP

v AspP
[SE]
lo Asp

Asp VP

lo ...

Note that once this analytical option is adopted, leaving SE outside the system of clitics of
Spanish, it could also be adopted to account for the position of SE when SE apparently performs
an argument function. For such cases one would only have to substitute the SE in the specifier
position within the verbal complex with an empty pronoun pro that has a feature [SE] that
needs to be licensed by insertion of SE, and claim that the spell out of SE happens in the high
head, not within the verbal complex.

209
 ANTONIO FÁBREGAS

(573) X'''P

X'''

X''' XP
[uSE, uphi]
lo X

X vP
[uD]
pro v
[SE, phi]
v AspP

lo Asp
[D]
Asp ...

In the Note that this ellaboration of Kempchinsky's proposal approaches the view of SE as
a probe, a mere introductor of unvalued features that can be used for agreement. Therefore, SE
would be an additional agreeing head, something that might help us explain and integrate in
the system the cases of spurious SE in a way similar to the one that we adopted in our discussion
of Walkow (2012) in §10.4.
Note however that the approach changes the view of what SE is as a morphosyntactic
element: it is not an aspectual element, but a probe located in a high position that is necessary
to license elements that otherwise would not be licensed by the regular functional heads, either
because the element carries a feature that otherwise cannot be licensed (call it [SE]) or because
the element requires licensing by phi features because no other head can provide them to it.
From this perspective, SE is not an impoverisher of the verbal structure, but actually a
strengthener that provides agreement features in contexts where there is an element that cannot
be licensed, either because it carries a feature that only SE can license or because it does not
have any other agreeing head.
This is, as we will see in the next and final section of the article, the approach that we
consider currently the best apt to unify the uses of SE, combining insights from Kempchinsky's
work with the proposal of SE as a probe which goes back to Kayne (2000) and Reuland (2011)
and Volkova & Reuland (2014), who treat reflexives as elements whose function is to introduce
unvalued features. But before we go there, let us briefly revise one more approach to the
unification of SE: one that assumes one single position for SE but does not treat it as a probe.

12.2. Dovbrovie-Sorin (2021)

Dobrovie-Sorin (2021) shares with Kayne (2000) the idea that SE is one single element, and
all instances of SE have to be located in the same position; she also shares with Kayne the idea
that SE is related to a nominative position, which follows from his approach where SE is the
highest element in the clitic area and the one that licenses best a second agreement position for
the subject (Kayne 2010). In contrast, unlike Reuland (2011), SE has an impoverishing function
and does not act as a probe.
Although we have already revised the relevant structures in §8 above, we repeat them here
for convenience.

210
 SE IN SPANISH

(574) TP Anticausative

T VoiceP
se
Voice{ø} VP

V DP

(575) TP Passive

T VoiceP
se-arb
Voice{ø} VP

V DP

The game is performed by putting SE always in TP. In this position, the distinction between
passive and anticausative depend on whether SE introduces an arbitrary specification or not. In
the absence of an external argument located in Voice, ARB in T forces the interpretation that
there is an agent that however has an indefinite interpretation. Although Dobrovie-Sorin (2021)
does not discuss the impersonal construal, we can speculate that she could derive it from the
structure of the active (576a) by specifying the T head as ARB, which then only licenses
external arguments that are indefinite (576b).

(576) a. TP

T VoiceP

DP Voice

Voice VP

V DP

b. TP

T VoiceP
se-ARB
pro Voice

Voice VP

V DP

The immediate advantage of this approach is that SE will always be external to any other
clitic with respect to the verb.
There is one technical and one more serious problem with this analysis. The technical
problem is that TP should be in Romance a second way to introduce an external argument,

211
 ANTONIO FÁBREGAS

which creates the problem of why there are no simple sentences with two external arguments,
one in Voice and one in TP. The more serious problem is that if SE spells out T and nothing
lower than it it becomes impossible to give account of the predicate-affecting uses of SE (§3-
§6) where the clitic might influence the lexical semantics of the verb, or even its argument
structure options.
Like in the case of Kempchinsky, there are ingredients of this analysis that can be kept,
particularly the hypothesis that SE spells out a head that is high in the tree, irrespective of its
function, and that it does not have a role impoverishing verbal heads, but rather presents a
possible solution for how to license arguments when the head is defective.

13. Conclusion: towards a possible unified explanation

It is time to wrap up this article. In this final section we will highlight which aspects of SE
are more established, which ones are still problematic and how the uses could be unified, that
is, under which premises the unification is possible.
When it comes to the variety of SE uses, three aspects can be globally highlighted given the
rich empirical picture that we obtain from §3-§10 above:

a) The role of SE cannot be simply claimed to be that of an impoverisher. The impoverishing

function of SE is built over some core cases –anticausative, passive, middle, impersonal– where
it seems to have the function to demote, remove or block the expression of an external
argument, or from cases where the case assigning possibilities are altered –antipassive, passive,
anticausative, middle–. In contrast, there are uses which are difficult to claim are impoverishing
cases: reflexives, aspectuals, agentives or factitives are the main cases, and in the case of
spurious SE the clitic can in fact be argued to have a strengthening function.
b) It is unlikely too, despite the historical evidence, that reflexive SE can be claimed to be
at the core of all uses, at least if taken in the sense that SE is a reflexive clitic in all these uses.
The reflexive analysis is very problematic in the case of anticausatives, for instance, and no SE
use beyond the reflexive (and under some different conditions, the spurious SE) allows
doubling with an overt pronoun.
c) There are no empirical reasons to propose that paradigmatic SE and non-paradigmatic SE
constitute two different types of SE elements. We have seen that the fixation of non-
paradigmatic SE in a third person can be treated as an epiphenomenon due to SE being
associated to a third person indefinite pronominal, an independent ban on personal pronouns
as subjects of the construction, or the necessity that there is another 3rd person element to
produce a clash with a 3rd person element.
d) The inherent SE cases can be treated as involving idiomatic constructions belonging to
the other established classes and do not constitute a separate class.

With respect to the nature of SE, here are the main conclusions:

a) The variety of constructions where SE appears is presumably an effect of its minimal

status as a syntactic object: it has the smallest set of features from the set of clitics in Spanish,
which makes it a small element that is ambiguous between a head and an XP and therefore can
find a place in a diversity of syntactic configurations.
b) It is unclear whether SE should or can be integrated in the system of clitics in Spanish,
that is, it is not clear that SE should have the same status as clitics like le, me or lo. Even if SE,
as a reflexive, can alternate with person-marked clitics, it can cooccur with them and the
presence of a reflexive amilirates PCC effects (§2.4), showing that SE should not automatically
be identified with such clitics and two positions should be distinguished for them.

212
 SE IN SPANISH

c) Integrated in the system of clitics or not, SE does not carry its own reference and seems
to behave as a set of nominal features without DP or NP layers.

With respect to the syntactic analysis of SE, here are the main conclusions:

a) The analyses that play around with the position of SE in order to explain its different uses
face one problem: SE as a clitic has to be located always in the same external position with
respect to the verb, irrespectively of the use. The problem can be solved for approaches where
SE is a pronoun if there is a high position that only licenses SE and no other head can license
SE, but there are (to the best of our knowledge) no analyses with this assumption, where SE is
always in an argument position.
b) The analyses that treat SE as a marker for a particular flavour of a verbal head also face
the same problem, even stronger, if they treat SE as materialised in the verbal head. In order to
derive the clitic ordering such analyses need to abandon cyclicity or assume that the ordering
of clitics is imposed outside syntax.
c) The analysis of SE as probe introducing agreement features seems inescapable in the case
of spurious SE, where there is evidence that SE is present as such in the syntax and semantics
and is not just the result of a morphological operation on spell out.

With this background in mind, we will now try to group the SE constructions in a reduced
number of configurations, drawing from the analyses presented above. Here we will sketch an
analysis that respects the premises above and present alternative analyses within the three
options mentioned:

i. SE is related to an argument
ii. SE is related to a verbal head
iii. SE is an agreeing element high in the structure

Then, to conclude, we will overview the loose ends of this possible unification.

13.1. Group A: SE as a head that attaches to a verbal head

There is a first group of SE structures that can be related to an operation in the head that
introduces the external argument, Voice or v, and whose role is to impoverish such head by not
letting it expand its full set of elements. Here, for convenience, we will use the label Voice for
that head, but nothing follows from this terminological choice, although we favour accounts
where Voice is related to a head that does not assign a theta role and the theta role comes from
other heads in the complement of Voice (see Pylkkänen 2002, Harley 2017, Ramchand 2018,
Fábregas & Putnam 2020). The facts related to impersonals and passives suggest that case and
the external argument have to be kept as separate properties introduced in different heads, so
we will also differentiate case from argument structure.
That said, the first group involves anticausative, passive, impersonal, middle and possibly
agentive, factitive and reflexive. The common property is that SE is related to an impoverished
head in Voice. The starting point is Ordóñez & Treviño's (2011, 2016) proposal to unify
impersonal and passive SE as (577) –of course changing the label for the relevant head–.

213
 ANTONIO FÁBREGAS

(577) VoiceP Se vive bien.

SE lives well
proarb Voice 'One lives well'

Voice vP
SE
v VP...

The value of the SE head as necessarily third person depends on the value of the specifier,
here a pro, and the distinction is assumed to follow from case assignment. Note that the
complement of Voice here is little v meaning CAUSE or DO, which allows the interpretation
that the external argument is an agent.
If passive SE does absorb accusative case, one could propose that the head related to SE is
not just Voice but starts from the head where accusative case is assigned, impoverishing it, and
from there it moves up to Voice (or vice versa, depending on where that head is in the analysis
assumed in each case).

(578) VoiceP

proarb Voice

Voice AccP
SE
Acc vP
SE
v VP

VP ...

The anticausative and the middle version could correspond to the same head, but where v
cause is not present and an agent theta role cannot be assigned to the specifier of Voice; in this
configuration, the patient moves to spec, VoiceP to give features to the SE head.

(579) VoiceP La televisión se rompió.

the TV SE broke
DP Voice

Voice VP
SE
V DP

One could further assume that the difference between anticausative and middle depends on
the absence / presence of a modal head and the interpretation of the verbal stem as implying an
agent or not (alternatively, one could derive the middle from the passive assuming presence of
a vP layer and letting the patient move to the position of spec, VoiceP, which is orthogonal to
the unification that we attempt here).
It is in principle possible to adopt Labelle (2008) and treat the reflexive / reciprocal as
another type of impoverished Voice head that allows for letting an internal argument positon
unsaturated, provided it is bound by the external argument.

