According to the GB assumptions all sentences must have a subject projection (EPP). EPP guarantees that a subject position will always be available even though it may not be relevant to a predicate’s argument structure. In these cases expletives or dummy’ subjects occupy the empty-generated subject position as in (60b) below. Their presence in the sentence has no bearing on semantics. However, as it is well known no similar dummy subjects appear in Greek as shown in (60a) below:
(60) (a) Vrechi. rains-3s
It is raining.’ (b) It is raining
Thus, while the English (60b) requires a dummy subject in the absence o f a proper one, the Greek (60a) does not require this. According to Speas (1990, ch.3 cited in Horrocks 1994: 84) absence o f pleonastics is a property associated with non-configurational languages for reasons described above. So too, according to Speas, related to this point is the fact that such languages do not also display subject raising, a problem which I shall
return to later on.
2.5.3.7.1 Lack of PRO
F o llo w in g the E P P outlined above, all clauses m ust c ontain a subject position even though this m ay not be filled by an ov ert subject. In this case, an an a p h o ric pronom inal e m p ty c a te g o ry is postu lated , called P R O , o r hi ^ P R O . T h e main evid en c e fo r P R O c o m e s from so-called control data. As an exam ple, c o n sid e r (6 1 ) below:
(6 1 ) J o h n p e rsu a d e d M ary to c o m e to the party.
In (61) above the understood subject o f the ifinite clause is the sam e as the object o f the matrix verb ‘p ersuade’. In this case, it is said that ‘M a r y ’ is the controller o f the subject o f the subordinate clause. A c c o rd in g to B resn a n (1982: 372), co n tro l depicts a relation o f referential dependency betw een an unexpressed subject, called the c o ntrolled elem ent, and an ex p re sse d controller. It is fu rth e rm o re a rg u e d that P R O is in c o m p le m e n ta iy distribution with overt NPs, the reason being that nominals need to be c a se -m a r k e d u n d er government, while P R O is only licensed in u n g o v e r n e d e n v iro n m e n ts (h e n c e n o n -C a se - m arked). This is the so-called PRO theorem, w h ich basically entails that P R O cannot o c c u r as the subject o f a finite clause ( C h o m s k y 1981, Manzini 1983).
In relation to G reek now , th e re are tw o ty p es o f analyses co n c e rn in g P R O O n e assu m e s the standard m ain strea m G B a p p r o a c h o utlined above, the p ro b lem s o f which h a v e been p ointed o u t by Philippaki and Catsimali (1 9 9 5 ) T h e o th e r claims that P R O d o e s n o t ex ist at all in G re e k , since G re e k has no infinitives, and thus P R O c a n n o t be licensed (Philippaki 1987, Philippaki and Catsimali 1995). This a p p r o a c h fu rth e rm o re advocates that control in G re e k can be easily a c c c o u n te d for by re fere n ce to sem an tico - p r a g m a tic inform ation. A lth o u g h a detailed desc rip tio n o f th ese tw o ty p es o f analyses goes beyond the scope o f my discussion, I shall briefly refer to the main a r g u m e n ts o f this second analysis, for it incorporates so m e useful em pirical and theoretical insights, which are relevant to the overall arg u m e n t o f this part. C o n s id e r first the d a ta in (62):
(62) (a) O Petros theli na pai sinema. [the Peter]nom want-3 s to go-3s cinema ‘Peter wants to go to the cinema.’ (b) Peter; wants [PRO; to go to the cinema]
In (62a) above, the embedded verb agrees in number and person with the matrix subject. Thus, it might be argued that the understood subject is a PRO coreferential with the matrix subject, as in the corresponding English example in (62b). However, unlike English, the embedded verb may vary in number, person, as illustrated in (63) below:
(63) (a) O Petros theli na pas sinema.
[the Peterjnom want-3s to go-2s cinema ‘Peter wants you to go to the cinema ’
(b) * Peter, wants [PROj to go to the cinema]
Na-clauses are finite clauses, introduced by ‘na’, an inflectional element showing subjunctive modality (Philippaki and Veloudis 1985)^”. If PRO was licensed in the embedded clause in (63a), then it would be governed by the finite inflection, which should presumably result in ungrammaticality, contrary to fact. According to Philippaki and Catsimali, therefore, PRO is not licensed in Greek. Instead, what is licensed inside the embedded clause is a small pro, which, however, is o f a sublexical affix inside the VP, carrying features o f number, person, tense (Philippaki 1987). I return to the details o f this in chapter 3.
