AND A-CHAINS
ADINA DRAGOMIRESCU
1Abstract. After a brief presentation of the Romanian data regarding the lexical subject of non-finite verbal forms and the types of control allowed in Romanian, the author discusses supine configurations with controlled covert subjects and with lexical overt subjects. The lexical subject is very limited from a syntactic point of view: it can only appear in supine relative clauses and in tough-constructions with a passive embedded supine. Apparently, in these configurations the supine assigns Case. In Romanian control structures, the overt subject can be lexicalized in different slots in the main clause or in the embedded clause, a fact which raises problems for most of the current approaches to control. After reviewing the main theoretical proposals, the author concludes that the most appropriate theory for these problematic data is the one put forth by Alboiu (2007) for the Romanian subjunctive, i.e. the position of the lexical subject in control configurations is determined by pragmatic factors. The comparison between the subjunctive control configuration and the supine control configuration shows that these two constructions apparently follow the same rules. Although pragmatics is the key-notion for the choice of the slot of the lexical subject, we are far from understanding the exact relation between pragmatic effects and the position of the lexical subject.
Key-words: Supine, Control, Subject, Pragmatic Effects.
1. INTRODUCTION
In Romanian, there are four non-finite verbal forms: three of these are found in all the Romance languages (the infinitive, the gerund/present participle, and the past participle), and one is specific to Romanian (the supine). The Romanian supine has a controversial history: some researchers consider that this form is directly inherited from Latin (Bourciez 1946: 250, Diaconescu 1971: 151, Lombard 1974: 302, etc.), while others believe that it developed in Romanian (Caragiu- Marioţeanu 1962, Brâncuş 2007: 167, Vasiliu, Ionescu-Ruxăndoiu 1986: 196−198), out of the past participle, as a consequence of the loss of the verbal value of the
1 “Iorgu Iordan − Al. Rosetti” Institute of Linguistics and University of Bucharest, [email protected].
This paper is supported by the Sectorial Operational Programme Human Resources Development (SOP HRD), financed from the European Social Fund and by the Romanian Government under the contract number SOP HRD/89/1.5/S/59758.
RRL, LVI, 4, p. 371–392, Bucureşti, 2011
infinitive. In other studies (see Brâncuş 2007: 168 and the references therein), the supine is considered a Balkan Sprachbund feature.
All the Romanian non-finite forms − the infinitive (1), the gerund (2), the past participle (3), and the supine (4) − accept lexical subjects, which is obligatorily postverbal (Dobrovie Sorin 2000: 115). The lexicalization of the subject of the supine is rare and syntactically limited (see section 2.2.).
(1) dorinţa de a cânta Ion desire.DEF DE AINF sing.INF Ion.NOM
‘the desire for Ion to sing’
(2) Venind Ion, petrecerea s-a încheiat coming.GER Ion.NOM party.DEF CL.REFL.ACC.3SG=has ended
‘With Ion’s coming the party ended’
(3) Odată plecat Ion, a început petrecerea once left.PPLE Ion.NOM has started party.DEF
‘The party started after John left’
(4) masă de stat patru persoane table DESUP sit.SUP four persons.NOM
‘table for six people to sit at’
Romanian has obligatory control (OC) in finite structures with subjunctives (5a) and in non-finite structures with infinitives (5b), and non-obligatory control (NOC) with subjunctives (5c) and infinitives (5d). As Alboiu (2007: 193) has shown, aspectual and implicative verbs in Romanian have (exhaustive) OC, while desiderative verbs have NOC. The supine is selected only by the first class of verbs and, consequently, displays only OC (5e). Romanian does not exhibit partial control (Alboiu 2007: 193, Alexiadou et al. 2010).
(5) a. Încep [PRO să scriu]
(I)start SĂSUBJ write.SUBJ.1SG b. Încep [a scrie PRO]
(I)start AINF write.INF
‘I start writing’
c. Vrea [PRO să plece]
(he)wants SĂSUBJ leave.SUBJ.3SG d. Vrea [a pleca PRO]
(he)wants AINF leave.INF
‘He wants to leave’
e. Se apucă [de scris PRO]
CL.ACC.3SG starts DESUP write.SUP ‘He starts writing’
This paper is structured as follows: in section 2, we present the data regarding the subject of supine clause in Romanian, i.e., on the one hand, the controlled (covert) subject (section 2.1.) and, on the other hand, the lexical (overt) subject (section 2.2.). Section 3 deals with the analyses previously proposed for control phenomena, and with the possibility of following such analyses in order to explain the Romanian supine data. In section 4, we discuss the relevance of pragmatic factors for control phenomena in Romanian supine constructions, by contrasting them with the Romanian subjunctive. In section 5, we draw the conclusions.
2. THE DATA REGARDING THE SUBJECT OF THE SUPINE CLAUSE IN ROMANIAN
In this section, we present the relevant data regarding the controlled subject of the supine in Romanian, and the syntactic constraints on the presence of a lexical subject in the supine clause.
2.1. The controlled subject of the supine
In Romanian, there are many constructions that contain the supine (see Pană Dindelegan 2008: 512−522, in GALR I). The control of the embedded subject of the supine occurs in various supine constructions.
(i) The fully verbal supine is introduced by de, which has been considered either a supine marker (Pană Dindelegan 2003: 143) or as a complementizer and a mood marker (Cornilescu, Cosma 2011). The supine is selected by modal (6) or aspectual verbs (7) or by adjectives (8).
(6) Ioni are de scris PROi o carte Ion has DESUP write.SUP a book
‘Ion has a book to write’
(7) Ioni a terminat de scris PROi o carte Ion has finished DESUP write.SUP a book
‘Ion finished writing a book’
(8) a. Ioni este demn de admirat PROi (de către colegi) Ion is worthy DESUP admire.SUP by colleagues
‘John is worthy of admiration (from his colleagues)’
b. Eai este gata de plecat PROi la şcoală she is ready DESUP go.SUP to school
‘She is ready to go to school’
c. Eii sunt numai buni de făcut PROi bucătari they are just good.PL DESUP make.SUP cooks
‘They are suited enough to be made/to become cooks’
In examples (6)−(8) above, control is obligatory and exhaustive, the subject of the main clause being strictly identical to the subject of the embedded supine clause (Pană Dindelegan 1992: 132). If the supine is selected by an adjective, the mechanism of control functions in a different manner (Pană Dindelegan 2003:
147). In (8a), PRO is the controlled subject of a passive supine, while in (8b) PRO is the subject of an active supine of an unaccusative verb. The example in (8c) has two different interpretations: (i) the subject of the embedded supine clause is coreferential with the subject of the main clause (‘they become cooks’) or (ii) the direct object of the embedded supine clause is coreferential with the subject of the main clause (‘someone wants to make them cooks’).