214
 SE IN SPANISH

(580) VoiceP

DP Voice

Voice vP
SE
v VP

The distinction between naturally reflexive verbs that do not allow doubling and the
reflexive verbs that allow it, from this perspective, could be operationalised with the tools of
Labelle (2008), proposing that the presence of that double is neutral for argument structure and
adding that its addition introduces the expectation that the event would not be performed by
the subject on itself, letting the distinction be lexically defined.
It is less clear whether agentive and factitive verbs would also allow a reduction to a
structure like (577), but here we will try to do so. The agentive / factitive reading, which in
§6.2 were argued to be essentially the same type of reading as the factitive comes from the
conceptual semantic possibilities of the predicate, share with the impersonal SE the
interpretation that the subject must be human. One might speculate, then, that the agentive /
factitive SE is a subcase of (577) where there is a vP layer that assigns CAUSE and where the
pro is not an indefinite pronoun, but an overt nominal.

(581) VoiceP

DP Voice
(human)
Voice vP
SE
v VP

V ...

The same SE reading that one obtains in anticausatives, where the sole responsible of the
event is the undergoer, would emerge with agentive SE, with the only difference that in such
cases the subject must be human and get a theta role assigned which forces the reading that the
agent was the only participant that triggers the event; it is possible that the prototypical reading
of this restricted value with human subjects that are interpreted as causers is precisely the
volitional agent reading.
Let us now consider the three versions of this analysis. In the unified approach to SE as a
pronoun, SE in such cases would be a pronominal head attached to the relevant verbal head,
where it would have the role to absorb part of its specification and impoverish it. It will contain
unvalued phi features for person and number, letting it agree with the element in spec, VoiceP.

215
 ANTONIO FÁBREGAS

(582) VoiceP

Voice

Voice ...

SE0 Voice0

From there, given the ambiguity between X and XP, it would move as an XP to the specifier
of the high clitic position, accounting for the order. In the case of the impersonal, a second
forcefull step would be to move to T or TP. An obvious alternative within this analysis is to
propose that SE is an XP argument introduced in the Voice position, but this would have to be
complemented with a theory that explains why anticausative, middle and reflexive SE have
subjects that act as being in specifier positions while passive SE suggests that the subject can
stay in a complement position; thus, it would not be straightforward to locate SE in all these
cases in the spec, VoiceP position and the analysis would have to locate it as originating from
a lower vP internal position for most cases.
The version of the analysis where SE marks the head, as we saw, would be possible but has
the problem that the clitic ordering would have to be non cyclic. We see no way to propose that
SE is the spell out of a predicate head that derives the order respecting cyclicity; either cyclicity
is broken or one admits that the ordering of clitics has nothing to do with syntax. Note that we
still need to assume unvalued person and number features.

(583) VoiceP

Voice

Voice ...
SE

The third approach, SE as a high probe, would involve marking the relevant heads that get
impoverished with a feature [SE] that needs licensing, and marks the reduction in their
capabilities. SE licenses that value and is spelled out high; in the case of the impersonal SE,
the licensing forces SE to further move to the T head. The unvalued features that agree with
the spec, VoiceP are placed necessarily in the high probe, although this does not force the head
to have valued features: these could also be non interpretable, provided that the [SE] feature is
interpretable.

(584) seP

se ...VoiceP
[uSE, uPhi]
Voice

Voice ...
[SE, phi]

That [SE] feature could be coming from the grammaticalisation of reflexive SE as a valency
reductor, following the path of grammaticalisation that eventually turns it into a feature of some
verbal heads that share the property of not licensing full argument structures or restricting the

216
 SE IN SPANISH

interpretation of the arguments. Note that the specific operation depends on what one assumes
about features: one could also assume Pesetsky & Torrego's (2007) distinction between
(un)interpretable and (un)valued features and propose that SE in seP is interpretable but
unvalued, while in the head VoiceP the head would be uninterpretable [SE] but valued through
agreement with the external argument. Of course, what we have sketched here is only a
possibility that allows many elaborations depending on additional theoretical tenets.

13.2. Group B: SE as an argument that incorporates or not

The second group of constructions share the property that the role of SE seems to influence
the argument structure of the verb by apparently introducing new participants, or altering the
properties of existing participants without affecting the presence of a full-fledged external
argument and therefore not involving Voice. Here we can group the aspectual SE cases both
with transitive and intransitve predicates, the antipassive SE and the figure SE.
In such cases SE is related to an argument position; starting with the antipassive SE and the
transtive SE, the constituent related to SE is introduced in the head that provides a position for
the internal argument, and might incorporate to it.

(585) vP

v VP

XP V

V SE

The difference between the transitive and the intransitive use might follow from
incorporation: in the antipassive SE, the element does not incorporate to the verbal head and
therefore requires case (586a); in the transitive SE use the element incorporates and case is not
necessary (586b).

(586) a. vP aprovechar-se *(de) algo

profit-SE (of) something
v VP

PP V

V SE

b. vP comer-se la habitación
eat-SE the room
v VP

DP V

SE+V SE

The same configuration works for intransitive aspectual cases, where the effect is to force
that the internal argument of the unaccusative projects as an external argument.

217
 ANTONIO FÁBREGAS

(587) a. vP salir
exit
v VP

V DP

b. vP salir-se
exit-SE
DP v

v VP

V SE

The analysis of transitive aspectual SE could behave as MacDonald's (2017) configuration

and provide a similar configuration, where SE is an affected argument.

(588) VoiceP comer-se dos bocadillos

eat-SE two sandwiches
DP Voice

Voice vP

ve VP

DP V

V PP

P SE

Figure SE would involve an argument introduced as the specifier of a PP.

(589) VP meter-se en la cama

put-SE in the bed
V PP

SE P

P DP

The analysis of these cases in the proposal where SE is an argument is straightforward for
most cases, assuming movement to the high XP position to obtain clitic ordering. The analysis
however would have trouble explaining the difference in case assignment between the
antipassive and the transitive through incorporation, as in the second case SE would have to
excorporate to move to XP. Note that here the features of the argument cannot be non-
interpretable in all cases, as in some of the configurations that predicts that SE would co-vary

218
 SE IN SPANISH

with the internal argument, not the subject, when they are distinct. One could, however, force
that the argument is coindexed with the subject, with or without additional operations.

(590) VoiceP

DPi Voice

Voice vP

ve VP

DP V

V PP

P SEi

The analysis that treats SE as the projection of a verbal head is less straightforward, because
the role that the verbal head would have in these cases would not involve impoverishment in
the same sense as before, except for the antipassive case when it comes to case assignment.
However, assuming that SE could mark P and V heads without impoverishing them, but rather
linking them as contaning an argument coindexed with an external argument, the approach can
be kept, with the known problem of the clitic ordering.
The analysis that treats SE as materialising a high probe in the clause would need to propose
that the arguments introduced are sets of phi features, thus defective arguments, that have to be
coindexed with the external argument and agree in features with it –possibly through the feature
[SE] that forces this interpretation. Once coindexed with the relevant argument, SE would
license the features in the usual way. The coindexation might be an effect of the absence of
referential properties of SE, caused by [SE].

(591) seP

se ...vP
[uSE, uphi]
v VP

V [SE, phi]

13.3. Group C: SE as a probe

The final group has only one member, the stand-alone case of spurious SE. In this
configuration, SE marks or solves an agreement problem. The SE-as-argument analysis is
forced to the conclusion that SE is in this case the spell out effect of a defective situation in
morphology, and is not present syntactically, with the known problems pointed out in. Alcaraz
(2017).
The SE as involving a verbal head and SE as a high probe are more difficult to compare in
this respect. The reason is that the two approaches are crucially differentiated empirically by
clitic ordering, and the spurious SE cases, by definition, do not accept any other clitic in the
sequence except the accusative third person. Personal pronouns introduced as direct objects
(592) do not trigger the reduction and are difficult to evaluate given PCC effects, and personal

219
 ANTONIO FÁBREGAS

pronouns introduced as non-selected objects produce ungrammatical results in this

configuration (593, Cuervo 2013).

(592) a. Se lo quité.
SE it took.away
b. *Te le quité.
you it took.away
(593) a. ¿En serio le robaste eso?
really him.dat stole that
b. ¿En serio se lo robaste?
really SE it stole
c. ¿En serio te le robaste eso?
really you him.dat stole that?
'You really stole that for yourself?'
d. *¿En serio se te lo robaste?
really SE you it stole

Thus, the approach where SE spells out a low impoverished verbal head could in principle
spell out a head that usually assigns dative case and marks that the assignment in this case is
more defective; if SE has an impoverishing function we should expect that dative case, or other
properties of the argument, should be impoverished, and this can match the data noted in
Alcaraz (2017, 2021) where spurious SE is not referential, in contrast to datives.
The approach where SE is a high probe is unproblematic for this type of SE, as it performs
in a pure way the function that the approach would claim for all SE cases: to provide a probe
that licenses an argument that otherwise could not be licensed.

13.4. The loose ends

Thus, to conclude, the approach above seems to suggest that the most straightforward way
of accounting for all SE structures is the SE as probe approach (Reuland 2011) that treats SE
outside the clitic system and singles it out by two properties:

a) Containing a feature that no other head contains, [SE], which might be the
grammaticalisation of the reflexive original function, that forces the element that it licenses to
not introduce a referent of its own or to impoverish one of the argument positions that it
otherwise introduces.
b) Being a set of nominal phi features that does not include a D feature, even when associated
to elements in argument position.

The analysis just sketched above, as it is obvious, leaves a number of problems unaddressed.
The most significant one of them refers to the lexical variation in the use of SE as applied to
anticausative, agentive and aspectual SE cases.
As discussed in §4.1 above not all verbs with an anticausative pair mark the anticausative
with SE –in some verbs SE is impossible, and in others, it is possible but not compulsory–. If
the analysis of inherent SE cases is right, it is not even true that a verb must have a causative
pair to mark SE. Moreover, which verbs use SE are subject to some individual variation among
speakers. The problem, from the perspective of the analysis just sketched, could be stated as
follows:

(594) Some verbs can have defective Voice that is licensed without SE and some verbs require
licensing with SE for defective Voice.

