The depictive secondary predication construction has two subject-predicate relationships in one clause, providing two propositional interpretations. This article proposes that the primary predication, which consists of a main verb, and the secondary predication, made up of a secondary predicate, are simultaneously derived in separate derivational planes; this proposal reflects a property of the secondary predicate that is not included in the θ-grid of the primary predicate. The idea of using Parallel Merge to merge these two planes that share a common element allows us to understand the secondary predication construction in a novel way.
1 Spanish Depictive Secondary Predicates
The aim of this article is to propose a new and economical derivational structure for Spanish depictive secondary predication in the Minimalist framework. It is well-known that depictive secondary predicates (henceforth, DSPs), which are not subcategorized for by the matrix verb, can establish a predication relation with the object or the subject of the matrix clause.1
Juan comió la carne cruda.
Juan ate the meat.FEM.SG raw.FEM.SG
‘Juan ate the meat raw.’
Marta tomó el café descalza.
Marta drank the coffee barefoot.FEM.SG
‘Marta drank the coffee barefoot.’
In this article, I propose a new derivational structure for the Spanish DSP construction. In building my proposal, I use Bowers’s (1993, 2001) PrP structure for each predication (i.e., the primary predication of the matrix verb and the secondary predication of the DSP) to efficiently capture the predication relation between the predicate and its subject. Bowers’s analysis is illustrated in (2).
Johni sketched the modelj nudej drunki.
[PrP Johni [Pr´ [Pr´ sketchk [VP the modelj [V´ tk [PrP PROj nude]]]] [PrP PROi drunk]]] (Bowers 2001:327)
As shown in (2b), Bowers assumes PRO in order to establish the predication relation between the DSP and its subject. However, I will present a new analysis that accounts for the predication relation of DSP constructions without the need to introduce PRO into the derivation.2
Given the structure of (2b), DSPs such as drunk and nude cannot move out of the adjoined PrP since PrP forms an adjunct island. However, there are some cases in which DSPs undergo wh-movement out of PrPs in Spanish, as shown in (3) and (4). As (5) shows, this movement is impossible in English.
¿Qué tan crudai comió Juan la carne ti?
how much raw.FEM.SG ate Juan the meat.FEM.SG
‘How raw was the meat that Juan ate?’
¿Qué tan emocionadai recibió Ana el premio ti ?
how much happy.FEM.SG received Ana the prize
‘How happy was Ana when she received the prize?’
¿En cuántas partesi encontraste roto ti el libro?
in how.many parts found torn.up.MASC.SG the book.MASC.SG
‘Into how many pieces did you find the book torn up?’
¿De qué farosi compró desprovisto ti el coche?3
of what lights bought lacking.MASC.SG the car.MASC.SG
‘What lights did he buy the car without?’
*How rawi did John eat the meat ti?
*How happyi did Ana receive the prize ti?
Here, I provide a new structure for the Spanish DSP construction, which does not postulate PRO yet allows the DSP’s wh-movement in (3) and (4) to be satisfactorily analyzed. Concretely, I propose a multidominance structure for the Spanish DSP construction without assuming an adjoined structure such as (2b).
2 Parallel Derivation of the DSP Construction
The DSP constructions in (1a–b) each consist of two predication relations, which share an element that functions as object (1a) or subject (1b) of the matrix clause and, at the same time, as the DSP’s subject. To account for this property, I propose that the two predication structures are derived in separate planes and that the two syntactic objects that represent their proper predication relation form a multidominance structure (e.g., Wilder 1999, 2008, Citko 2005, Van Riemsdijk 2006, De Vries 2009). More precisely, in the course of the derivation, the two syntactic objects are joined together, taking a shared element as a grafting point (borrowing a term from Van Riemsdijk 2006) via Parallel Merge (Citko 2005). This means that the primary predication and the secondary predication are formed separately.4 In this sense, we can paraphrase (1a) as Juan comió la carne y la carne estaba cruda ‘Juan ate the meat and the meat was raw’. Descriptively speaking, the DSP construction is one in which two different derivations with clausal interpretation share a nominal phrase and have their own predicates.