If the main verb is impersonal (tough-constructions included), the subject of the embedded supine clause either has an arbitrary reading, without control (9), or it is controlled by another constituent of the main clause: the indirect object (10) or a genitive/possessive phrase (11) (Pană Dindelegan 1992: 132, 2003: 147). These cases also illustrate OC.
(9) a. Rămâne de terminat PROarb lucrarea (it)remains DESUP finish.SUP paper.DEF
‘The paper remains to be finished’
b. E greu de păstrat PROarb un prieten (it)is hard DESUP keep.SUP a friend
‘It is hard to keep a friend’
(10) a. Îmii rămâne de terminat PROi lucrarea CL.DAT.1SG (it)remains DESUP finish.SUP paper.DEF
‘It remains for me to finish the paper’
b. Îmii e greu de păstrat PROi un prieten CL.DAT.1SG (it)is hard DESUP keep.SUP a friend
‘It is hard for me to keep a friend’
(11) Este la îndemâna meai de făcut PROi asta (it)is at hand my.F.SG DESUP do.SUP this
‘It is at hand for me to do this’
In examples like (12), analyzed by Pană Dindelegan (2003: 146), control is non-obligatory.
(12) Mii se pare greu de acceptat PROi/j CL.DAT.1SG CL.REFL.ACC.3SG (it)seems hard DESUP accept.SUP propunerea
proposal.DEF
‘I think it is hard for me/you/us, etc. to accept the proposal’
Romanian also allows raising with impersonal verbs in tough-constructions;
the main verb agrees with the raised subject. This construction was analyzed as subject raising (Pană Dindelegan 1982), as predicate union (Hill 2002: 508) and as A-movement if the subject agrees with the main verb or Ā-movement if there is no agreement between the subject and the main verb (Dye 2006).
(13) a. Lucrările rămân de terminat lucrările (compare with (9a)) papers.DEF remain DESUP finish.SUP papers.DEF
‘The papers remain to be finished’
b. Prietenii sunt greu de păstrat prietenii (compare with (9b)) friends.DEF are hard DESUP keep.SUP friends.DEF
‘Friends are hard to keep’
In the GB analysis, control and raising display a few different properties: in raising constructions, the raised subject bears only a theta-role, while in control constructions the subject bears two theta-roles; raising predicates are formed via movement rules, while control structures are the result of the Equi NP deletion rule (in the standard theory) or of some construal rules linking a phonetically null DP (PRO) with its antecedent (in late GB analyses) − see Hornstein (1999: 70, 2003:
7−11). The movement analysis of control (in the spirit of Hornstein − see section 3.4. below) does not need this distinction anymore.
(ii) The supine is a mixed category, verbal and nominal, introduced by de or by other prepositions (pe ‘on’, la ‘at, in’, din ‘from’, spre ‘towards, for’, etc.). In the following examples, control is obligatory and exhaustive.
(14) a. Ioni se apucă de citit PROi Ion CL.REFL.ACC.3SG starts DESUP read.SUP b. Ioni se pune pe citit PROi Ion CL.REFL.ACC.3SG puts on read.SUP
‘Ion starts reading’
c. Ioni trăieşte din cerşit PROi
Ion lives from beg.SUP
‘Ion lives out of begging’
(iii) A special type of control occurs when a fully verbal supine, obligatorily introduced by de, is selected by a nominal (see Williams 1980). The subject of the supine is controlled by a genitive (15a) or by a possessive phrase (15b). Hornstein (2003: 48) claims that these constructions do not involve control, but are the result of the “aboutness relation” imposed by the genitive. In Romanian, this type of control (if this is indeed control) is rare, in most of the cases PRO having an arbitrary reading (16).
(15) a. pofta [lui Ion]i de mâncat PROi gogoşi desire.DEF Ion.GEN DESUP eat.SUP donuts
‘Ion’s desire to eat donuts’
b. cheful meui de muncit PROi dimineaţa disposition.DEF my.M.SG.3SG DESUP work.SUP morning
‘my disposition to work in the morning’
(16) a. pofta de mâncat PROarb gogoşi desire.DEF DESUP eat.SUP donuts
‘the desire to eat donuts’
b. cheful de muncit PROarb dimneaţa disposition.DEF DESUP work.SUP morning
‘the disposition to work in the morning’
A problematic situation. All these data can be accounted for in different ways in the current theories of control. But the possibility of different slots for the lexical subject raises problems for most of the analyses of control. Alboiu (2007) and Alexiadou et al. (2010) have taken into account the contexts in which the embedded subject of the subjunctive, coreferential with the subject of the main clause, is lexicalized in the subordinate clause. The subject of the embedded supine has the same possibilities (18).
(17) a. Ion se apucă să citească Ion.NOM CL.REFL.3SG starts SĂSUBJ read.SUBJ.3SG b. Se apucă Ion să citească
CL.REFL.3SG starts Ion.NOM SĂSUBJ read.SUBJ.3SG c. Se apucă să citească Ion CL.REFL.3SG starts SĂSUBJ read.SUBJ.3SG Ion.NOM ‘Ion starts reading’
(18) a. Ion se apucă de citit Ion.NOM CL.REFL.3SG starts DESUP read.SUP b. Se apucă Ion de citit
CL.REFL.3SG starts Ion.NOM DESUP read.SUP c. Se apucă de citit Ion CL.REFL.3SG starts DESUP read.SUP Ion.NOM
‘Ion starts reading’
In sections 3 and 4 we will look at different approaches to control in order to see if they can account for examples of this kind.
2.2. The lexical subject of the supine
In Romanian, the supine that takes a lexical subject is syntactically conditioned. There are only two supine constructions that take this type of subject
(see Pană Dindelegan 2011b). Neither of these structures involves a control configuration which would block Case assignment in the embedded domain (see section 3.6. above).