220
 SE IN SPANISH

Several options come to mind as potential ways to explore from this perspective:

a) Theories where the syntax of anticausative verbs with SE differs from the syntax of those
that don't. Remember that in §4.2 we overviewed some evidence that anticausative without SE
differ from anticausatives with SE through dative causers.

a. One could posit that the verbs that do not carry SE contain some type of vP layer that
is able to license the defective [SE] feature, proposing then that the syntax of the anticausative
verbs without SE is different from those with SE;
b. One could alternatively posit that the defective Voice that these verbs carry is not
marked by SE, making the presence of SE unnecessary
c. One could finally argue that anticausatives without SE lack VoiceP altogether and lack
an external argument because of this, or even the lack of vP, thus making SE unnecessary
because there is no Voice to mark with [SE].

b) Theories where the syntax of anticausative verbs with SE and those without SE is largely
identical. Remember the lexical variation with respect to which verbs take anticausative SE,
suggesting that perhaps speakers do not play around with the syntactic or semantic
construction. In this approach, SE would be present in the syntax but would either be spelled
out as zero with some verbs or spelled out together with the verbal stem exponent, through
synthetic exponence.

Remember that a similar approach is taken in Vivanco (2021), where she argues that SE in
anticausatives marks a particular type of aspect that involves a bounded incremental change,
in alternation with an exponent zero that marks the other cases. There is solid evidence that in
at least some groups of anticausative SE cases this generalisation is correct, but our approach
cannot straightforwardly account for this semantic effect. In general, in fact, the analysis
sketched above does not capture the semantic effects of SE in aspect, which seem to be quite
systematic (Nishida 1994, García Fernández 2015). If SE is a high head in the clausal structure
and its role is to provide agreement features, it should not have any direct effect on or sensitivity
to the Aktionsart properties of the predicate. Similarly, it is unlikely that the aspectual
properties of the verb –located in whichever head one decides– should obtain a compulsory
bounded incremental reading with addition of the feature [SE], that in such cases would not
have an obvious role in impoverishing the referentiality of a participant or removing
participants –remember that aspectual SE does not remove an internal argument–. For this
problem, there are several potential ways out that might be explored in the future:

a. One could explore the possibility that the accomplishment interpretation is the default
interpretation of aspect for transitive verbs, so that [SE] could be argued to impoverish a V or
v layer.
b. One could claim that the telic incremental interpretation is obtained with SE because SE
impoverishes the dynamic part of the event, so that event has to define its progress forcefully
through the referential properties of the internal argument. Thus, the internal argument has to
be coextensive with the verbal event and therefore measures it; in other cases, the scale
provided by an adjective would act as such referentially impoverished argument that SE forces.

One could claim that the approach sketched largely ignores semantic issues, or treats them
through a very vague principle whose application in individual cases is unclear. This is the
reason why the agentive or factitive SE cases are difficult to treat in this approach, because in

221
 ANTONIO FÁBREGAS

them one needs to treat the low referentiality of [SE] as triggering a prototypical human agent
or instigator reading, which seems to be an effect that is at a different level from the effect that
it produces in impersonals or passives, where the agent cannot be expressed by a distinct
nominal.
But perhaps the most serious problem of this approach and any other approach that builds
the reasoning starting from the minimal role of SE in structures, as a minimally specified entity
within the system, is to restrict the SE uses in a way that one does not overgenerate
constructions that are not attested.
For instance, we have a figure SE where the subject and the entity that moves coincide
(595a), but why cannot we have a ground SE where the subject coincides with the region where
another entity moves (595b), as we saw in §6.4 above? As we suggested there, one can claim
that the ground position is for locations, and the subject is not a locative, but note that nothing
prevents an overt structure where a human is that region (595c). What makes a pronoun related
to SE unable to appear in the place of the nominal in (595c)?

(595) a. Juan se infiltró en la asociación.

Juan SE infiltrated in the association
b. #Juan se puso a Pedro.
Juan SE put DOM Pedro
Intended: 'Juan put Pedro in a location defined by him'
c. Juan puso a Pedro ante él.
Juan put DOM Pedro before him
'Juan put Pedro in a place before Juan'

Despite its length, we believe that this article just scratches the surface of SE structures in
Spanish. We have been unable to provide answers to most of the questions that this element
poses in Spanish, but at least we hope to have been able to show what the main analytical
challenges and problems are, and to have managed to put some order in the apparent chaos of
constructions, interpretations and analytical options that the issue of SE produces in
contemporary Spanish.

Antonio Fábregas
Department of Language and Culture
Faculty of Humanities, Social Sciences and Education
University of Tromsø
Tromsø, N-9037
Norway
[email protected]

References
Ackema, P. & M. Schoorlemmer. 1995. Middles and nonmovement. Linguistic Inquiry 26 (2):
173-197.
Alcaraz, A. 2017. Spurious vs. dative. Ms., available at lingbuzz/003293.
Alcaraz, A. 2021. Agreement, binding and the structure of Spanish clitics. PhD dissertation,
University of the Basque Country.
Alexiadou, A. & E. Anagnostopoulou. 2004. Voice morphology in the causative-inchoative
alternation: Evidence for a non-unified structural analysis of unaccusatives. In A.
Alexiadou, E. Anagnostopoulou & M. Everaert (eds.), The unaccusativity puzzle:
Explorations of the syntax-lexicon interface. Oxford: Oxford University Press, 114-136.
https://doi.org/10.1093/acprof:oso/9780199257652.003.0005

222
 SE IN SPANISH

Alexiadou, A., E. Anagnostopoulou & F. Schäfer. 2015. External arguments in transitivity

alternations. A layering approach. Oxford: Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199571949.001.0001
Alexiadou, A., E. Anagnostopoulou, & F. Schäfer. 2006. The properties of anticausatives
crosslinguistically. In M. Frascarelli (ed.), Phases of interpretation. Berlín: Mouton,
187-211. https://doi.org/10.1515/9783110197723.4.187
Anagnostopoulou, E. (2003). The syntax of ditransitives: Evidence from clitics. Berlin: Walter
de Gruyter.
Anderson, S. R. 2005. Aspects of the Theory of Clitics. Oxford: Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199279906.001.0001
Aoun, J. 1981. The formal nature of anaphoric relations. Doctoral dissertation, Massachusetts
Institute of Technology.
Armstrong, G. & J. MacDonald. 2021. A guide to understanding SE constructions: where they
come from and how they are connected. In G. Armstrong & J. E. MacDonald (eds.),
Unraveling the complexities of SE. Dordrecht: Springer, pp. 1-33.
https://doi.org/10.1007/978-3-030-57004-0_1
Armstrong, G. & P. Kempchinsky. 2021. Causative SE: a transitive analysis. In G. Armstrong
& J. E. MacDonald (eds.), Unraveling the complexities of SE. Dordrecht: Springer, pp.
219-243. https://doi.org/10.1007/978-3-030-57004-0_9
Armstrong, G. 2013. Agentive reflexive clitics and transitive se constructions in Spanish.
Borealis 2 (2): 81–128. https://doi.org/10.7557/1.2.2.2526
Balari, S. & N. Bel. 1990. Se impersonal: por qué no es un clítico de sujeto. In M. A. Álvarez
Martínez (ed.), Actas del congreso de la Sociedad Española de Lingülstica. XX
aniversario, Madrid, Gredos, vol. 1, pp. 318-328.
Barbosa, P. 2000. Clitics: a Window into the Null Subject Property. In J. Costa (ed.),
Portuguese Syntax: Comparative Studies, pp. 31-93. New York: Oxford Press.
Basilico, D. 2010. The se clitic and its relation to paths. Probus 22, pp. 271-302.
https://doi.org/10.1515/prbs.2010.010
Beavers, J. & A. Koontz-Garboden. 2013. In defense of the reflexivization analysis of
anticausativization. Lingua 131, pp. 199–216.
https://doi.org/10.1016/j.lingua.2012.10.009
Béjar, S. & M. Rezac. 2003. Person licensing and the derivation of PCC effects. Amsterdam
studies in the theory and history of linguistic Science series 4, pp. 49-62.
https://doi.org/10.1075/cilt.244.07bej
Belletti, A. & L. Rizzi. 1988. Psych verbs and theta theory. Natural Language and Linguistic
Theory 6, pp. 291–352. https://doi.org/10.1007/BF00133902
Belletti, A. 1982. Morphological passive and pro-drop: The impersonal construction in Italian.
Journal of Linguistic Research 2 (4), pp. 1–34. https://doi.org/10.1515/tlir.1982.2.2.101
Bello, A. 1847. Gramática de la lengua castellana destinada al uso de los americanos.
Alicante: Biblioteca Virtual Miguel de Cervantes (2002).
Benveniste, É. 1960. Problems in general linguistics. Ed. Florida: University of Miami Press.
Bleam, T. 1999. Leísta Spanish and the Syntax of Clitic doubling. Doctoral Dissertation,
University of Delaware
Bogard, S. 1996-1997. El sentido anticausativo y su formalización sintáctica en el español.
RPL 3, pp. 271-289.
Bogard, S. 1999. Duplicación y clausura argumental: dos funciones del clítico reflexivo en
español. Español actual 71, pp. 41-48.
Bonet, E. 1991. Morphology after syntax: Pronominal clitics in Romance. PhD diss., MIT.
Bonet, E. 1995. Feature structure of romance clitics. Natural Language and Linguistic Theory
13, pp. 607–647. https://doi.org/10.1007/BF00992853

223
 ANTONIO FÁBREGAS

Borer, H. 1984. Parametric syntax. Dordrecht: Foris. https://doi.org/10.1515/9783110808506