Under my proposal, a parallel derivation for (1a), a Spanish DSP construction with an object- oriented predicate (OOP), would look like this:
The derivation of the category PrP for primary predication in plane 1 and that of the same category for secondary predication in plane 2 proceed in parallel and share an element (the direct object la carne ‘the meat’) that functions as a pivot between them.
In the same vein, the derivation of (1b), a DSP construction with a subject-oriented predicate (SOP), is represented as in (7). Taking a shared element (the subject Marta) as a grafting point, the two syntactic objects are formed in parallel.
PrP1 and PrP2 , derived in parallel in (6) and (7), are merged when a common element between them enters into the derivation. Parallel Merge takes this common element as a grafting point and joins the two independent derivations.
The following constituency tests at the VP level support my proposal that the DSP is derived on a separate plane from primary predication:
Juan cantó el lied contento y Pedro disgustado.
Juan sang the lied happy and Pedro angry
‘Juan sang the lied happy and Pedro angry.’
(Demonte 1988:11, (23a))
Juan comió la carne cruda y Pedro cocida.
Juan ate the meat raw and Pedro cooked
‘Juan ate the meat raw and Pedro cooked.’
Juan comió la carne disgustado y Ana lo hizo contenta.
Juan ate the meat angry and Ana 3.CL did happy
‘Juan ate the meat angry and Ana did it happy.’
Juan comió la carne cruda y Pedro lo hizo cocida.
Juan ate the meat raw and Pedro 3.CL did cooked
‘Juan ate the meat raw and Pedro did it cooked.’
As shown in (8), the DSP can be left behind when gapping is applied to the DSP construction. This means that the SOP disgustado and the OOP cocida can be separated from VP. Examples (9a–b), in which the hacerlo ‘do it’ test is applied to the DSP construction, are grammatical—also showing that VP can be separated from the SOP contenta and the OOP cocida.
The following evidence also supports my proposal that PrP2 is derived on an independent plane. More than one DSP predicated of the same subject can appear in the same clause, as shown in (10) (see Rothstein 1985).
Juan comió la carnei crudai, tiernai.
Juan ate the meat.FEM.SG raw.FEM.SG tender.FEM.SG
‘Juan ate the meat raw, tender.’
Juani jugaba descalzoi , casi desnudoi.
Juan played barefoot.MASC.SG almost naked.MASC.SG
‘Juan used to play barefoot, almost naked.’
The cooccurrence of DSPs shows that (a) a DSP is not an element lexically selected by the matrix verb and (b) each syntactic object that represents a proper predication relation between a DSP and its subject is constructed independently on its own derivational plane.
3 Wh-Movement in Spanish DSPs
In (6) and (7), PrP1 and PrP2 only share one common element and they are floating around in the derivational space. Subsequently, PrP1 of the primary predication and PrP2 of the secondary predication form a single syntactic object, PrP1+2 , as shown in (11). One might ask why PrP1 and PrP2 should be assembled at the level of PrP. Following Chomsky (2001), the interpretation/evaluation of the derivation takes place by phase. If we assume that PrP is a phase, we have two phases in parallel in (6) and (7): PrP1 and PrP2 . These cannot be interpreted or evaluated since they violate the Single Syntactic Object Condition.5 Therefore, they should be assembled at the level of PrP.
In (11), two syntactic objects, PrP1 and PrP2 , form a multidominance structure and they are assembled into a symmetrically labeled constituent PrP1+2.6 The matrix Asp selects PrP1+2. I argue that the functional head Asp simultaneously enters into an Agree relation with all matched goals, namely, Pr1 and Pr2 (Multiple Agree; Hiraiwa 2000), and that both Pr1 and Pr2 move to the same head Asp, given that V-to-T movement takes place in Spanish. For this multiple head movement, I propose a head cluster formation, adopting the Cluster Hypothesis suggested by Sabel (2001).7 Concretely, the feature of Pr1 that is attracted by Asp attracts the same type of feature of Pr2 , and then the head cluster formed in Asp moves to T as a complex head, Pr1 + Pr2 , as shown in (11).8 Following Gallego’s (2006, 2010) phase sliding proposal, a nonphase head may inherit phasehood when a phase head moves to it. Therefore, I assume that in Spanish, AspP functions as a phase by Pr-to-Asp movement. As a result, in (11), AspP inherits phasehood from PrP1 and PrP2 and displays phase effects.