2.2.1. The first type of construction that accepts a lexical subject is represented by a fully verbal supine, in a supine relative clause (SRC), introduced by de. Old Romanian (19) and present-day Romanian (20) display the same restrictions.
(19) a. loc de cinat şase înş (Corbea, 232) place DESUP dine.SUP six people
‘a place where six people can dine’
b. scaun de şezut şase oameni (Corbea, 232) chair DESUP sit.SUP six people
‘a chair for six people to sit on’
c. loc de alergat caii (Corbea, 234) place DESUP run.SUP horses
‘a place where horses can run’
d. locuri de iernat oştile (Corbea, 237) places DESUP winter.SUP armies
‘places where the armies can winter’
e. avè şi vreme de vinit mojâcii (Neculce, 254) (it)had also time DESUP come.SUP churls
‘it was the time for the churls to came’
(20) a. măsuţe de jucat copiii (idealbebe.ro) tables DESUP play.SUP children
‘little tables for children to play at’
b. corturi de jucat copiii (idealbebe.ro) tents DESUP play.SUP children
‘tents for children to play in’
c. ham de mers copiii (www.okazii.ro) harness DESUP walk.SUP children
‘a harness for children to walk’
The supine relative clause was analyzed by Cornilescu, Cosma (2011), who identified two subtypes of SRC.
(i) The first type is represented by a SRC based on the relativization of the internal argument (21), in which the lexical subject is not allowed.
(21) prăjituri de servit musafirilor cakes DESUP serve.SUP guests.DAT
‘cakes to be served to the guests’
(ii) The second type of SRC is based on the relativization of a locative adjunct. This type accepts a lexical subject, as in (19) and (20), and has the following properties: de is a preposition which can be replaced by other prepositions (pentru ‘for’ − (22a)); relativization is strictly local (22b); very limited, the head can be resumed as a pronoun (22c).
(22) a. măsuţe pentru jucat copiii tables for play.SUP children
‘little tables for children to play at’
b. *măsuţe de încercat de jucat copiii tables DESUP try.SUP DESUP play.SUP children c. măsuţe de jucat copiii la ele
tables DESUP play.SUP children.DEF at them.F
‘little tables for children to play at’
2.2.2. The other supine construction with a lexical subject is the passive construction, where the argument of the supine can be considered the subject only in the presence of a by-phrase (23). If the by-phrase is not lexicalized and the supine has an active reading, the argument of the supine qualifies as a direct object (24).
(23) E greu de rezolvat problema de către toţi copiii (it)is hard DESUP work-out.SUP problem.DEF.F.SG by all children.DEF
‘~ It is hard for all the children to work out the problem’
(24) E greu de rezolvat problema
(it)is hard DESUP work-out.SUP problem.DEF.F.SG
‘It is hard to work out the problem’
2.2.3. The fact that the non-finite verbal forms (especially the infinitive and the gerund/present participle) can have their own subject has been explained in different ways. In older studies (Rosenbaum 1967, Lakoff 1968, 1971), the absence of the overt subject from infinitival clauses was explained by means of the rule of Equi-NP Deletion or Obligatory NP Deletion, which require that the subject of a subordinate clause be deleted at the surface level if it is identical with the subject of the main clause (Schulte 2007: 122).
The occurrence of the lexical subject was explained as the effect of the functional category Tense. Tense is also responsible for licensing pronominal clitics (Lois 1990: 253). According to Dobrovie-Sorin (2001: 49), the ability of infinitives to take lexical subjects is the effect of the fact that [–finite] Inflection can assign Case. The supine is special in that it can take a lexical subject, but it cannot host a clitic, contrasting with infinitives and gerunds/present participles, which not only take lexical subjects, but can also host clitics.
Dobrovie-Sorin (2000: 116) accounted for the postverbal position of the subject of non-finite verbal forms in Romanian by proposing that nominative Case is assigned in postverbal position, and not in Spec,IP. In clauses with agreement of the finite verb with the subject, there is an optional rule predicting that the subject has to undergo movement in a preverbal position. This rule is a parametrical choice, since there are languages such as Spanish (especially Caribbean Spanish) which allow for preverbal overt subjects with the infinitive (see Schulte 2007:
153).
3. APPROACHES TO CONTROL
The precise characterization of the control phenomenon and the analyses put forth for explaining it represent one of the most debated topics of present-day generative linguistics, especially because taking into account data from numerous languages generates difficulties for all the proposed theories.
3.1. Control in Government and Binding Theory (GB)
In GB (Chomsky 1986: 191−193), the non-overt subject (PRO) of non-finite forms is a pronominal element without a phonological matrix. PRO is similar to overt pronouns because it does not ever have an antecedent within its clause (or NP), and it resembles anaphors because it does not have intrinsic reference. PRO’s reference is either assigned by means of an anaphoric relation with an antecedent or it is arbitrary (PRO sometimes lacks specific reference, as examples (9) and (16) show). Therefore, PRO qualifies both as a pronoun and as an anaphor, these being elements that observe different binding principles: as an anaphor, PRO should be bound in its governing domain, while as a pronoun it should be free in its governing domain. This generates a contradiction. For eliminating this contradiction, Chomsky (1986: 191) formulates the following principle, which represents an essential property of PRO: PRO is ungoverned. Consequently, unlike all the other nominal expressions (phonologically realized or null), PRO does not receive Case at all.
The theory of the null element PRO was criticised precisely because it is unable to explain many data from various languages. Take Romanian, for instance:
there are various cases where PRO alternates with a lexical subject (compare, for example, (5a) to (25a) or (5e) to (25b)); this is unaccounted for in a GB approach (see Barrie, Pittman 2004: 78, Alboiu 2007: 193).