Bosque, I. 1992. Anáforas distributivas: la gramática de sendos. Miscellanea Antverpiensia.
Tübingen: Max Niemeyer, pp. 59-92.
Bosque, I. 1999. Repaso de sintaxis tradicional. Ejercicios de autocomprobación. Madrid:
Arco / Libros.
Brucart, J. M. 2012. Copular alternations in Spanish and Catalan attributive sentences.
Linguística: Revista de Estudos Linguísticos da Universidade do Porto 7, pp. 9.43.
Bruhn de Garavito, J. 1999. The 'se' constructions in Spanish and near-native competence.
Spanish Applied Linguistics 3, pp. 247-96.
Bull, William E. 1952. The intransitive reflexive: ‘ir’ vs. ‘irse’. The Modern Language Journal
36(8), pp. 382–386. https://doi.org/10.1111/j.1540-4781.1952.tb06587.x
Büring, D. (2005). Binding theory. Cambridge University Press.
https://doi.org/10.1017/CBO9780511802669
Campos, H. 1989. Impersonal passive SE in Spanish. Linguisticae Investigationes 13, pp. 1-
21. https://doi.org/10.1075/li.13.1.02cam
Cano Aguilar, R. 1981. Estructuras sintácticas transitivas en el español actual. Madrid:
Gredos.
Cardinaletti, A. 2000. On different types of clitic clusters. Workshop on the Antisymmetry
Theory, Cortona, May 15-17.
Cardinalletti, A. & M. Starke. 1999. The typology of structural deficiency: A case study of the
three classes of pronouns. Clitics in the languages of Europe, 8.
https://doi.org/10.1515/9783110804010.145
Carlson, G. 1977. Reference to kinds in English. PhD dissertation, University of Massachusetts.
Cartagena, N. 1972. Sentido y estructura de las construcciones pronominales en español.
Concepción: Publicaciones del Instituto Central de Lenguas.
Cennamo, M. 1999. Late Latin pleonastic reflexives and the unaccusative hypothesis.
Transactions of the Philological Society 97 (1), pp. 103–150.
https://doi.org/10.1111/1467-968X.00046
Cennamo, M., T. Eythórsson & J. Bardal. 2015. Semantic and (morpho)-syntactic constraints
on anticausativization: Evidence from Latin and Old Norse-Icelandic. Linguistics 53
(4): 677–730. https://doi.org/10.1515/ling-2015-0015
Charnavel, I. 2019. Locality and logophoricity: A theory of exempt anaphora. Oxford: Oxford
University Press. https://doi.org/10.1093/oso/9780190902100.001.0001
Chierchia, G. [1989] 2004. A semantics for unaccusatives and its syntactic consequences. In
Alexiadou, A., Anagnostopoulou, E. & Everaert, M. (eds.), The unaccussativity puzzle:
Explorations of the syntax-lexicon interface. Oxford: Oxford University Press, pp. 22-
59. https://doi.org/10.1093/acprof:oso/9780199257652.003.0002
Chung, S. & W. Ladusaw. 2003. Restriction and saturation. Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/5927.001.0001
Cinque, G. 1988. On si construction and the theory of arb. Linguistic Inquiry 19 (4), pp. 521-
581.
Clements, J. C. 2006. Transitivity and Spanish non-anaphoric se. In J. Clancy Clements &
Jiyoung Yoon (eds.), Functional approaches to Spanish syntax, 236–264. London:
Palgrave Macmillan. https://doi.org/10.1057/9780230522688_11
Condoravdi, C. 1988. The middle: where semantics and morphology meet. Student conference
in linguistics. MIT working papers in linguistics 11, pp. 16-30.
Contreras, H. 1973. Grammaticality versus acceptability. The Spanish se case. Linguistic
Inquiry 4, pp. 83-88.
Contreras, H. 1974. Indeterminate subject sentence in Spanish. Bloomington: Indiana
University Linguistic Club.

224
 SE IN SPANISH

Contreras, L. 1966. Significados y funciones de 'se'. Zeitschrift for Romanische Philologie 82,
pp. 298-307. https://doi.org/10.1515/zrph.1966.82.3-4.298
Creissels, D. 2006. Syntaxe générale. Une introduction typologique. Paris: Hermès.
Cuervo, M. C. 2003. Datives at large. Doctoral dissertation, Massachusetts Institute of
Technology.
Cuervo, M. C. 2013. Spanish clitic clusters: Three of a perfect pair. Borealis. An International
Journal of Hispanic Linguistics 2 (2), pp. 191–220. https://doi.org/10.7557/1.2.2.2692
Cuervo, M. C. 2014. Alternating unaccusatives and the distribution of roots. Lingua 141, pp.
48-70. https://doi.org/10.1016/j.lingua.2013.12.001
Cuervo, R. J. 1874. Notas a la Gramdtica de la lengua castellana de don Andrés Bello. Bogotá:
Instituto Caro y Cuervo.
D'Alessandro, R. 2007. Impersonal Si constructions. Berlin: Mouton de Gruyter.
https://doi.org/10.1515/9783110207514
Darlymple, M., M. Kanazawa, S. Mchombo & S. Peters. 1994. What do reciprocals mean?
SALT IV, pp. 61-78. https://doi.org/10.3765/salt.v4i0.2466
De Benito Moreno, C. 2015. Las construcciones con 'se' desde una perspectiva variacionista
y dialectal. PhD Thesis, Universidad Autónoma de Madrid.
De Benito Moreno, C. 2021. Is there really an aspectual SE in Spanish? Folia Linguistica 55,
pp. 195-230. https://doi.org/10.1515/flin-2020-2074
De Cuyper, G. 2006. La estructura lexica de la resultatividad y su expresión en las lenguas
germánicas y románicas. München: Lincom.
De Kock, J. & E. Gómez Molina. 1990, Las formas pronominales del verbo y la pasiva. In De
Kock (ed.), Gramática española, vol. 11. Salamanca: Universidad de Salamanca.
De Mello, G. 1978. On the use of por + agent with se constructions. Hispania 61, pp. 323-327.
https://doi.org/10.2307/340879
De Mello, G. 1996. Morir frente a morirse en el español hablado contemporáneo. Zeitschrift
für Romanische Philologie 112 (2), pp. 277-292.
https://doi.org/10.1515/zrph.1996.112.2.277
De Miguel, E. & M. Fernández Lagunilla. 2000. El operador aspectual se. Revista Española de
Lingüística 30 (1), pp. 13-43.
De Miguel, E. 1992. El aspecto en la sintaxis del español Madrid: UAM.
Déchaine, A.M. & M. Wiltschko. 2020. A formal typology of reflexives. Studia Linguistica
71, pp. 60-106. https://doi.org/10.1111/stul.12072
Deguchi, A. 1978. Antipassive and reflexive passive in Spanish. Lingüística Hispánica
(Osaka) 1, pp. 87-101
Deguchi, A. 1982. On reflexive structures. Lingüística Hispánica 5, pp. 33-45.
Devís Márquez, P. 1993. Esquemas sintáctico-semánticos: el problema de la diátesis en
español. Cádiz: Universidad de Cádiz.
Di Tullio, Ángela. 2012. La construcción de la lectura agentiva del se no argumental. In
Valeriano Bellosta von Colber & Marco García García (eds.), Aspectualidad –
Transitividad – Referencialidad. Las lenguas románicas en contraste, 69–85.
Frankfurt: Peter Lang.
Diesing, M. 1992. Indefinites. Cambridge (Mass.): MIT Press.
Dimitriadis, A. 2004. Discontinuous reciprocals. Ms., University of Utrecht. Available from
http://www.let.uu.nl/~Alexis.Dimitriadis/personal/papers/
Dimitriadis, A. 2007. Irreducible symmetry in reciprocal constructions. In König, Ekkehard &
Gast, Volker (eds.), Reciprocals and reflexives: crosslinguistic and theoretical
explorations. Berlin: Mouton de Gruyter.

225
 ANTONIO FÁBREGAS

Dobrovie-Sorin, C. 1998. Impersonal se construction in Romance and the passivization of

unergatives. Linguistic Inquiry 29, pp. 399-437.
https://doi.org/10.1162/002438998553806
Dobrovie-Sorin, C. 2006. The se anaphor and its role in argument realization. In The Blackwell
companion to syntax, ed. M. Everaert and H. van Riemsdijk, vol. 4, 118–179. Oxford:
Blackwell. https://doi.org/10.1002/9780470996591.ch56
Dobrovie-Sorin, C. 2021. Implicit agents and the person constraint on SE-passives. In G.
Armstrong & J. E. MacDonald (eds.), Unraveling the complexities of SE. Dordrecht:
Springer, pp. 111-137. https://doi.org/10.1007/978-3-030-57004-0_5
Doron, E. 2003. Agency and voice: The semantics of semitic templates. Natural Language
Semantics 11, pp. 1-67. https://doi.org/10.1023/A:1023021423453
Dougherty, R. C. 1974. The syntax and semantics of each other constructions. Foundations of
Language, 1 2 : 1-47.
Dowty, D. 1979. Word meaning and Montague grammar. The semantics of verbs and time in
generative semantics and Montague’s PTQ. Dordrecht: Reidel.
https://doi.org/10.1007/978-94-009-9473-7
Embick, D. 2004. Unaccusative Syntax and Verbal Alternations. In A. Alexiadou, E.
Anagnostopoulou, & M. Everaert (eds.), The unaccusativity puzzle: explorations of the
syntax–lexicon interface. Oxford: Oxford University Press, pp. 137–58.
https://doi.org/10.1093/acprof:oso/9780199257652.003.0006
Escobar, L. & Teomiro, I. 2016. The gradual acquisition of clitic ‘se’ in Spanish L2. Topics in
Linguistics 17, pp. 17–29. https://doi.org/10.1515/topling-2016-0002
Everaert, M. 1999. Types of anaphoric expressions: reflexives and reciprocals. In Frajzyngier,
Zygmunt & Curl, Traci S. (eds.), Reciprocals: Forms and functions. Amsterdam:
Benjamins, pp. 63-83. https://doi.org/10.1075/tsl.41.04eve
Fábregas, A. & M. Putnam. 2020. Passives and Middles in Mainland Scandinavian. Berlin: De
Gruyter. https://doi.org/10.1515/9783110670912
Fábregas, A. & T. Cabré. 2020. Ways of being a dative across Romance varieties. In A. Pineda
& J. Mateu (eds.), Dative constructions in Romance and Beyond. Language Science
Press, pp. 395-411.
Fábregas, A. & T. Cabré. 2021. Towards a syntactic account of ungrammatical clitic sequences
and their repairs. In M. O. Hinzeling, N. Pomino & E. M. Remberger (eds.), Formal
approaches to Romance Morphosyntax. Berlin: De Gruyter, pp. 91-117.
https://doi.org/10.1515/9783110719154-004
Fábregas, A. 2013. Differential object marking in Spanish: State of the art. Borealis–An
International Journal of Hispanic Linguistics 2 (2), pp. 1–80.
https://doi.org/10.7557/1.2.2.2603
Fábregas, A. 2018. Determiners and quantifiers in Spanish. Borealis 7, pp. 1-108.
https://doi.org/10.7557/1.7.2.4546
Fagan, S. M. 1988. The English Middle. Linguistic Inquiry 19 (2), pp. 181-203.
Fagan, S. M. 1992. The syntax and semantics of middle constructions: a study with special
reference to German. Cambridge: Cambridge University Press.
Feliu, E. 2003. Morfología derivativa y semántica léxica. Madrid: Universidad Autónoma de
Madrid.
Fellbaum, C. 1986. On the middle construction in English. Indiana University Linguistic Club.
Fernández Ordóñez, I. 1999. Leísmo, laísmo y loísmo. In I. Bosque & V. Demonte (dirs.),
Gramática descriptiva de la lengua española. Madrid: Espasa Calpe, pp. 1317-1398.
Fernández Ramírez, S. [1957] 1986. Gramática española, vol 4. El verbo y la oración. Madrid:
Arco Libros.