Here, I assume that the DSP’s wh-movement actually consists of two movements. The first one is the DSP’s movement to the edge of AspP, triggered by the edge feature of Asp;9 and the second one is the movement from the edge of AspP to Spec,C triggered by the wh-feature, as shown in (11).10 In this sense, it can be supposed that DSP movement is possible in languages in which V-to-T movement (i.e., Pr-to-Asp movement in this article) causes phase sliding, but DSP movement is not possible in languages such as English in which V-to-T movement does not exist.11 I believe that the difference between English and Spanish with respect to DSP movement (see (3)–(5)) can be explained by phase sliding and V-to-T movement in my framework. Concretely, given that V-to-T movement does not occur in English, phases are not extended and movement to the edge of an extended phase is impossible. As a result, under my proposal, the extraction is possible from either PrP1 or PrP2 , given that PrP1 and PrP2 form a multidominance structure and the matrix Asp agrees with the symmetrically labeled constituent PrP1+2 (see Culicover and Jackendoff 1999, Hiraiwa and Bodomo 2008, Citko 2011). Therefore, the structure proposed in this article, in which two syntactic objects form a multidominance structure, is free from the Coordinate Structure Constraint (Ross 1967).
4 Evidence for Parallel Derivation of the DSP Construction
In this section, I provide two pieces of empirical evidence for the claim that the DSP construction is derived in separate planes.
First, it is generally assumed that two focus elements cannot cooccur in the left periphery of a clause (see Cinque 1983, 1990, Rizzi 1997, Zubizarreta 1998, 1999). In this sense, the focus element is not compatible with wh-words.
Turning to the DSP construction, we see that the focus elements in (13) do cooccur with wh-interrogatives.
¿Quién (CASI) BORRACHO condujo el coche?
who (ALMOST) DRUNK.MASC.SG drove the car
‘Who drove the car almost drunk?’
¿Quéplato (TODAVÍA) CALIENTE le sirvió Marta?
what dish.MASC.SG (STILL) HOT.MASC.SG 3.CL served Marta
‘What dish did Marta serve still hot to him/her?’
¿Qué canción (CASI) EMOCIONADA está cantando Marta?
what song (ALMOST) EXCITED.FEM.SG is singing Marta
‘Which song is Marta singing almost excited?’
The grammaticality contrast between (12) and (13) is naturally explained under the parallel derivation analysis of the DSP construction. Concretely, in (13a), the probe C seeks focus elements of each plane, finding quién and BORRACHO and simultaneously entering into an Agree relation with them (Hiraiwa 2000), since this instance of Agree does not trigger defective intervention effects (Chomsky 2000).12 The probe C raises them to the CP domain, forming a complex focus phrase. I propose (adopting the Cluster Hypothesis; see Rudin 1988, Grewendorf 2001, Sabel 2001) that this phrase is constructed by focus cluster formation. (13a) can be represented as in (14).
Concretely, I assume that a focus cluster is formed in Spec,Asp by adjunction of a focused DSP to a moved wh-element with the same type of feature (i.e., a focus feature) and that the final landing site of the focus cluster is Spec,C.13 In fact, the movement order of the focused DSP and the wh-phrase to Spec,Asp cannot be determined by closeness to Asp, given that the two elements are derived on different planes. Nevertheless, the reason why the focused DSP should be adjoined to the moved wh-phrase for cluster formation as shown in (14) is that the wh-operator cannot bind its variable in adjoined position to the moved focus element. However, examples like (12a–b) with two focus elements derived in one plane are ruled out by the Defective Intervention Constraint.14 When C Agrees with one of the two focus elements, a checking relation with the other focus element is not established, given that one focus element asymmetrically c-commands the other.15
Furthermore, multiple focus is observed in the following examples, which involve nouns that can have their own argument structures:
¿De qué cantante BORROSAS salieron las fotos?
of what singer BLURRY came.out the photographs
‘What singer’s photographs came out blurry?’