(25) a. Încep să scriu eu (I)start SĂSUBJ write.SUBJ.1SG I.NOM
‘I start writing’
b. Se apucă de scris Ion CL.ACC.3SG (hr)starts DESUP write.SUP Ion.NOM
‘Ion starts writing’
3.2. Control in Principles and Parameters Framework
Discussing the fact that PRO, unlike other nominal expressions, does not have Case, Chomsky and Lasnik (1995: 119) postulate that PRO has a special Case, namely null Case, which is different from the familiar Cases (nominative, accusative, etc.). Nevertheless, PRO has an exceptional status, being the only NP that can bear null Case. From an interpretative point of view, PRO is a minimal NP argument, lacking independent phonologic, referential or other properties. Null Case is somehow similar to nominative Case. According to Chomsky and Lasnik (1995: 120), nominative Case is standardly checked in Spec,IP, with I having tense and agreement features. This is thus an instance of the Spec−Head relation, the head being I. Similarly, null Case is an instance of the same relation, I lacking tense and agreement features: “the minimal I checks null Case, and the minimal NP alone [i.e. PRO] can bear it”. The authors also assume that, more generally, the infinitival element (with null agreement) and the head Ing of gerundive nominals check null Case. This can also be assumed in the case of Romanian for infinitival suffixes (-a, -ea, -e, -i, -î), gerund/present participle suffixes (-ind, -înd) or for supine suffixes, which are homonymous with the past participle suffixes (-at, -ut, -s, -t, -it, -ât). However, this theory does not account neither for the possibility of non-finite forms to take lexical subjects nor for the fact that there are different slots of lexicalizing the subjects.
Martin (2001) puts forward a refined theory for null Case, which he claims to be explanatory enough to account for the distribution of overt and non-overt subjects in infinitival clauses. His hypothesis is that the ability of non-finite T to check null Case depends on the temporal properties of T. Martin (2001: 146) shows that Chomsky and Lasnik’s analysis, more precisely the assumption that non-finite T always checks null Case, indirectly predicts that PRO can be the subject of any kind of infinitive, and that raising out of infinitives is never possible (which is contrary to the fact). Martin’s solution − which goes back to Stowell’s (1982) proposal that control infinitives are [+ tense], modal or future oriented, whereas raising infinitives are [− tense] − is that T in control infinitives checks null Case, whereas T in raising infinitives does not. However, this very interesting proposal does not hold for Romanian infinitives and supines: in Romanian, the raising untensed infinitive can take a lexical subject (26) and the untensed supine (i.e., the supine with an anaphoric tense − see Cornilescu, Cosma 2011) can take its own lexical subject in structures with (23) or without raising (4), (19), (20).
(26) Copiii par a fi bucuroşi/
children.DEF seem.IND.PRES.3PL AINF be.INF happy.PL Par a fi bucuroşi copiii
seem.IND.PRES.3PL AINF be.INF happy.PL children.DEF
‘The children seem happy’
As Landau (2007: 309) shows, PRO-based approaches cannot explain how PRO can be licensed and interpreted in a position higher than the one of the controller. This is validated by languages like English, where the shared argument in control constructions is forced to occupy a position in the main clause. As it is shown in examples (17) and (18), this is not the case for Romanian (see Alboiu 2007: 190).
3.3. Control as an Abstract Agree Relation
In Minimalism, there are two important trends in analyzing control: the one put forth by Landau, who claims that control is independent of movement, being an abstract agree relation, and the one proposed by Hornstein, who claims that control is an instance of movement.
In Landau’s (1999) approach, there are two types of OC: exhaustive control and partial control. In exhaustive control complements, Tense is null and PRO is referentially identical with the controller, while in partial control complements, Tense is contentful and PRO only has to include the controller. Thus, OC either enters an Agree relation between a matrix functional head and PRO (in exhaustive control), or undergoes infinitival Agree (in partial control). Partial control occurs only in tensed complements, where control via Agree blocks the transmission of semantic number from the controller to PRO. We will not go into details here, since Landau’s proposal does not offer any suggestion for the Romanian data (i.e., the possibility of lexicalizing the controlled subjects in different positions).
Gallego (this issue) is developing an analysis of control couched in the theory of phases that shares some ideas with Landau’s proposal, in particular, the fact that control involves (long-distance) Agree.
3.4. The Movement Theory of Control
Starting from O’Neil’s (1995) intuitions, Hornstein (1999) is the first to put forth an analysis where OC is conceived as movement/raising, i.e. OC consists in NP movement into a Case position, without the need of a PRO module (see Landau 2007: 293 and Gallego, this issue). Hornstein’s theory is based on two strong assumptions: Deep Structure has to be eliminated from Grammar, and theta-roles are features that trigger movement. The simplification goes one step further, in that
both the distinction between PRO and NP-trace and the one between control and raising are eliminated. As Hornstein (2003: 20) shows, PRO in OC constructions is identical with NP-trace, and it is the residue of an overt A-movement. In a raising construction, movement proceeds from the embedded clause to a matrix non-theta position, while in control structures, (one of) the landing site(s) of movement is obligatorily a matrix thematic position. If we assume that the subject positions of all non-finite clauses are not Case marking positions, A-movement from these positions is allowed. In contrast, A-movement from Case positions is prohibited;
thus, if PRO from OC constructions is the residue of A-movement, then we should never find a PRO in a Case position. Hornstein (2003: 22) also claims that control, like raising, is due to movement triggered by Case necessities, and that control is not necessarily restricted to non-finite subject positions: if some of the non-subject positions are not Case positions, they can be also occupied by PRO. This observation enables the subjunctive in OC constructions to have a PRO subject, because the subjunctive can also be Tense-deficient (Boeckx, Hornstein 2006:
123). PRO in OC constructions differs from PRO in NOC constructions: the first type resembles reflexives, while the later resembles pronouns (Hornstein 2003: 26).
For an appropriate analysis of the Romanian data (see section 3.6. below), we should keep in mind the fact that, as a result of movement via multiple theta- positions, we end up with a chain bearing multiple theta-roles, i.e., a chain saturating several distinct argument positions.
Boeckx, Hornstein (2006: 121) show that the movement theory of control explains the locality of control, more exactly, the fact that PRO occurs only in the highest subject position (in the embedded clause), and that the relation between the controller and PRO generally observes the Principle of Minimal Distance. But what remains problematic is not locality of control, but precisely non-locality of control, as illustrated by a large amount of data from Romanian. The debate between Hornstein and Landau has shown that Hornstein’s theory has other weak points that concern partial control or backward control. Hornstein (2003: 42−43, 52) shows that partial and backward control are not problematic for the movement theory of control. As to backward control, a point of interest for explaining the Romanian data, Hornstein claims that it represents a situation in which the controlled PRO asymmetrically c-commands its antecedent, and this situation can be accounted for in the movement theory of control. Nevertheless, situations like (17c) and (18c), in which the relation between PRO and its antecedent does not observe Minimal Distance, are not taken into account by Hornstein.
