The gray area of supine clauses *.
This paper aims to clarify the categorial and syntactic status of supine forms in Romanian. It appears that the supine is a [+N] [+V] category that lacks [PHI]-features and differs from adjectives with reference to distribution, thematic properties, and syntactic checking. Crucially, the absence of [PHI]-features reflects on the syntactic configuration, which becomes valid in only two instances: within a DP, or within a CP structure. Supine complementation is the optimal choice for tough constructions. Checking conditions on [N] features in these constructions have as side effect a form of clause union. The result is a monoclausal structure, more economical in operations than tough constructions with infinitive or subjunctive complements.
Nonfinite verbal forms in Romance have been shown to generate clauses with nominal properties (Raposo 1987). Generally, this nominal property relates to the anaphoric tense of the nonfinite inflection and coincides with restrictions on complementizer (C) selection. For example, infinitive clauses may have the complementizers de or a, but not que, in all varieties of French, from Old French to Modern French (see Martineau 2000 among others). Most nonfinite clauses in Romanian present the same properties, with the exception of supines. The latter lack anaphoric tense, although they exhibit an obligatory C de. The investigation of these nonfinite forms helps clarify the clausal nominalization and the selectional restrictions on compatible Cs. Crucially, nominalization is dissociated here from anaphoric T: bare supines trigger specific syntactic processes that involve checking of nominal [N] features specified on the verb stem, as opposed to clustering on a functional head (i.e. T[ense]). Furthermore, the constraints on [N] checking prompt a clause-union configuration. This indicates that a clause union arises as an effect of nominalization in nonfinite clauses and not necessarily as a property of T (Kayne 1991).
1. The supine problem
Romanian supines look like invariable past participles, as in Table 1. The supine occurs in two environments: in a noun phrase, as in (1a), or in a clausal form preceded by de `of', as in (1b). In both configurations the supine has an active value and takes a direct complement. Note that these structures are productive in the language.
(1) a. [Culesul porumbului] e din ce in ce mai gathered-SUP-the maize-the-GEN is from that in that more greu. difficult `The gathering of maize is more and more difficult.' b. E din ce in ce mai greu [de cules is from that in that more difficult of gathered-SUP porumbul.] maize-the `It is more and more difficult to gather maize.'
Several questions arise from the description in Table 1 and the examples in (1). With respect to morphology, one has to define the process that accounts for the supine form. If the derivation is inflectional, through the suffix -t/-s, as in Romance past participles, one must expect a passivization of the verb; however, the supine form stays active. Moreover, the past participle inflection implies activation of agreement features, whereas the supine is invariable. (1) With respect to syntax, an explanation is in order for the facts illustrated in (2).
--Clausal supine must have a prepositional subordinator; that is de `of' in most cases:
(2) a. * E greu cules porumbul. is difficult gathered-SUP maize-the
--De `of' may not select a nominalized supine when it functions as a sentential complement: (2)
(2) b. * E greu de culesul porumbului. is difficult of gathered-SUP-the maize-the-GEN
--De supines are the choice for tough constructions in Romanian, whereas other Romance languages use infinitives and other Balkan languages use subjunctives.
(2) c. Problemele sunt greu [de rezolvat.] problems-the are difficult of solve-SUP `The problems are difficult to solve.'
Within a minimalist framework, this paper argues that the supine form has the -t/-s ending and lacks [PHI]-features in the lexicon. Categorially, the supine has [+V] [+N] features, like adjectives; however, absence of [PHI]-features determines a different syntactic behavior. In minimalism, [N] features are interpretable due to the presence of [PHI]-features and do not need syntactic checking. Along these lines, lack of [PHI]-features renders the [N] feature uninterpretable, and therefore it needs deletion through syntactic checking. This brings up a crucial difference between the syntax of adjective and the syntax of supines. The latter and not the former must undergo syntactic checking on an uninterpretable [N] feature, and this checking is fulfilled in two ways: the supine may pass the computation either as a noun formation (determiner phrase, DP) or as a verb formation (complementizer phrase, CP). The syntactic idiosyncrasies of supines seem to make them more economical in tough constructions and may explain their choice over infinitives or over subjunctives.
2. Morphological considerations
2.1. Supine nouns as agentive
Derivational morphology yields a specific class of nouns that have sprung from the supine form. Specifically, the definite article -(u)l for singular masculine and neuter encliticizes on the supine, as in (3a).
(3) a. Supine: pictat `painting' [arrow right] Noun: pictatul `the painting'
Most deverbal nouns based on the supine have a corresponding regular noun. The former designates the action, as in (3b), while the latter designates the result of that action, as in (3c).
(3) b. Critica cu mult patos pictatul peretilor critized-3sg with much passion painting-SUP-the walls-the-GEN de genul grafitti. of genre-the grafitti `He was passionately criticizing the grafitti-style wall painting.' c. Pictura moderna accepta genul grafitti. painting-the modern accepts genre-the grafitti `Modern painting accepts the grafitti style.'