¿Qué parte del cuerpo CON ANESTESIA GENERAL le operaron a what part of.the body WITH ANESTHESIA GENERAL 3.CL operated to Marta?
‘What part of the body did they perform an operation on Marta under general anesthesia?’
In (15a), the representational noun foto, like a verb, can have its own argument structure that consists of object, agent, and possessor (e.g., Grimshaw 1990, Demonte and Masullo 1999, Picallo 1999, Suñer 1999). The noun phrase las fotos and the wh-phrase de qué cantante that functions as its object are merged on an independent plane. The wh-phrase de qué cantante, extracted out of DP, can move to Spec,C (Ticio 2005) and can cooccur with the focus element BORROSAS, derived on a different plane. In (15b), the wh-phrase qué parte del cuerpo, which functions as direct object of the matrix clause, is a body-part noun. A body-part noun can have its own argument structure, taking a possessor argument (here, Marta), and this possessor can establish a predication relation with its predicate (here, CON ANESTESIA GENERAL) (see Demonte and Masullo 1999, Rigau 1999, Suñer 1999). The possessor and its predicate form an independent syntactic object. (15b) also shows that the wh-phrase qué parte del cuerpo and the focus element CON ANESTESIA GENERAL, derived on separate planes, can form a multiple focus. This evidence supports my proposal that two focus elements derived on different planes can undergo multiple focus fronting.
A second piece of evidence showing that the DSP construction is derived in separate planes is this. It is generally assumed that a pronoun must be located in the c-command domain of a quantifier to receive a bound reading (Reinhart 1983). Now consider examples that involve a subject quantifier and a pronoun embedded in an OOP.
Pintó cada pintor a la modelo sentada en la silla.
painted each painter.MASCA the model.FEM seated.FEM.SG on the chair
‘Each painter painted the model seated on the chair.’
Lanzó cada amigo a la amiga vestida con el abrigo.
threw each friend.MASCA the friend.FEM dressed.FEM.SG in the coat
‘Each friend threw the friend dressed in the coat.’ (Context: People are throwing others overboard to save them from a burning ship.)
*Pintó cada pintori a la modelo sentada en sui silla.
painted each painter.MASC A the model.FEM seated.FEM.SG on his chair
‘Each painter painted the model seated on his chair.’
*Lanzó cada amigoi a la amiga vestida con sui abrigo.
threw each friend.MASCA the friend.FEM dressed.FEM.SG in his coat
‘Each friend threw the friend dressed in his coat.’ (Context: People are throwing others overboard to save them from a burning ship.)
While (16a–b) without a bound pronoun are grammatical, (17a–b) with one are not. This means that the ungrammaticality is due to the quantifier binding. In (17a–b), the postverbal subject is structurally higher than the OOP and, in conventional terms, the subject c-commands into the OOP.16 Thus, it is predicted that a bound reading should be possible between the subject quantifier and the pronoun in the OOP. However, this prediction is not borne out. This phenomenon is accounted for if it is assumed that the quantifier and the pronoun are not derived in the same plane. The two syntactic objects (i.e., PrP1 and PrP2 ) have been constructed separately in a three- dimensional space in (6) and (7), and they are two independent c-command units. A c-command unit is a syntactic object, such as PrP1 and PrP2 , formed by the monotonic application of Merge (Uriagereka 1999). This process is related to a multidimensional derivation in which each syntactic object forms a different c-command unit. Therefore, the elements of PrP1 cannot enter into a c-command relation with those of PrP2 , and no binding relation is established between PrP1 and PrP2 (see Espinal 1991, De Vries 2012, Griffiths and De Vries 2013). In this sense, the binding facts observed in the DSP construction support my proposal. That is to say, the impossibility of a bound reading can be explained by the claim that the quantifier cannot c-command the pronoun located in another plane in (17a–b). Furthermore, the subjects in (17a–b) are intentionally postverbal. This is to demonstrate that the subject quantifier and the pronoun are derived on different planes, since I suppose that the postverbal subject occupies a position in PrP1.17
In addition, consider examples that involve an indirect object quantifier and a pronoun embedded in an SOP.
Marta le cantó a cada amigo el lied entusiasmada con el
Marta 3.CL sang to each friend.IO.MASC.SG the lied.DO excited.FEM.SG at the éxito.