226
 SE IN SPANISH

Fernández Ramírez, S. 1964. Un proceso lingüístico en marcha. Presente y futuro de la lengua

española, Madrid, III, pp. 277-285.
Fernández Soriano, O. 1989. Rección y Ligamiento en español: Aspectos del Parámetro del
Sujeto Nulo. Doctoral dissertation, PhD Thesis. Universidad Autónoma de Madrid.
Fiengo, R. & H. Lasnik. 1973. The logical structure of reciprocal sentences in English.
Foundations of Language, 9(4):447-468.
Fiengo, R. & R. May. 1994. Indices and identity. Cambridge (Mass.): MIT Press.
Fillmore, C. 1970. The grammar of hitting and breaking. In R. A: Jacobs & P. S. Rosenbaum,
(eds.), Readings in english transformational grammar. Waltham, MA: Ginn, pp. 120-
133.
Folli, R. 2001. Constructing telicity in English and Italian. PhD dissertation, Oxford
University.
Fontana, J. M. 1993. Phrase structure and the syntax of clitics in the history of Spanish.
Doctoral dissertation, University of Pennsylvania, Philadelphia.
Franco, J. 1993. Conditions on clitic doubling: the agreement hypothesis. Anuario del
Seminario de Filología Vasca "Julio de Urquijo" 27(1), pp. 285-298.
Franco, J. 2000. Agreement as a Continuum. Clitic phenomena in European languages 30, 147.
https://doi.org/10.1075/la.30.07fra
García Fernández, L. & D. Gómez Vázquez. 2015. More than a Copula: Complex Predicates
with estar and the Clitic se. In New Perspectives on the Study of Ser & Estar, ed. by
Isabel Pérez Jiménez, Manuel Leonetti and Silvia Gumiel, pp. 23–50. Amsterdam, John
Benjamins. https://doi.org/10.1075/ihll.5.02gar
García Fernández, L. 2011. Algunas observaciones sobre se aspectual”. In J. Cuartero Otal, L.
García Fernández & C. Sinner (eds.), Estudios sobre perífrasis y aspecto. München:
Peniope, pp. 43-71.
García Fernández, Luis. 2015. Some reflections on verbs with clitic increase: Verbs of motion.
In Elisa Barrajón López, José Luis Cifuentes Honrubia & Susana Rodríguez Rosique
(eds.), Verb classes and aspects, 264–287. Amsterdam: John Benjamins.
https://doi.org/10.1075/ivitra.9.12gar
García Negroni, M. M. 2002. La construcción media con 'se'. In C. Sánchez López (ed.), Las
construcciones con 'se'. Madrid: Visor, pp. 276-311.
García Pardo, A. 2021. Light verbs and the syntactic configurations of SE. In G. Armstrong &
J. E. MacDonald (eds.), Unraveling the complexities of SE. Dordrecht: Springer, pp.
243-265. https://doi.org/10.1007/978-3-030-57004-0_10
García Tejada, A., A. Cuza & E. G. Lustres Alonso. 2021. The production and comprehension
of Spanish se use in L2 and Heritage Spanish. Second Language Research. Online first,
pp. 1-31. https://doi.org/10.1177/02676583211016537
Gehrke, B. 2012. Passive states. In V. Demente & L. McNally (eds), Telicity, Change, and
State: A Cross-Categorial View of Event Structure. Oxford: Oxford University Press,
pp. 185–211. https://doi.org/10.1093/acprof:oso/9780199693498.003.0008
Gianollo, C. 2014. Labile verbs in Late Latin. Linguistics 52 (4), pp. 945–1002.
https://doi.org/10.1515/ling-2014-0013
Gili Gaya, S. 1943. Curso superior de sintaxis española. Madrid: Bibliograph.
Gómez Soler, I. 2015. Acquisitional patterns of Spanish anticausative se: The end of the road.
Revista Española de Lingüística Aplicada 28, pp. 349–81.
https://doi.org/10.1075/resla.28.2.01gom
Gómez Torrego, L. 1992. Valores gramaticales de se, Madrid, Arco Libros.
Gómez Vázquez, D. & L. García Fernández. 2013. Estados controlados. In Actas del X
Congreso de Lingüística General, ed. by J. F. Val Álvaro et al., 335–346. Universidad
de Zaragoza.

227
 ANTONIO FÁBREGAS

Grimshaw, J. 1982. On the Lexical Representation of Romance Reflexivie Clitics. In Joan

Bresnan (ed.), The Mental Representation of Grammatical Relations. Cambridge
(Mass.): MIT Press, pp. 87-148.
Grimshaw, J. B. 1997. Projection, heads, and optimality. Linguistic Inquiry 28(3), pp. 373–
422.
Hale, K. & S. J. Keyser. 1986. Some transitivity alternations in English. Lexicon Project
Working Papers 7. Cambridge (Mass.): Center for Cognitive Science, MIT.
Hale, K. & S. J. Keyser. 1987. A view from the middle. Lexicon Project Working Papers 10.
Cambridge (Mass.): Center for Cognitive Science, MIT.
Harley, H. 2017. The bundling hypothesis and the disparate functions of little v. In R.
D'Alessandro et al. (eds.), The verbal domain. Oxford: Oxford University Press, pp. 3-
28. https://doi.org/10.1093/oso/9780198767886.003.0001
Harris, J. 1995. The Morphology of Spanish Clitics. In H. Campos & P. Kempchinsky (eds.),
Evolution and revolution in linguistic theory. Washington: Georgetown U. Press.
Harris, J. W. & M. Halle. 2005. Unexpected plural inflections in Spanish. Linguistic Inquiry
36, pp. 195-222. https://doi.org/10.1162/0024389053710710
Harris, J. W. 1983. Syllable structure and stress in Spanish. Cambridge (Mass.): MIT Press.
Härtl, H. 2003. Conceptual and grammatical characteristics of argument alternations: The case
of decausative verbs. Linguistics 41 (5), pp. 883–916.
https://doi.org/10.1515/ling.2003.029
Haspelmaht, M. 1987. Transitivity alternations of the anticausative type. Cologne: Institut für
Sprachwissenschaft.
Haspelmath, M. 1990. The grammaticization of passive morphology. Studies in Language.
Amsterdam: John Benjamins. https://doi.org/10.1075/sl.14.1.03has
Haspelmath, M. 1993. More on the typology of inchoative / causative alternations. In B.
Comrie & M. Polinsky (eds.), Causatives and transitivity. Amsterdam: John Benjamins,
pp. 87-120. https://doi.org/10.1075/slcs.23.05has
Hatcher, A. G. 1957. Se hace algo vs. algo se hace. Modern Language Notes 73, pp. 102-107.
https://doi.org/10.2307/3043048
Haverkate, H. 1980. La construccion pseudorreflexiva y la estructura antropocéntrica de la
oración española. In Diálogos hispánicos en Amsterdam, 1. Los clíticos en el español
actual.
Heap, D. 1998. Optimizing Iberian Clitic Sequences. In J. Lema & E. Treviño (eds.),
Theoretical Analyses on Romance Languages. Amsterdam: John Benjamins.
https://doi.org/10.1075/cilt.157.13hea
Heggie, L. & F. Ordóñez (eds). 2005. Clitic and affix combinations. Theoretical perspectives.
Amsterdam: John Benjamins. https://doi.org/10.1075/la.74
Heidinger, S. 2014. The persistence of labile verbs in the French causative-anticausative
alternation. Linguistics 52, pp. 1003-1024.
Heim, I., H. Lasnik & R. May. 1991. Reciprocity and plurality. Linguistic Inquiry 22, pp. 63-
101.
Hernández Sacristán, C. 1986. En torno a los verbos pronominales y su doble origen.
Lingüistica Española Actual VIII, pp. 59-72.
https://doi.org/10.1515/iber.1991.1991.33.16
Hernández Sacristán, C. 1991. Configuración funcional y construcción refleja en español.
Iberorromania 33, pp. 16-27.
Higginbotham, J. 1980. Pronouns and bound variables. Linguistic Inquiry 11(4), pp. 679-708.
Hoeckestra, T. & I. Roberts. 1993. Middle constructions. In E. Reuland & W. Abraham (eds.),
Knowledge of language II. Lexical and conceptual structure. Dordrecht: Kluwer.