According to this view, then, the coreferential or non-coreferential interpretation between the embedded and matrix subject in (62) and (63) above is semantically licensed. This kind o f analysis is also advocated by Joseph (1992), and Philippaki and Catsimali (1995) offer rather convincing data in support o f their view that it is the pronominal subject suffix, rather than PRO, which controls the nominative case o f the embedded nominal, as shown by the data in (64):
(6 4 ) (a) E vala to n K o sta na majirepsi m o n o s tu
p u t - I s [the K o s ta s ja c c to c o o k by him self-nom ‘I m a d e C o s ta s to c o o k by him self,’
(b) E pisa to n P e tro na ekfonisi to lo g o afto s o idjos p e r su a d e -Is [the Peterjacc to give [the talk Jacc [this the sa m e jn o m ‘I p e r s u a d e d P e te r to give the talk h i m s e l f ’
T he point o f the data s h o w n in (6 4 ) ab o v e is that the n o m inative ca se o f the e m b e d d e d clause can only be a c c o u n te d for by c o in d ex a tio n with the built-in subject sufilx o f the e m b e d d e d verb.
2.S.3.9 Some more argum ents
This last part brings to atten tio n so m e additional areas in w hich I think the ca se for the structural difference b e tw e e n G re ek and English c l a u s e can be reinforced T h e s e mainly c o n c e rn nominal raising and passivization
2.5.3.9.1 NP-raising ?
Recall from the discussion in section 2 .5 .3 .7 that raising is pre d ic ted to be absent in languages w here the subject d o e s not enjoy a special syntactic status. C o n s id e r th e N P - raising d a ta from English in (6 5 ) and the equivalent G re e k ex a m p le s d epicte d in (66):
(6 5 ) (a) *It seems[„, t the children to be intelligent]
(b) It seem s [(.,, [(, that [„. the children are intelligent ]]] (c) [T h e children; seem [„. f to be intelligent]]
(6 6 ) (a) F e n o n te na ine eksipna ta pedhia.
seem -3pl to be in telligent-nom -neut [the ch ild re n jn o m (b) T a pedhia fe nonte -3pl na ine eksipna.
According to a G B analysis (6 5 a ) is un g ra m m a tica l b e c a u se th e subject o f the infinitival clause, ‘the child’ is not assigned Case. N o te that the [-finite] inflection in (6 5 a ) cannot
qualify as a g o v e r n o r and, c onse quently, as a C a s e -a ssig n e r to the relevant NP. If, however, the N P is raised to the e m pty subject p osition o f the m atrix v erb ‘s e e m s ’, then t h e N P will be C a s e - m a rk e d by the [+finite] g o v e r n o r I o f the m atrix verb, and the s e n t e n c e b e c o m e s gra m m a tica l (65c). T h e fact that the positio n b e fo re ‘s e e m s ’ is g e n e r a t e d e m p ty is p ro v e d by (6 5 b ) in w hich the expletive ‘it’ o c c u p ie s the subject position. ‘S eem s’ is called a raising’ verb since its inherent p ro p e r ty is alw ays to trigger raising o f the N P o c c u p y in g th e subject position o f its infinitival c o m plem ent.
C o n s id e r in g n o w (6 6 ) w e see that b o th the ra is e d ’ and the ‘n o n - r a is e d ’ construction are equally gra m m a tica l in G re e k and, th erefore, no structura l re q u irem en t similar to that obtaining in English is at w ork here N ote that in the G B analysis o f raising d iscussed here, (6 6 ) w o u ld be ta k e n as an exa m p le o f raising (A g o u r a k i 1993). O n e o f th e arg u m e n ts s u p p o rtin g this putativ e ‘raising’ analysis is that the n u m b er and person o f the ‘ra is e d ’ N P is re g istered in th e inflection o f the ‘ra isin g ’ v erb ‘f e n e te ’ N ote, h o w e v e r , that this a rg u m e n t is not conclusive, for as s h o w n in (6 7 ) below , ‘fe n e te ’ (se em ) could also be left u n ch a n g e d in te rm s o f num ber, irre sp ectiv e o f the n u m b e r o f the ‘ra is e d ’ n oun
(6 7 ) (a) F e n e te na ine eksipna ta pedhia. se e m -3 s to be in telligent-nom -neut [the c h ild re n jn o m (b) T a ped h ia fe nete-3s na ine eksipna.