‘Marta sang the lied to each friend excited at the success.’
Juan le contaba a cada amiga la historia orgulloso del
Juan 3.CL told to each friend.IO.FEM.SG the story.DO proud.MASC.SG of.the buen hacer.
‘Juan used to tell each friend the story proud of the good deed.’
??Marta le cantó a cada amigoi el lied entusiasmada con sui
Marta 3.CL sang to each friend.IO.MASC.SG the lied.DO excited.FEM.SG at his
‘Marta sang the lied to each friend excited at his success.’
??Juan le contaba a cada amigai la historia orgulloso de sui
Juan 3.CL told to each friend.IO.FEM.SG the story.DO proud.MASC.SG of her buen hacer.
‘Juan used to tell each friend the story proud of her good deed.’
While (18a–b) without a bound pronoun are grammatical, (19a–b) with a pronoun bound by a quantifier are judged as degraded because of the quantifier binding. Let us assume that the indirect object has moved to a higher position than the SOP in (19a–b), escaping the VP domain (see Ordóñez 1998).18 Note that under the conventional structure (see footnote 16), the raised indirect object quantifier c-commands into the SOP, and it is predicted that a bound reading between them will be established. However, actually a bound reading between the quantifier and the pronoun in the SOP renders (19a–b) degraded. This binding fact is accounted for if we assume that the indirect object and the SOP are located in different planes. (19) can be represented as in (20).
Concretely, I assume that the indirect object, a cada amigo in (19a) and a cada amiga in (19b), has raised to a higher position than the postverbal subject in PrP1 , Marta in (19a) and Juan in (19b), before a single syntactic object PrP1+2 is formed, and that the raised indirect object still included in PrP2 cannot c-command the SOP located in PrP2.
If the discussion in this article is on the right track, we can conclude that a DSP and its subject are formed on a separate plane. This leads to a novel way of accounting for the derivation of depictive secondary predication.
In this article, I have argued for a multidimensional derivation for primary and secondary predication. Two syntactic objects that represent predication relations are formed separately, each on its own plane. These two syntactic objects are formed by means of a multidominance structure, taking the shared element as a grafting point. This proposal is supported by empirical data involving focus fronting and quantifier binding.
I would like to thank the two anonymous LI reviewers for valuable comments and insightful suggestions. This work was supported by the Hankuk University of Foreign Studies Research Fund.
2Bowers (2001) analyzes the predication relation of the DSP construction using PRO and control theory. However, I analyze it by establishing a direct relation between the DSP and its subject, without postulating PRO. Note that I do not deny the existence of an empty category PRO; I simply attempt to capture the predication relation directly.
4 No relation of derivational precedence holds between the primary predication and the secondary predication. In this sense, the terms primary and secondary are used here only for convenience; they do not refer to derivational order.
5 The assembling of PrP1 and PrP2 is based on the Single Syntactic Object Condition proposed by Chomsky (1995:226)
(i) At the LF interface, can be interpreted only if it consists of a single syntactic object.
7Sabel (2001:533) proposes the following Cluster Hypothesis for complex X0 and XP categories (multiple verb incorporation, multiple wh-constructions, multiple focus constructions, etc.):
(i) Cluster Hypothesis
A feature F that is attracted by K attracts a feature of the same type F.
When a functional head attracts α and β that have the same feature, they can move to the same position, forming a cluster.
8 The postulation of a functional head Asp in (11) is based on the aspectual correlation between the primary and secondary predicates.
Juan compró el coche contento.
Juan bought the car happy
‘Juan bought the car happy.’
*Juan compró el coche inteligente.
Juan bought the car intelligent
‘Juan bought the car intelligent.’
*Juan sabe inglés contento.
Juan knows English happy
‘Juan knows English happy.’
As shown in (i), the primary and secondary predicates should be compatible in terms of aspectual features. Moreover, postulating Asp provides a useful explanation for the DSP’s movement in (ii).
Juan comió cruda la carne.
Juan ate raw the meat
En Joan va menjar crua la carn.
Joan ate raw the meat
‘John ate the meat raw.’