228
 SE IN SPANISH

Horvath, J. & T. Siloni. 2011. Causatives across components. Natural Language and Linguistic
Theory 29: 657-704. https://doi.org/10.1007/s11049-011-9135-3
Horvath, J. & T. Siloni. 2013. Anticausatives have no cause(r): A rejoinder to Beavers &
Koontz-Garboden. Lingua 131, pp. 217–30.
https://doi.org/10.1016/j.lingua.2013.02.013
Iglesias Bango, M. 1991. La voz en la gramdtica española. León: Universidad de León.
Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris.
https://doi.org/10.1515/9783112420225
Jaeggli, O. 1986. Three issues in the theory of clitics: Case, doubled NPs, and extraction. The
syntax of pronominal clitics, pp. 13-42. London: Brill.
https://doi.org/10.1163/9789004373150_003
Jespersen, O. 1927. A Modern English Grammar: On Historical Principles (Vol. Part III
Syntax. Second Volume). Copenhagen: Heidelberg.
Jiménez-Fernández, Á. & M. Tubino. 2014. Variación sintáctica en la causativización léxica.
Revista Española de Lingüística 44 (1): 7–37.
Jiménez-Fernández, Á. & M. Tubino. 2019. Causativity in southern peninsular Spanish. In The
syntactic variation of Spanish dialects, ed. Á. Gallego, 181–212. Oxford: OUP.
https://doi.org/10.1093/oso/9780190634797.003.0007
Jordan, P. G. 1973. La forma se como sujeto indefinido en español. Hispania 56, pp. 597-603.
https://doi.org/10.2307/339248
Kanski, Z. 1992. lmpersonal constructions as a strategy for second-order predication. In M.
Kefer & J. van der Auwera (eds.), Meaning and grammar. Cross-linguistic
perspectives. Berlin: Mouton de Gruyter, pp. 95-122.
Kayne, R. S. 1969. The transformational cycle in French syntax. PhD dissertation, MIT.
Kayne, R. S. 1975. French syntax: The transformational cycle. Cambridge (Mass.): MIT Press.
Kayne, R. S. 1989. Null subjects and clitic climbing. The null subject parameter. Springer:
Dordrecht, pp. 239-261). https://doi.org/10.1007/978-94-009-2540-3_8
Kayne, R. S. 1994. The antisymmetry of syntax. Cambridge (Mass.): MIT Press.
Kayne, R. S. 2000. Chapter 8. Person morphemes and reflexives in Italian, French and related
languages. In Parameters and universals. Oxford: Oxford University Press.
Kayne, R. S. 2005. Movement and silence. Oxford: Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780195179163.001.0001
Kayne, Richard S. 2010. Toward a syntactic reinterpretation of Harris and Halle (2005). In
Brigitte Kampers-Manhe, Reineke Bok-Bennema & Bart Hollebrandse (eds.), Romance
languages and linguistic theory 2008: Selected papers from ‘Going Romance’
Groningen 2008, 145–170. Amsterdam: John Benjamins.
https://doi.org/10.1075/rllt.2.09kay
Kempchinsky P. 2000. Aspect projections and predicate types. In H. Campos et al. (eds),
Hispanic linguistics at the turn of the millennium. Somerville, MA: Casacadilla Press,
pp. 171–83.
Kempchinsky, P. 2004. Romance se as an aspectual element. In J. Auger et al. (eds.),
Contemporary approaches to Romance linguistics”. Amsterdam: John Benjamins, pp.
239-256. https://doi.org/10.1075/cilt.258.13kem
Kempchinsky, P. 2006. Teasing apart the middle. In Andolin gogoan/Homenaje a Andolin
Eguzkitza, eds. Itziar Laka and Beatriz Fernández, 532–547. University of the Basque
Country Press.
Keyser, S. J. & T. Roeper. 1984. On the middle and ergative constructions in English.
Linguistic Inquiry 15 (3), pp. 381-416.
Klein, P. 1987. The syntax and semantics of Spanish SE. Hispanic Journal 9, pp. 149-162.

229
 ANTONIO FÁBREGAS

Koontz-Garboden, A. 2007. States, changes of state, and the monotonicity hypothesis. PhD
dissertation, Stanford University.
Koontz-Garboden, A. 2009. Anticausativization. Natural Language and Linguistic Theory 27,
pp. 77–138. https://doi.org/10.1007/s11049-008-9058-9
Kovacci, O. 1986. Estudios de gramática española. Buenos Aires: Hachete.
Kratzer, A. 1996. Severing the External Argument from its Verb. In J. Rooryck & L. Zaring
(eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer, pp. 109–37.
https://doi.org/10.1007/978-94-015-8617-7_5
Labelle, M. & E. Doron. 2010. Anticausative derivations (and other valency alternations) in
French. Probus 22 (2), pp. 303–16. https://doi.org/10.1515/prbs.2010.011
Labelle, M. 1992. Change of state and valency. Journal of Linguistics 28, pp. 375–414.
https://doi.org/10.1017/S0022226700015267
Labelle, M. 2008. The French reflexive and reciprocal se. Natural Language & Linguistic
Theory 26(4). https://doi.org/10.1007/s11049-008-9053-1
Lakoff, G. & S. Peters. 1969. Phrasal conjunction and symmetric predicates. In Modern studies
in English: Readings in transformational grammar, ed. David A. Reibel and Sanford
A. Schane, 113–142. Englewood Cliffs, NJ: Prentice-Hall.
Lakoff, G. 1971. Irregularities in syntax. New York: Holt, Rinehart and Winston.
Landa, Alazne. 1995. Conditions on null objects in Basque Spanish and their relation to leísmo
and clitic doubling. Los Angeles: University of Southern California dissertation.
Langendoen, D. T. 1978. The logic of reciprocity. Linguistic Inquiry 9, pp. 177-197.
Lekakou, M. 2005. In the Middle, Somewhat Elevated. The semantics of middles and its
crosslinguistic realization. Ph.D. dissertation, University of London.
Levin, B. & M. Rappaport Hovav. 1995. Unaccusativity. At the syntax-semantics interface.
Cambridge (Mass.): MIT Press.
Levin, B. 1993. Verb Classes in English. Cambridge (Mass.): MIT Press.
Llorente, A. 1977. Las construcciones de carácter impersonal en español. Estudios ofrecidos a
Emilio Alarcos, 1, Oviedo, pp. 107-125.
López, L. 2007. Locality and the architecture of syntactic dependencies. London: Palgrave-
MacMillan. https://doi.org/10.1057/9780230597471
Lozano, A. G. 1970. Non reflexiviry of the indefinite se in Spanish. Hispania 53, pp. 452-457.
https://doi.org/10.2307/337966
Lozano, A. G. 1972. The indefinite se revisited. Hispania 55, pp. 94-95.
https://doi.org/10.2307/338249
Lundquist, B. 2011. Restrictions on reflexive and anti-causative readings in nominalizations
and participles. In T. Strahan, Nordlyd 38, Related to Reflexives. Tromsø: CASTL, 1-
44. https://doi.org/10.7557/12.2028
MacDonald, J. & J. Melgares. 2021. On (un)grammatical clitic sequences in impersonal SE
constructions. In G. Armstrong & J. E. MacDonald (eds.), Unraveling the complexities
of SE. Dordrecht: Springer, pp. 87-111. https://doi.org/10.1007/978-3-030-57004-0_4
MacDonald, J. E. 2017. An implicit projected argument in Spanish impersonal and passive se
constructions. Syntax 20, pp. 353–383. https://doi.org/10.1111/synt.12146
MacDonald, J. E., & M. Maddox. 2018. Passive se in Romanian and Spanish: A linguistic
cycle. Journal of Linguistics 54, pp. 389–427.
https://doi.org/10.1017/S002222671700038X
Maddox, M. L. 2021. The development of SE from Latin to Spanish and the reflexive object
cycle. In G. Armstrong & J. E. MacDonald (eds.), Unraveling the complexities of SE.
Dordrecht: Springer, pp. 33-57. https://doi.org/10.1007/978-3-030-57004-0_2
Manteca Alonso-Cortés, A. 1976. En torno al se impersonal. Revista de la Sociedad Española
de Lingüística, 6, pp. 167-180.

230
 SE IN SPANISH

Manzini, M. R. 1986. On Italian si. In H. Borer (ed.), Syntax and semantics 19. The syntax of
pronominal clitics. New York: Academic Press, pp. 241-262.
https://doi.org/10.1163/9789004373150_011
Marantz, A. 1984. On the nature of grammatical relations. Linguistic Inquiry Monographs 10.
Cambridge (Mass.): MIT Press.
Marcos Marín, F. 1978. Estudios sobre el pronombre. Madrid: Gredos.
Marín, R. & A. Fábregas. 2018. Spanish 'estarse' is not only agentive, but also inchoative. In
Janine Berns, Haike Jacobs and Dominique Nouveau (eds.), Romance Languages and
Linguistic Theory 13. Amsterdam: John Benjamins, pp. 209-223.
https://doi.org/10.1075/rllt.13.13mar
Marín, R. 2011. Spanish adjectives within bounds”. In P. Cabredo Hofherr & O. Matushansky
(eds.), Formal analyses in syntax and semantics. Amsterdam, John Benjamins, pp. 307-
332. https://doi.org/10.1075/la.153.09mar
Martín Zorraquino, M. A. 1978. Desviaciones del sistema y de la norma de la lengua en las
construcciones pronominales españolas. Zaragoza: Universidad de Zaragoza.
Martín Zorraquino, M. A. 1979. Las construcciones pronominales en español: paradigmas y
desviaciones. Madrid: Gredos. https://doi.org/10.1075/z.9.iii52con
Martínez Martín, J. 1989. Las expresiones fijas de verbo pronominal en español: verbo +
complemento prepositivo. In J. Borrego Nieto et al. Philologica. Homenaje a A.
Llorente, Salamanca, págs. 179-193.
Martínez, J. A. 1981. Sobre el se en español. Revista de Filologia de la Universidad de La
Laguna, 0, pp. 87-92.
Maslova, Elena. 2007. Reflexive encoding of reciprocity: cross-linguistic and language-
internal variation. In König, Ekkehard & Gast, Volker (eds.), Reciprocals and
reflexives: cross-linguistic and theoretical explorations. Berlin: Mouton de Gruyter.
Masullo, P. 1990. Toward an Incorporation Account o f Argument Alternations in Spanish,
M.A. Thesis, University of Washington
Masullo, Pascual José. 1992. Incorporation and case theory in Spanish. A cross-linguistic
perspective. PhD diss., University of Washington.
McGinnis, M. 1998. Locality in A-movement. Doctoral dissertation, MIT.
Mendikoetxea, A. & A. Battye. 1990. Arb se/si in transitive contexts: A comparative study.
Rivista Di Grammatica Generative 15, pp. 161–195.
Mendikoetxea, A. 1992. On the nature of agreement: the syntax of ARB SE in Spanish. PhD
dissertation, University of York, GB.
Mendikoetxea, A. 1999. Construcciones inacusativas y pasivas. In I. Bosque & V. Demonte
(eds.), Gramática descriptiva de lengua española. Volumen 2, Madrid, Espasa Calpe,
pp. 1575-1630.
Mendikoetxea, A. 2000. Relaciones de interficie: los verbos de cambio de estado. Cuadernos
de Lingüística VII, pp. 125-144.
Mendikoetxea, A. 2002. La semántica de la impersonalidad. In C. Sánchez López (ed.), Las
construcciones con 'se'. Madrid: Visor, pp. 239-276.
Mendikoetxea, A. 2008. Clitic impersonal constructions in Romance: Syntactic features and
semantic interpretation. In Impersonal constructions in grammatical theory, A.
Siewierska (ed.) Special issue of the transactions of the philological society, 106, 2, pp.
290–336. Oxford: Blackwell. https://doi.org/10.1111/j.1467-968X.2008.00210.x
Mendikoetxea, A. 2012. Passives and se constructions. In J. I. Hualde, A. Olarrea & E.
O'Rourke (eds.), The handbook of Spanish linguistics. London: Wiley-Blackwell, pp.
477-502. https://doi.org/10.1002/9781118228098.ch23