3.5. Backward Control
Backward control (BC) is a biclausal control configuration in which the lower coindexed subject is expressed and the thematic subject in the higher clause is unpronounced (Polinsky, Potsdam 2002: 261). This type of control was studied
by Farrell (1995) for Brazilian Portuguese, Polinsky, Potsdam (2002) for Tsez, a language from the Nakh-Daghestanian family (it is also present in two other languages from the same family, Tsaxur and Bezhta), Ordóñez (1999) and Gallego (this issue) for Spanish, Alexiadou et al. (2010) for Greek and Romanian subjunctives, etc. These genetically unrelated languages share a series of properties: they are pro-drop languages with a relatively free word order, and, as Alexiadou et al. noticed, they display a wide range of clitic doubling configurations.
The most influential theoretical approach is Polinsky, Potsdam (2002). These authors show that the Principles and Parameters approach cannot explain this kind of phenomena, but the movement theory of control allows BC, because it does not ban movement from a thematic position. BC occurs as a result of covert movement of the “controller” DP to its matrix thematic position. Landau (2007: 309) comments that the claim that BC exists in natural language is perhaps the most interesting contribution of the reductionist camp to the debate on the nature of OC.
Thus, if OC is A-movement, and A-movement can be covert, then the existence of BC is an unavoidable possibility.
Alexiadou et al. (2010) claim that in Greek and Romanian all OC verbs can also exhibit BC. The situation in Greek and Romanian differs from the one in Tsez in two important respects:
(i) in Tsez only two aspectual verbs display BC, while in Greek and Romanian all OC verbs allow BC;
(ii) Tsez has either obligatory forward control with most of the OC verbs, or obligatory BC with aspectuals, which means that in Greek and Romanian BC is optional while in Tsez BC is obligatory with aspectuals.
Alexiadou et al. (2010) have shown that BC in Greek and Romanian brings new evidence for the movement theory of control, precisely for the existence of a copy in the higher clause: in BC constructions, modifiers like ‘alone’ can be licensed in the matrix clause, while the DP they modify resides in the embedded clause:
(27) A învăţat singuri să-şi rezolve Ioni has learn alone SĂSUBJ=CL.REFL.DAT.3SG solve.SUBJ.3SG Ion.NOM problemele
problems.DEF
‘Ion learned to solve his problems all by himself’
Ordóñez (2009) analyzes Spanish examples like (28a, b), in which the subject of the infinitive is postverbal. From this perspective, Spanish differs from Catalan, where this type of post-infinitival subject is not available. Ordóñez’s conclusion is that Spanish differs from Greek and Romanian with respect to BC, and that Spanish does not have BC. Rather, these examples illustrate the formation of verbal complexes (where subjects are licensed by the matrix negation (28c,d), i.e. these
are matrix subjects and not embedded subjects, and thus the postverbal infinitival subject is not in situ, but illustrates an instance of movement) and should be re- examined from a remnant movement perspective.
(28) a. Antes de comprar (Luis) manzanas (Luis) (Spanish)
‘Before (Luiz) buy apples’
b. Hoy no deben (los estudiantes) leer (los estudiantes) las novellas (los estudiantes)
‘Today the students do not have to read the storyes’
c. No olvidó no tocarse nadie la nariz
‘Nobody forgot to touch his nose’
d. *?Olvidó no tocarse nadie la nariz
Another case of apparent BC in Spanish is discussed by Gallego (this issue, Gallego’s examples in (52)).
3.6. Control, A-Chains and pragmatic effects
It is generally accepted that the control theory goes beyond the purely syntactic domain, involving other factors such as theta-roles, properties of the verb, and some pragmatic aspects (Chomsky 1986: 76). In fact, starting from Jackendoff (1969, 1972), there is a standing tradition of proposals based on semantics. The main idea of these proposals is that the reference of the subject of the infinitival clause is dependent on the lexical semantics of the verb/the predicate that selects the respective infinitival complement. The semantic approaches were criticised because it was shown that OC is not a lexical property and thus cannot be reduced to the lexical properties of the verb (Jordan 2009: 145). The pragmatic approaches (Comrie 1984, 1985, Pountain 1998) have the same spirit: the meaning of the verb is involved in determining the type of control, and some control constructions are more likely to make an utterance felicitous than others. These models are not able to correctly indicate the subject of the infinitival clause. Schulte (2007) puts forth an integrated syntactic-pragmatic approach, a model which contains syntactic units, pragmatic criteria and Entrenchment Restrictions (or “fossilized pragmatics”, i.e.
the final step of the mechanism in which the control patterns established on syntactic and pragmatic bases are checked using some rules and restrictions, this eliminating the control patterns implausible in most of the discourse contexts).
Schulte’s model is very interesting, but (again) it does not say anything about the possibility of multiple slots for lexical subjects in Romanian.
An appropriate explanation for the Romanian data is offered by Alboiu (2007). The author puts forth an analysis for Romanian OC subjunctives which follows Hornstein’s main ideas, but which contains a few amendments in order to explain the optional subject displacement, and semantic and pragmatic effects
associated therewith. After presenting Alboiu’s proposal for subjunctives, we will check if it holds for Romanian supine as well (section 4.).
Alboiu (2007: 195) shows that Romanian OC subjunctives are non-phasal, which is proven by the following facts: (i) a lexical complementizer is absent, (ii) the T head is Tense-unsaturated (and even phi-incomplete), and (iii) the embedded domain is not able to value the Case of the DP subject. All these are properties of C, a phase head, and not of T. The absence of these properties unambiguously shows that OC subjunctives are not phases. We should note that the same argumentation holds for Romanian supine OC constructions (de is not a lexical complementizer, T is unsaturated − see 3.2. above − and the embedded domain is not able to value the Case of the DP subject). Romanian subjunctives can be phasal, but only in the presence of the lexical complementizer ca, which prevents DP-movement out of the embedded clause. In conclusion, OC subjunctive structures and OC supine structures are not phasal domains in Romanian. This is very important, because in recent minimalist work Case valuation is a property of phasal domains rather than of Agreement, and, consequently, the two OC structures cannot value Case and cannot licence a DP subject. Because OC subjunctives cannot satisfy the Case necessities of the embedded DP, this DP is forced to move elsewhere. The creation of A-Chains is not equivalent to movement but to the instantiation of an Agree operation. In the derivation of OC subjunctive structures (largely explained by Alboiu 2007: 203−205), the shared DP subject enters at least two A-Chains: a thematic chain and a Case chain.