9 The DSP should move to the edge of AspP since its base-generated position is invisible to the higher phase head C.
10 I assume that the edge of AspP is an Ā-position and that this position is related to subextraction from the DSP. Haegeman, Jiménez-Fernández, and Radford (2014:113) claim that, as shown in (i), subextraction out of a subject is possible when the subject occupies an Ā-position (i.e., Spec,Foc), while extraction out of a subject that remains in situ in Spec,v—that is, an A-position—is ungrammatical.
¿De qué pintori causaron literalmente un gran escándalo [varios dibujos ti]? of which painter caused literally a big scandal several paintings
*¿De qué pintori causaron un gran escándalo [varios dibujos ti] literalmente? of which painter caused a big scandal several paintings literally
‘Of which painter did several paintings literally cause a big scandal?’
11 V-to-T movement does not always imply phase sliding. For example, V-to-T movement does not extend the phase vP to TP in French. Concerning this matter, Alexiadou, Anagnostopoulou, and Wurmbrand (2014) propose that V-to-T movement causes phase sliding only when feature valuation takes place, which does not happen in French.
13 Following Sabel (2001), the focus cluster is formed prior to movement to Spec,C. I suggest that focus cluster formation takes place in Spec,Asp, as shown in (14). According to Chomsky’s (2001) Phase Impenetrability Condition, the base-generated position of the focused DSP in (14) is invisible to the higher phase head C. Thus, the edge of AspP functions as an intermediate landing site for the DSP.
14 For example, sentence (12b) has two focus elements, cua´ndo and las MANZANAS, derived in one plane as shown in (i).
I assume that the focus element las MANZANAS moves to the edge of PrP and becomes visible to the higher phase head C. When the probe C Agrees with the closest matching goal las MANZANAS, an Agree relation is not established with the other focus element, cuándo, owing to the Defective Intervention Constraint.
15 The question arises why multiple wh-elements are not available in the Spanish DSP construction. I have proposed that two focus elements derived in different planes form a focus cluster. In languages without multiple wh-fronting such as Spanish, wh-cluster formation is not permitted because one wh-element cannot be a host for the other wh-element (see Stoyanova 2008).
(i) *¿Qué regalo, qué tan emocionada, recibió Marta? what gift how much excited.FEM.SG received Marta
‘What gift, how excited, did Marta receive?’
17 If we replace the postverbal subjects of (17a–b) with preverbal subjects, the sentences become grammatical, as shown in (i).
ada pintori pintó a la modelo sentada en sui silla.
each painter.MASC painted A the model.FEM seated.FEM.SG on his chair
‘Each painter painted the model seated on his chair.’
Cada amigoi lanzó a la amiga vestida con sui abrigo.
each friend.MASC threw A the friend.FEM dressed.FEM.SG in his coat
‘Each friend threw the friend dressed in his coat.’ (Context: People are throwing others overboard to save them from a burning ship.)
In (ia–b), a bound reading is possible between the subject quantifier and the pronoun in the OOP since the preverbal subject moves out of PrP1+2 to a higher position, and this moved subject c-commands the pronoun in the OOP. López (2013:92) also points out an asymmetry between preverbal and postverbal subjects regarding bound readings with quantifiers. A quantifier within a preverbal subject can bind the pronoun within the dislocated object, while a quantifier within a postverbal subject cannot.
Cada padrei trajo un regalo a sui hijo.
each father brought a gift to his son
‘Each father brought his son a gift.’
Ayer, cada padrei le trajo un regalo, a sui hijo.
yesterday each father 3.CL brought a gift to his son
‘Yesterday, each father brought a gift, to his son.’
Ayer le trajo cada padrei un regalo, a su*i hijo.
yesterday 3.CL brought each father a gift to his son
‘Yesterday each father brought a gift, to his son.’
18Ordóñez (1998) assumes that the raised indirect object c-commands the postverbal subject, since, as shown in (i), a bound reading between the raised indirect object quantifier and the pronoun in the subject is available.
(i) ¿Quéle regaló a cada niñoi su amigoi para su cumpleaños?
what 3.CL gave to each boy.IO his friend for his birthday
‘What did his friend give to each boy for his birthday?’