231
 ANTONIO FÁBREGAS

Michaelis, S. 1998. Antikausativ als Brücke zum Passiv: fieri, venire und se im
Vulgärlateinischen und Altitalienischen. In Neuere Beschreibungsmethoden der Syntax
romanischer Sprache, ed. Wolfgang Dahmen et al., 69–98. Tübingen: Narr.
Molina Redondo, J.A. 1974. Usos de se. Cuestiones sintácticas y léxicas. Madrid, SGEL.
Monge, F. 1955. Las frases pronominales de sentido impersonal en español. Archivo de
Filología Aragonesa VII, pp. 1-102.
Moreno Cabrera, J. C. 1990. Processes and actions: internal agentless impersonals in some
european languages. In J. Betrechrt, G. Bernini & C. Buridant (eds.), Toward a typology
of european languages. Empirical approaches to language typology 8. Berlín, Walter
de Gruyter, pp. 255-272. https://doi.org/10.1515/9783110863178.255
Morimoto, Y. 2008. 'Me estuve quieto': el concepto de estado y el llamado se aspectual. In
Actas del XXXVII Simposio Internacional de la Sociedad Española de Lingüística, ed.
by Inés Olza Moreno, Manuel Casado Velarde and Ramón González Ruiz, pp. 591–
599. Pamplona: Servicio de Publicaciones de la Universidad de Navarra.
Morimoto, Y. 2011. El control en los predicados estativos. In Estados y estatividad, ed. by
Ángeles Carrasco Gutiérrez, pp. 122–141. Muenchen: LINCOM.
Nedjalkov, V. P. 1969. Nekotorye verojatnostnye universalii v glagol'nom slovoobrazovanii.
In I. F. Vardul' (ed.). Jazykovye versalii i lingvisticeskaja tipologija. Moscú: Nauka, pp.
106-114.
Nedjalkov, Vladimir P. (ed.). 2007. Reciprocal constructions. Amsterdam: John Benjamins.
https://doi.org/10.1075/tsl.71
Nevins, A. 2007. The representation of third person and its consequences for person-case
effects. Natural Language & Linguistic Theory 25 (2), pp. 273–313.
https://doi.org/10.1007/s11049-006-9017-2
Nevins, A. 2011. Multiple agree with clitics: Person complementarity vs. omnivorous number.
Natural Language & Linguistic Theory, 29(4), 939-971.
https://doi.org/10.1007/s11049-011-9150-4
Nevins, A. 2012. Haplological dissimilation at distinct stages of exponence. In Jochen
Trommer (ed.), The morphology and phonology of exponence: The state of the art, 84–
116. Oxford: Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199573721.003.0003
Nishida, C. 1994. The Spanish reflexive se as an aspectual class marker. Linguistics 32 (3), pp.
425-458. https://doi.org/10.1515/ling.1994.32.3.425
Oca, E. 1914. El pronombre se en nominativo. Boletín de la Real Academia española 1, pp.
573-581.
Oltra-Massuet, I. 2014. Deverbal adjectives at the interface. Berlin: De Gruyter.
https://doi.org/10.1515/9781614510659
Omori, Y. 1997. Ambiguity in the constructions with se. Sophia Linguistica 41, pp. 115–131.
Ordóñez, F. & E. Treviño. 2011. Impersonals with passive morphology. In Selected
proceedings of the 13th Hispanic Linguistics symposium, ed. Luis A. Ortiz-López, pp.
314–324.
Ordóñez, F. 2021. On the nature of impersonal SE: why Italian is not like Catalan and Spanish.
In G. Armstrong & J. E. MacDonald (eds.), Unraveling the complexities of SE.
Dordrecht: Springer, pp. 137-161. https://doi.org/10.1007/978-3-030-57004-0_6
Ordoñez, Francisco 2012. Clitics in Spanish. In J.I. Hualde, A. Olarrea & E. O’Rourke (eds.)
The handbook of Hispanic linguistics. Wiley-Blackwell.
https://doi.org/10.1002/9781118228098.ch21
Ordóñez, Francisco, and Esthela Treviño. 2016. Agreement and DOM with impersonal Se: A
comparative study of Mexican and Peninsular Spanish. In The Morphosyntax of

232
 SE IN SPANISH

Portuguese and Spanish in Latin America, ed. M. Kato and F. Ordóñez. Oxford: Oxford
University Press. https://doi.org/10.1093/acprof:oso/9780190465889.003.0010
Ormazabal, J. & J. Romero. 2013. Object clitics, agreement and dialectal variation. Probus:
International Journal of Latin and Romance Linguistics 25(2), pp. 301-344.
https://doi.org/10.1515/probus-2013-0012
Ormazabal, J. & J. Romero. 2019. The formal properties of non-paradigmatic SE. Borealis:
International Journal of Hispanic Linguistics 8, pp. 55-84.
https://doi.org/10.7557/1.8.1.4704
Ormazabal, J. & J. Romero. 2021. Deconstructing SE constructions: number agreement and
post-syntactic variation. Ms., University of the Basque Country and University of
Extremadura.
Otero, C. 1966. Gramaticalidad y normativismo. A propósito de algunos escritos de A.
Rosemblat. Romance Philology XX, pp. 53-68.
Otero, C. 1986. Arbitrary subjects in finite dauses. In I. Bordelois et al. (eds.), Generative
studies in Spanish syntax. Dordrecht: Foris, pp. 81-109.
https://doi.org/10.1515/9783110859232-007
Otero, C. 1999. Pronombres reflexivos y recíprocos. In I. Bosque & V. Demonte (dirs.),
Gramática descriptiva de la lengua española, pp. 1427–1517. Madrid: Espasa-Calpe.
Otero, C. 2002. Facetas de 'se'. In C. Sánchez López (ed.), Las construcciones con 'se'. Madrid:
Visor, pp. 168-210.
Perlmutter, D. 1971. Surface structure constraints in syntax. Linguistic Inquiry, 1, pp. 187-256.
Pescarini, D. 2010. Elsewhere in romance: Evidence from clitic clusters. Linguistic Inquiry 41,
pp. 427–444. https://doi.org/10.1162/LING_a_00003
Pesetsky, D. & E. Torrego. 2007. The syntax of valuation and the interpretability of features.
Phrasal and clausal architecture: Syntactic derivation and interpretation, pp. 262-294.
https://doi.org/10.1075/la.101.14pes
Pesetsky, D. 1995. Zero syntax: experiencers and cascades. Cambridge (Mass.): MIT Press.
Piñón, C. 2001. Modeling the causative-inchoative alternation. Linguistische Arbeitsberichte
76, pp. 273-293. https://doi.org/10.3765/salt.v11i0.2858
Pollard, C. & I. A. Sag. 1992. Anaphors in English and the Scope of Binding Theory. Linguistic
Inquiry 23, pp. 261-303.
Postal, P. 1966. On so-called pronouns in English. Monograph series on language and
linguistics 19, pp. 177-206.
Preminger, O. 2014. Agreement and its failures. Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/9780262027403.001.0001
Pujalte, M. & A. Saab. 2012. Syncretism as PF-repair: The case of se-insertation in Spanish.
In The end of argument structure? ed. M.C. Cuervo & Y. Roberge, 229–260. Bingley:
Emerald Press. https://doi.org/10.1163/9781780523774_011
Pujalte, Mercedes. 2013. Argumentos (no) agregados. Indagaciones sobre la morfosintaxis de
la introducción de argumentos en español. Ph.D. thesis, University of Buenos Aires.
Pylkkänen, L. 2002. Introducing arguments. PhD dissertation, MIT.
Quine, W. V. O. 1961. Variables explained away. Proceedings of the American Philosophical
Society 104(3), pp. 343-347
RAE & ASALE. 2009. Nueva gramática de la lengua española. Madrid: Espasa.
Ramchand, G. 2008. Verb meaning and the Lexicon: a first phase syntax. Cambridge:
Cambridge University Press. https://doi.org/10.1017/CBO9780511486319
Ramchand, G. 2018. Situations and syntactic structures. Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/9780262037754.001.0001
Raposo, E. & J. Uriagereka. 1996. Indefinite SE. Natural Language & Linguistic Theory 14
(4), pp. 749–810. https://doi.org/10.1007/BF00133363

233
 ANTONIO FÁBREGAS

Rappaport Hovav, M. & B. Levin. 2010. Reflections on Manner/Result Complementarity. In