As Alboiu (2007: 203) shows, the position of the DP subject is a semantic and pragmatic effect, independent of the syntactic satisfying of OC. The author starts from the assumption that Romanian exploits syntactic structure to encode sentence pragmatics: phrases may be dislocated in order to obtain interpretative effects (such as Theme, Rheme), i.e. technically speaking, an occurrence feature (OCC) is optionally present in the derivation.
Alboiu (2007: 205−207) identifies five situations in which the lexical DP subject occupies different slots, depending on the desired pragmatic effect.
(i) The shared argument remains in situ (i.e. in the Spec,vP of the subjunctive) if the embedded VP is interpreted as new information, and there is no OCC feature present in the derivation. The fact that Victor is in an A-position is proven by the possibility of being replaced by the quantifier cineva ‘someone’ (see Alboiu 2007: 211, endnotes 23 and 24, for details).
(29) − Ce e gălăgia asta?
what is noise.DEF this
‘What’s all this noise?’
− Încearcă să cânte Victor la trombon tries SĂSUBJ sing.SUBJ.3sg Victor.NOM at trombone
‘Victor is trying to play the trombone’
(ii) The shared argument is a part of the rhematic domain of the matrix clause, and in this case the embedded subject is displaced to the matrix Spec,vP.
This displacement is due to the presence of an OCC feature on the higher v predicate, which requires that the shared DP surface next to higher v and not next to the embedded v. In (30) too, Victor can be replaced by cineva ‘someone’.
(30) − Ce se întâmplă?
what CL.REFL.ACC.3SG (it)happens
‘What’s going on?’
− Încearcă Victor să cânte la trombon tries Victor.NOM SĂSUBJ sing.SUBJ.3SG at trombone
‘Victor is trying to play the trombone’
(iii) If the shared argument consists of exclusively new information (it is the rhematic focus of the sentence), it will surface as maximally embedded in the subjunctive predicate. In (31) too, Victor can be replaced by cineva ‘someone’.
(31) − Cine încearcă să cânte la trombon?
Who tries SĂSUBJ sing.SUBJ.3SG at trombon
‘Who is trying to play the trombone?’
− Încearcă să cânte la trombon Victor
tries SĂSUBJ sing.SUBJ.3SG at trombone Victor.NOM
‘Victor is trying to play the trombone’
(iv) If the shared subject is known to both the speaker and the hearer, it is interpreted as a Topic. If the subject is lexicalized, it surfaces in the matrix sentence preverbal domain, outside of the main clause predicate Rheme.
Displacement happens due to an OCC requirement on matrix C/T domain. In (32), Victor cannot be replaced by cineva ‘someone’, and this indicates that it is in a Topic Ā-position.
(32) − Mihai, ce face Victor?
Mihai what does Victor
‘Mihai, what’s Victor doing?’
−(Victor) încearcă să cânte la trombon Victor tries SĂSUBJ sing.SUBJ.3SG at trombone
‘Victor is trying to play the trombone’
(v) The shared argument can be contrastively focused. In these cases, dislocation is not crucial, the only requirement being heavy prosodic stress.
(33) − Mihai încercă să cânte la trombon?
Mihai tries SĂSUBJ sing.SUBJ.3SG at trobone
‘Is Mihai trying to play the trombone?’
− (VICTOR) încearcă (VICTOR) să cânte (VICTOR) la Victor.NOM tries Victor.NOM SĂSUBJ sing.SUBJ.3SG Victor.NOM at trombon (VICTOR)
trobone Victor.NOM
‘It’s Victor who’s trying to play the trombone (not Mihai)’
Alboiu’s (2007: 208) conclusion is that the shared argument of Romanian OC constructions only “moves forward” to ensure novel semantic and pragmatic effects. In the next section we will see if these pragmatic effects are recognized by all speakers and if OC with supines follows the same rules as OC with subjunctives.
4. WHAT PRAGMATICS CAN EXPLAIN
In this section, we present the results of a linguistic experiment in which we tested 15 speakers with linguistics training. With this experiment, we checked the pragmatic effects described by Alboiu (2007) for the lexicalization of the controlled subject of the subjunctive and we compared the controlled subjunctive with the controlled supine. The questionnaire contains a brief presentation of the communicative context (“you hear music; Victor has been playing the trombone for two minutes”); the subjects are asked to answer 5 questions, using for variant (a) the subjunctive and for variant (b) the supine. Because Romanian is a pro-drop language, the subjects are asked to use – if possible – the lexical subject Victor. The five questions are those suggested by Alboiu (2007), illustrated above in (29)–(33).
In the case of question (1) (Ce e gălăgia asta? ‘What’s all this noise?’), according to the prediction above in (29), the subject should be placed after the embedded verb, which is contrary to the results. None of the 15 answers contains the subject only in the envisaged position (i.e., there was no answer of the type S-a apucat să cânte/de cântat Victor la trombon). In most of the answers (7 answers), the subject is placed immediately after the main verb (S-a apucat Victor să cânte/de cântat la trombon); one of the answers has, for the embedded subjunctive, the variant in which the subject follows the embedded verb as well (S-a apucat să cânte Victor la trombon). In 5 of the answers, the subject occupies the initial position of the main clause (Victor s-a apucat să cânte/de cântat la trombon), but 2 of the answers contain only the subjunctive (and not the supine) in the embedded clause. Two of the answers contain three variants for the position of the subject ((Victor) s-a apucat (Victor) să cânte/de cântat (Victor) la trombon), and one answer contains all four variants for the position of the subject (the preceding three ones plus the embedded clause final position: S-a apucat să cânte/de cântat la trombon Victor), with the speaker’s comment that intonation matters in all the cases.
For question (2) (Ce se întâmplă? ‘What’s going on?’), the prediction is that the subject is placed after the main verb (as in (30)). Only in 4 of the answers this prediction is borne out (S-a apucat Victor să cânte la trombon); one of the respondents also accepts for the supine the variant with the subject placed after the embedded verb (S-a apucat de cântat Victor la trombon). In other 4 answers, the subject occurs in the initial position of the main clause (Victor s-a apucat să cânte/de cântat la trombon); one of the answers contains only the subjunctive variant (the supine one is excluded). Three of the answers contain both word order possibilities above, and, of these, one contains only the subjunctive. One of the answers contains three word order possibilities ((Victor) s-a apucat (Victor) să cânte/de cântat (Victor) la trombon), and two answers contain four possibilities (the preceding three ones plus S-a apucat să cânte/de cântat la trombon Victor).