In this case, ‘fenete’ is the impersonal 3rd p e rso n singular verb w h o s e subject, ac c o rd in g to the traditional analysis, is the whole lo w e r na-clause. M o re o v e r, since G re ek d o e s not m ake use o f pleonastic elements, the diagnostic for an e m p ty g e n e ra te d n o n -th eta . C ase- assigned p o sition p re c e d in g th e raising verbs ca n n o t be e m p lo y ed either.
NP-raising thus appears to be totally optional in G re ek , and totally un re la ted to
^'Furthennore, Agouraki (1993) points out that the (îreek verb ‘Icncte' meets other diagnostic criteria for raising verbs quoted below (ibid: 133):
(i) no selectional restnctions imposed by ‘fenete’ on the lower subject (ii) ‘fenete’ lacks an object
(iii) ‘fenete’ does not assign an external theta-rolc (iv) ‘fenete’ is not similar to the English ‘look like’
a ny g ra m m a tica l co n d itio n s on N P s such as C a s e assignm ent, A g o u r a k i ’s point (Agouraki 1993; 1 3 1) quoted here is, I think, indicative o f this; “ . . w i t h re spect to o th e r la n g u a g e s the possibility o f raising from +fm clau se s has been n o te d by G ro su and H orvath (1984) and R iv ero (1992). [. ]T h e fact th a t th e re is raising from +fin clauses is p r o b le m a tic for all cu rren t th eo rie s o f m o v e m e n t T h e q u e stio n o f the m otivation o f raising seem s to be equally p ro b le m a tic ” ^ ’
In sum , I a r g u ed that o b ligatory N P -ra isin g d o e s not obtain in G reek which naturally follows from the fact that N P subjects d o n o t enjoy a special syntactic status in M G This situation parallels exactly that o f passivization to which I shall turn belo w
2.5.3.9.2 Passivization
Further evidence against the ex isten c e o f a c o m p u ls o ry subject position in G reek c o m e s from facts relating to passivization C o n sid e r (6 8 ) and (6 9 ) below;
(6 8 ) (a) *It is c o n sid e re d Jo h n to be intelligent (b) Jo h n is considered[„, t to be intelligent]
(6 9 ) (a) T h e o r u n te ta a rch ea elinika na ine sinarpastika
are considered [the ancient G reek]nom -pl to be fa scinating-nom -neut-pl ‘A ncient G re e k is co n sid ered to be fa scin atin g .’
(b) T a a rch ea elinika th e o ru n te na ine sinarpastika.
(c) T h e o r ite ta a rch ea elinika na ine sinarpastika
is c o n sid e re d [the ancient G r e e k j n o m to be fa sc inating-nom -neut. (d ) T a a rc h e a elinika th eo rite na ine sinarpastika
In ( 6 8 a ) again the N P subject o f the infinitival cla u se tails to be C a s e - m a rk e d , on the
I he (Jin'erence between (67a and b), however, appears to be inteipretative and, perliaps, any potential explanation should take into account various pragmatic tenus invoked Thus, (67a) gives an impersonal llavoui to the proposition expressed by the domain of the low er clause, whereas (67b) expresses that the speaker is somehow invoked in the predication conceniing The child’ which is picked out ol the domain of the lower clause, lo r traiifonnational operations go\ enicd by discourse and contextual factors, among which subject-to-subject raising, see Creider ( 1979).
assumption that passive morphology deprives a verb o f its capacity to Case-mark its complement. Thus only raising o f the embedded subject NP to the Case-assigned matrix [spec, IP] position rescues the structure. By contrast, the Greek data in (69) suggest, once again, that no structural condition is responsible for the raising o f the embedded subject NP into the matrix [spec, IP] position. Indeed, as in the case with the ‘raising’ verb ‘fenete’ in section 2.5.3.9.1, the passive verb does not have to agree in number with the ‘raised’ noun, as shown in (69c, d) above. Here, theorite’ (it is considered) will be the 3rd person singular impersonal version o f the passive theorume’ (I am considered), even though the supposed passivized subject ‘ta archea elinika’ is plural.
Passivization in MG seems to be related more to semanticopragmatic factors rather than syntactic. This idea has been entertained by Philippaki (1976). According to Philippaki Greek passives express “. . .the involvement o f the subject in the meaning o f the verb. . .due to the fact that the same verb form covers the meaning o f both passive and middle or reflexive.” (ibid: 570). Note that passivization is a subject-creating construction (Gundel 1988). As seen above, however, passivization in Greek need not be employed as a subject-creating operation. This again follows if we assume that the subject need not be syntactically present in MG clause.