M. Rappaport Hovav, E. Doron, and I. Sichel (eds.), Syntax, Lexical Semantics, and
Event Structure. Oxford: Oxford University Press, 21–38.
https://doi.org/10.1093/acprof:oso/9780199544325.003.0002
Reinhart, T. & E. Reuland. 1993. Reflexivity. Linguistic Inquiry 24, pp. 657–720.
Reinhart, T. & T. Siloni. 2005. The lexicon–syntax parameter: reflexivization and other arity
operations. Linguistic Inquiry 36, pp. 389–436.
https://doi.org/10.1162/0024389054396881
Reinhart, T. 1996. Syntactic effects of lexical operations: reflexives and unaccusatives. OTS
working papers in linguistics, University ofUtrech.
Reinhart, T. 2000. The theta system: syntactic realization of verbal concepts, OTS Working
papers, 00.01/TL, University of Utrecht.
Reuland, E. 2011. Anaphora and language design. Cambridge (Mass.): MIT Press.
https://doi.org/10.1093/obo/9780199772810-0050
Ricós, A. 1998. La pasiva con se agentiva en los textos jurídicos administrativos: su incidencia
pragmática. Estudios lingülsticos de la Universidad de Alicante 12, pp. 195-209.
https://doi.org/10.14198/ELUA1998.12.12
Rigau, G. 1994. Les propietats dels verbs pronominals. Els Marges 50, pp. 29-39.
Rizzi, L. 1986. Null objects in Italian and the theory of 'pro'. Linguistic inquiry 17(3), pp. 501-
558.
Roberts, I. 1987. The representation of implicit and dethematized subjects. Dordrecht: Foris.
https://doi.org/10.1515/9783110877304
Roberts, I. 2010. Agreement and head movement: Clitics, incorporation, and defective goals.
Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/9780262014304.001.0001
Roca, F. 1992. Object clitics in Spanish and Catalan. Catalan Working Papers in Linguistics
2, pp. 245–280.
Rodríguez Díez, B. 1996. Valores de /se/. Estudios filológicos en homenaje a Eugenio de
Bustos Tovar, edited by J. A. Bartol Hernández, J. F. García Santos & J. de Santos
Guervós. Salamanca: Universidad de Salamanca, pp. 817-830.
Rodríguez Molina, J. 2010. La gramaticalización de los tiempos compuestos en español
antiguo. PhD thesis, Universidad Autónoma de Madrid.
Romero Pascual, C. & I. I. Teomiro García. 2012. La relación entre estructura eventiva y
papeles temáticos: el se aspectual del español. Revista de Filología Románica 29 (2),
pp. 233–243. https://doi.org/10.5209/rev_RFRM.2012.v29.n2.40156
Romero, J. & J. Ormazabal. 2019. The formal properties of non paradigmatic ‘se’. Borealis.
An International Journal of Hispanic Linguistics 8: 55–84.
https://doi.org/10.7557/1.8.1.4704
Saab, A. 2014. Syntax or nothing: Some theoretical and empirical remarks on implicit
arguments. Borealis: An International Journal of Hispanic Linguistics 3: 125–183.
https://doi.org/10.7557/1.3.2.2952
Saltarelli, M. 1994. Voice in Latin and romance: On the representation of lexical subjects. In
Issues and theory in romance linguistics, ed. Michael Mazzola, 445–478. Washington:
Georgetown University Press.
Sánchez López, C. & M. Vivanco. In press. La voz media: estructura argumental. In C.
Company Company (dir.), Sintaxis Histórica de la lengua española. [To appear]
México: Fondo de Cultura Económica.
Sánchez López, C. 2002. Las construcciones con se. Estado de la cuestión. In C. Sánchez López
(ed.), Las construcciones con ‘se’. Madrid: Visor Libros.

234
 SE IN SPANISH

Sanz, M. & I. Laka. 2002. Oraciones transitivas con se: el modo de acción en la sintaxis. In C.
Sánchez López (ed.), Las construcciones con se. Madrid: Visor.
Sanz, M. 1995. Telicity. objects and the mapping onto predicate types. A cross-linguistic
analisys of the role of syntax in processing. PhD dissertation, University of Rochester.
Schäfer, F. & M. Vivanco. 2016. Anticausatives are weak scalar expressions, not reflexive
expressions. Glossa: A Journal of General Linguistics 1(1), 18, pp. 1–36.
https://doi.org/10.5334/gjgl.36
Schäfer, F. 2007. On the nature of anticausative morphology: External arguments in change-
of-state contexts. PhD dissertation. Opus - Elektronische Hochschulschriften der
Universität Stuttgart, Universitätsbibliothek Stuttgart. https://doi.org/10.1075/la.126
Schäfer, F. 2008. The syntax of (anti)causatives. External arguments in change of state
contexts. Amsterdam: John Benjamins. https://doi.org/10.1075/la.126
Schäfer, F. 2009. The causative alternation. Language and Linguistics Compass 3 (2), pp. 641-
681. https://doi.org/10.1111/j.1749-818X.2009.00127.x
Schäfer, F. 2012. Two types of external argument licensing – the case of causers. Studia
Linguistica 66 (2), pp. 28–80. https://doi.org/10.1111/j.1467-9582.2012.01192.x
Schroten, J. 1979. Las oraciones finitas que carecen de sujeto léxico: análisis estructural y
análisis generativo. Archivuum XXIX-XXX, pp. 507-534.
Sells, P. 1987. Aspects of logophoricity. Linguistic Inquiry 18(3), pp. 445-479.
Smith, C. S. 1970. Jespersen’s ‘Move and Change’ Class and Causative Verbs in English. In
M. A. Jazayery, E. C. Polomé, & W. Winter (eds.), Linguistic and Literary Studies in
Honor of Archibald A. Hill. Vol. 2: Descriptive Linguistics. The Hague: Mouton de
Gruyter, pp. 101–9. https://doi.org/10.1515/9783110800432.101
Sportiche, D. 1996. Clitic constructions. In Johan Rooryck & Laurie Zaring (eds.), Phrase
structure and the lexicon, 213–276. Dordrecht: Kluwer. https://doi.org/10.1007/978-
94-015-8617-7_9
Stroik, T. 1992. Middles and movement. Linguistic Inquiry 23 (1), pp. 127-137.
Stroik, T. 1995. On middle formation: a reply to Zribi-Hertz. Linguistic Inquiry 26 (1), pp. 165-
172.
Strozer, J. 1976. Clitics in Spanish. UCLA Ph.D dissertation.
Suñer, M. 1973. Non paradigmatie ses in Spanish. PhD thesis, Indiana University at
Bloomington.
Suñer, M. 1974. Where does impersonal se come from? Proceedings of the third linguistic
symposium on Romance languages, Georgetown University Press. R. J. Campbel, M.
G. Goldin & M.E. Wang (eds.), pp. 146-157.
Suñer, M. 1983. Pro arb. Linguistic Inquiry 14, pp. 188-191.
Suñer, M. 1988. The role of agreement in clitic-doubled constructions. Natural Language and
Linguistic Theory 6: 391–434. https://doi.org/10.1007/BF00133904
Suñer, M. 1990. Impersonal Se passives and the licensing of empty categories. Probus 2 (2):
209–231. https://doi.org/10.1515/prbs.1990.2.2.209
Suñer, M. 1992. Subject clitics in the Northern Italian vernaculars and the matching hypothesis.
Natural Language and Linguistic Theory 10, pp. 641–672.
https://doi.org/10.1007/BF00133332
Suñer, M. 2002. Las pasivas con 'se' inpersonal y la legitimación de las categorías vacías. In C.
Sánchez López (ed.), Las construcciones con 'se'. Madrid: Visor, pp. 210-239.
Talmy, L. 2000. Toward a cognitive semantics. Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/6848.001.0001
Teomiro, I. (2005). Control theory and the timing of PRO. Master theory, Utrecht University.

235
 ANTONIO FÁBREGAS

Teomiro, I. (2010). Anaphors at the Interfaces. A Comparative Study of the Variation of the
Anaphoric Systems of English, Dutch and Spanish. Doctoral dissertation, Universidad
Autónoma de Madrid, Madrid.
Teomiro, I. (2011). Reflexivity and adjustment strategies at the interfaces. Nordlyd: Tromsø
Working Papers in Linguistics 38, pp. 119-149. https://doi.org/10.7557/12.2026
Torrego, E. 1995. From argumental to non-argumental pronouns: Spanish doubled reflexives.
Probus 7(2), pp. 221-241. https://doi.org/10.1515/prbs.1995.7.2.221
Torrego, E. 1998. The dependencies of objects. Cambridge (Mass.): MIT Press.
https://doi.org/10.7551/mitpress/2337.001.0001
Tremblay, A. 2006. On the second language acquisition of Spanish reflexive passives and
reflexive impersonals by French- and English-speaking adults. Second Language
Research 22, pp. 30–63. https://doi.org/10.1191/0267658306sr260oa
Trujillo, R. 1988. Sobre las construcciones pasivas. Lingüística Española Actual 10/2.
Uriagereka, J. 1988. On Government. Doctoral dissertation, dissertation, University of
Connecticut.
Uriagereka, J. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic
Inquiry 26(1), pp. 79-123.
Vivanco, M. 2021. Scalar constraints on anticausative SE: the aspectual hypothesis revisited.
In G. Armstrong & J. E. MacDonald (eds.), Unraveling the complexities of SE.
Dordrecht: Springer, pp. 291-321. https://doi.org/10.1007/978-3-030-57004-0_12
Vivanco, Margot. 2016. Causatividad y cambio de estado en español: la alternancia causativo-
inacusativa. PhD dissertation, Universidad Complutense de Madrid.
Volkova, A. & E. Reuland. 2014. Reflexivity without reflexives? The Linguistic Review 31,
pp. 587-644. https://doi.org/10.1515/tlr-2014-0012
Walkow, M. 2012. Goals, big and small. Cambridge, MA: University of Massachusetts
dissertation.
Walkow, M. 2013. Locating variation in person restrictions. University of Pennsylvania
Working Papers in Linguistics 19, pp. 247–256.
Wolfsgruber, A. C. 2021. Null subjects and SE revisited: what Medieval Romance varieties
reveal. In G. Armstrong & J. E. MacDonald (eds.), Unraveling the complexities of SE.
Dordrecht: Springer, pp. 57-87. https://doi.org/10.1007/978-3-030-57004-0_3
Wood, J. 2014. Reflexive -st verbs in Icelandic. Natural Language and Linguistic Theory 32,
pp. 1387–1425. https://doi.org/10.1007/s11049-014-9243-y
Wood, J. 2015. Icelandic morphosyntax and argument structure. Dordrecht: Springer
https://doi.org/10.1007/978-3-319-09138-9
Wunderlich, D. 1997. Cause and the structure of verbs. Linguistic Inquiry 28, pp. 27-68.
Zagona, K. 1996. Compositionality of aspect: Evidence from Spanish aspectual se. In C. Parodi
et al. (eds.), Aspects of romance linguistics. Selected from the SLSR XXIV. Georgetown:
Georgetown University Press, pp. 475-488.
Zribi-Hertz, A. 1987. La réfrexivité ergative en français moderne. Le Français Moderne 55
(1), pp. 23-52.
Zribi-Herzt, A. 1993. On Stroik’s analysis of English middle. Linguistic Inquiry 24 (3), pp.
583-589.

236