For question (3) (Cine s-a apucat să cânte/de cântat la trombon? ‘Who started playing the trombone?’), according to the rule in (31), the lexical subject should have the final position in the embedded clause. However, in most of the answers (11 answers), the subject is in the initial position of the main clause (Victor s-a apucat să cânte/de cântat la trombon); one of the answers contains only the subjunctive and another one only the supine. Two answers contain only the subject (Victor). One of the answers contains two variants: the subject is in the initial position of the main clause and in the final position of the embedded clause ((Victor) s-a apucat să cânte/de cântat la trombon (Victor)). One of the answers contains three variants: Victor; Victor s-a apucat să cânte/de cântat la trombon and S-a apucat Victor.
For question (4) (Mihai, ce face Victor? ‘Mihai, what’s Victor doing?’), the prediction from example (32) is that in the answer the subject should occupy the first position of the main clause. In this case, the majority of answers (9 answers) confirm the prediction; however, one contains only the subjunctive, and one contains only the supine. In one of the answers, the subject occupies the final position of the embedded clause. In two cases, there are two variants, with the subject in sentence-initial position and in the final position of the embedded clause ((Victor) s-a apucat să cânte/de cântat la trombon (Victor)). One of the answers contains all four word order possibilities, and two answers contain only the subject, without the rest of the utterance.
Finally, the answer for question (5) (Mihai încercă să cânte la trombon? ‘Is Mihai trying to play the trombone?’) should allow the subject to occupy any position, including the possibility of not uttering the subject (after the negation). In most of the answers (13 answers), the subject occupies the initial position of the main clause; in six cases, the subject is preceded by the sentence negation (Nu, Victor s-a apucat să cânte/de cântat la trombon). One of the answers contains two variants: in one case, the subject occupies the initial position on the main clause, in
the other, it occurs in sentence final position ((Victor) s-a apucat să cânte/de cântat la trombon (Victor)). One of the answers is as follows: Aş! Victor (‘Nope, Victor’).
The analysis of these answers yields three types of conclusions:
(i) The rules postulated by Alboiu (2007) for the pragmatic effects of the lexicalization of the subject in A-Chains in structures with the subjunctive are not confirmed by the received answers. However, what is confirmed is that the position of the lexical subject in these A-Chains is pragmatically relevant; very often, the discourse/pragmatic effects are not determined only by word order, but also by intonation.
(ii) In many cases, irrespective of the information structure targeted by the question, in the answer the subject occupies the sentence (main clause) initial position or, however, there is a preference for lexicalizing the subject in the main clause.
(iii) Most of the answers show that OC structures with the subjunctive and with the supine display the same pragmatic effects; most of the speakers have chosen the same word order (i.e. position of the subject) with both the subjunctive and the supine; only one respondent preferred the subjunctive (to the supine) for some of the answers and has commented on the aspectual differences between the subjunctive and the supine.
5. CONCLUSIONS
In this paper, we have presented and commented on the two types of subjects allowed in supine constructions: (i) the lexical, overt subject, obligatorily postverbal, limited to SRC constructions and to tough-construction with a passive embedded supine, and (ii) the covert subject in OC supine configurations. In these two constructions, the supine has different properties: in the cases where it can take a lexical subject, the supine is able to assign nominative Case due to the presence of the functional category responsible for nominative assignment (Tense or v). But it is not clear why in these constructions (as in any supine construction) clitics are not allowed, since the general assumption is that the subject and the clitics are licensed by the same functional category (Tense or v). This might provide a suggestion as to the clause structure of Romanian: it might be that, in Romanian, the subject and the clitics are not tied to the same functional category.
Romanian data are an argument against the theory of PRO, since the referential expression to which PRO is related is not obligatorily in a higher position of the clause. Data of this kind might be accounted for in the BC approach (like the one put forth by Alexiadou et al. for Romanian). However, it is better understood in a theory which postulates that control involves an A-Chain in which
any copy of the subject can be lexicalized, depending on some pragmatic interpretation (Alboiu 2007). Following Alboiu’s analysis for the subjunctive, we have shown that the same analysis holds for Romanian supine configurations.
However, the parallelism between a certain position of the lexical subject and a precise pragmatic effect rests beyond any precise rule. It seems that speakers prefer to place the subject in clause initial position independently of any pragmatic interpretation (which, by the way, is well supplemented by intonation).
CORPUS
Teodor Corbea, Dictiones latinae cum valachica interpretatione, edited by Alin-Mihai Gherman, vol.
I, Cluj-Napoca, Clusium, 2001 [1691−1697].
Ion Neculce, Letopiseţul Ţării Moldovei şi O samă de cuvinte, edited by Iorgu Iordan, Bucureşti, Editura de Stat pentru Literatură şi Artă, 1955.
Internet
REFERENCES
Alboiu, G., 2007, “Moving forward with Romanian backward control and raising”, in: W. D. Davies, S. Dubinsky (eds.), New horizons in the analysis of control and raising, Dordrecht, Springer, 187−211.
Alexiadou, A., E. Anagnostopoulou, G. Iordăchioaia, M. Marchis, 2010, “No objections to Backward Control?”, in Norbert Hornstein, Maria Polinsky (eds.), Movement Theory of Control, Amsterdam, John Benjamins, 89–118, http://ifla.uni-stuttgart.de/files/bc_paper.pdf
Barrie, M., C. M. Pittman, 2004, “Partial control and the movement towards Movement”, Toronto Working Papers in Linguistics, 22, p. 75−92.
Boeckx, C., N. Hornstein, 2006, “The virtues of control as movement”, Syntax, 9, 2, 118−130.
Bourciez, É., 1946, Éléments de linguistique romane, quatrième edition, Paris, Librairie C. Klincksieck.
Brâncuş, G., 2007 [1967], “O concordanţă gramaticală româno-albaneză: modul supin”, in Studii de istorie a limbii române, Bucureşti, Editura Academiei Române, 167−173.
Caragiu-Marioţeanu, M., 1962, “Moduri nepersonale”, Studii şi cercetări lingvistice, 1, 29−43.
Chomsky, N., 1986 [1981], Lectures on Government and Binding. The Pisa Lectures, fourth edition, Dordrecht, Riverton, Foris Publications.
Chomsky, N., H. Lasnik, 1995 [1993], The Theory of Principles and Parameters, in: N. Chomsky, The Minimalist Program, Cambridge, Massachusetts, London, England, The MIT Press, 13−127.
Comrie, B., 1984, “Subject and Object Control: Syntax, Semantics, Pragmatics”, in: C. Brugmann, M.
Macaulay (eds.), Proceedimgs of the Tenth Annual Meeting of the Berkley Linguistics Society, 450−464.
Comrie, B., 1985, “Reflections on Subject and Object Control”, Journal of Semantics, 4, 47−65.
Cornilescu, A., R. Cosma, 2011, “The structure of supine relative clauses”, The Annual Conference of the English Department, juin 3-4.
Diaconescu, I., 1971, “Supinul în limba română din secolele al XVI-lea − al XVIII-lea”, Analele Universităţii din Bucureşti. Limba şi literature română, 1−2, 151−163.
Dobrovie Sorin, C., 2000 [1994], Sintaxa limbii române. Studii de sintaxă comparată a limbilor romanice, Bucureşti, Editura Univers.
Dobrovie Sorin, C., 2001, “Head-to-Head Merge in Balkan Subjunctives and Locality”, in: M. L.
Riverso, A. Ralli (eds.), Comparative Syntax of Balkan Languages, Oxford, Oxford University Press, 44−73.
DSL − A. Bidu Vrănceanu, C. Călăraşu, L. Ionescu-Ruxăndoiu, M. Mancaş, G. Pană Dindelegan, 2005, Dicţionar de ştiinţe ale limbii, Bucureşti, Editura Nemira.
Dye, C., 2006, “A and Ā-Movement in Romanian Supine Constructions”, Linguistic Inquiry, 4, 37, 665−674.
Farrell, P., 1995, “Backward Control in Brazilian Portuguese”, in: J. Fuller, H. Han, D. Parkinson (eds.), Proceedings of Eastern States Conference in Linguistics, 94, 116−127.
Hill, V., 2002, “The gray area of supine clauses”, Linguistics, 40, 3, 495−517.
Hornstein, N., 1999, “Movement and Control”, Linguistic Inquiry, 30, 1, 69−96.
Hornstein, N., 2003, “On Control”, în: Randall Hendrick (ed.), Minimalist Syntax, Malden-Oxford- Carlton, Blackwell Publishing, 6−81.
Jackendoff, R., 1969, Some Rules of Semantic Interpretation for English, Cambridge, Mass., MIT Dissertation.
Jackendoff, R., 1972, Semantic Interpretation in Generative Grammar, Cambridge, Mass., MIT Press.
Jordan, M., 2009, Loss of Infinitival Complementation in Romanian Diachronic Syntax, PhD dissertation, Florida.
Lakoff, R. T., 1968, Abstract Syntax and Latin Complementation, Cambridge, Mass., MIT Press.
Lakoff, R. T., 1971, Irregularity in Syntax, New York: Holt, Rinehart & Winston.
Landau, I., 1999, Elements of Control, PhD thesis, MIT.
Landau, I., 2007, “Movement-resistant aspects of control”, în William D. Davies, Stanley Dubinsky, New horisons in the analysis of control and raising, Dordrecht, Springer, 293−325.
Lois, X., 1990, “Auxiliary selection and past participle agreement in Romance”, Probus, 2, 233−255.
Lombard, A., 1974, La langue roumaine. Une presentation, Paris, Éditions Klincksieck.
Martin, R., 2001, “Null Case and the Distribution of PRO”, Linguistic Inquiry, 32, 1, 141−166.
O’Neill, J., 1995, “Out of Control”, in: J. N. Beckman (ed.), Proceedings of NELS, 25, 361−371.
Ordóñez, F., 2009, “Backwards or in situ? Infinitival Overt Subjects in Romance”, 19th Colloquium on Generative Grammar, 1 april.
Pană Dindelegan, G., 1982, “Structura sintactică nominal + adverb (sau adjectiv) + supin”, Limba română, 1, 5−13.
Pană Dindelegan, G., 1992, “Supinul”, in Sintaxă şi semantică. Clase de cuvinte şi forme gramaticale cu dublă natură (adjectivul, adverbul, prepoziţia, formele verbale nepersonale), Bucureşti, Tipografia Universităţii din Bucureşti, 128−136.
Pană Dindelegan, G., 2003, “Precizări asupra determinanţilor supinului”, in Elemente de gramatică.
Dificultăţi, controverse, noi interpretări, Bucureşti, Humanitas Educaţional, 142−150.
Pană Dindelegan, G., 2008, “Supinul”, in: V. Guţu Romalo (ed.), Gramatica limbii române, vol. I Cuvântul, Bucureşti, Editura Academiei Române, 509−524.
Pană Dindelegan, G., 2011a, “Din istoria supinului românesc”, in: R. Zafiu, C. Uşurelu, H. Bogdan Oprea, (eds.), Limba română – ipostaze ale variaţiei lingistice, Bucureşti, Editura Universităţii din Bucureşti, 119–130.
Pană Dindelegan, G., 2011b, “The supine”, in: G. Pană Dindelegan (ed.), Romanian Grammar. A linguistic introduction, Oxford, Oxford University Press (to appear).
Polinsky, M., E. Potsdam, 2002, “Backward Control”, Linguistic Inquiry, 33, 2, 245−282.
Pountain, C. J., 1998, “Person and Voice in the Spanish Infinitive”, Bulletin of Hispanic Studies, 75, 393−410.
Rosenbaum, P., 1967, The Grammar of English Predicate Complement Construction, Cambridge, Mass., MIT Press.
Stowell, T., 1982, “The Tense of Infinitives”, Linguistic Inquiry, 13, 561−570.
Vasiliu, E., L. Ionescu-Ruxăndoiu, 1986, Limba română în secolele al XII-lea − al XV-lea (fonetică − fonologie − morfologie), Bucureşti.
Williams, E., 1980, “Predication”, Linguistic Inquiry, 11, 203−238.