The semantic dychotomy in (3) shows that the supine-based noun preserves some verbal properties, which impose an active, agentive reading.
2.2. Infinitive nouns as regular nouns
There is a second group of deverbal nouns in Romanian, derived from the old infinitive form. That is, Old Romanian has the infinitive suffix -re specific to Romance infinitives, for example culegere `to gather'. This suffix was dropped very early in the history of the language, leaving only the infinitive stem. However, the formation with -re became very productive in deverbal nouns, which add the definite article -a for singular feminine as an enclitic. Thus, the infinitive form culegere `to gather' becomes culegerea `gathering', `collection'. (3)
The comparison between supine and infinitive deverbal nouns is relevant for at least two reasons. First, it shows that when it comes to nominalization, the supine pairs with the infinitive and not with the past participle. Second, infinitive-based nouns behave on a par with regular nouns: they do not impose an agentive reading; sometimes they may even serve as "resultative" doubles for supines, as in (4).
(4) a. Ne-am sculat la cantatul cocosilor. us-have-1pl wake-up at singing-SUP-the roosters-the-GEN `We woke up when the roosters were singing.' b. Ne-a sculat cantarea/cantecul cocosilor. us-has-3sg wake-up song-INF-the/song-the roosters-the-GEN `The roosters' song woke us up.'
According to the readings on the deverbal nouns in (3) and (4), supine-based nouns are more verbal than infinitive-based nouns and regular nouns. This may serve as an empirical basis for the categorial status of supines like [+V][+N], which would contrast with the categorial status of infinitive-based nouns, which became [-V][+N]. The contrast between supine and infinitive deverbal nouns is further supported by tests of nominal inflection.
2.3. The degree of nominalization
Supine nouns contrast with the class of infinitive and regular nouns as follows: (4)
i. Indefinite articles modify infinitive nouns, as in (5b), but not supine nouns, as in (5a).
(5) a. Am participat la * un vopsit (de casa/al casei).// have-1sg participated to a painting of house/house-GEN// ok. vopsitul casei painting-the house-the-GEN `I participated in the painting of the house.' b. Am participat la o vopsire generala a casei.// have-1sg participated to a painting general of house-GEN// vopsirea casei. painting-the house-the-GEN `I participated to a general painting of the house.'
ii. Plural is possible with infinitive nouns, as in (5d), but not with supine nouns, as in (5c).
(5) c. Sing. masc/neuter: cantatul `singing' [left and right arrow] Pl. masc. * cantatii // Pl. neuter * cantaturile d. Sing. fem.: cantarea `song'; Pl. fem.: cantarile `songs'
As a consequence of the restrictions in (5a) and (5c), other indefinite modifiers are incompatible with supine nouns, as in (5e).
(5) e. * niste `some'/trei `three'/cateva `a few'/pictaturi `paintings'
iii. Adjectives compatible with an agentive reading may modify supine nouns, but they occur only on the right of the noun, as in (5f), whereas regular nouns, including infinitive-based nouns, may have the adjective on the left or on the right.
(5) f. Sup. ok. cautatul insistent al tainelor inquiry-the insistent of secrets-the Inf. ok. cautarea insistenta, a tainelor inquiry-the insistent of secrets-the Sup. * insistentul cautat al tainelor insistent-the inquiry of secrets-the Inf. ok. insistenta cautare a tainelor insistent-the inquiry of secrets-the `The insistent inquiry of secrets'
The examples in (5) show that supine nouns, unlike infinitive nouns, undergo a deficient nominalization. More precisely, the supine noun lacks the specifications related to number and possibly gender, allowing us to conclude that they lack [PHI]-features and undergo a low degree of nominalization. Adding to this the verbal behavior noted in the previous section, it seems justified to consider the supine a [+N], [+V] category. Syntactic checking on this category is necessary insofar as the absence of [PHI]-features renders the [N] feature uninterpretable. However, the [V] feature, fully specified in the lexicon, enters syntactic checking only as needed for the checking of functional features.
This prediction seems correct in light of checking constraints related to the restrictions on indefinite modifiers and adjective placement. Indefinite modifiers occupy the D slot, so that the nominal [D/N] feature is checked through feature matching against the noun in situ, as in (6a). In a DP with D -l adjectives may either stay in situ while the noun moves to D, yielding the noun-adjective order, as in (6b); or the adjective moves to Spec, DP and enclitizes the article, while the noun stays in situ, yielding the adjective-noun order, as in (6c). The common point of DPs with indefinite modifiers and DPs with adjectives is that in both cases there is a configuration in which the noun must stay in situ, and this configuration is ruled out by supine-based nouns. It appears that the supine is recognized as a nominal element only if it moves to a D occupied by a default article -l, as in (6d). So the poverty of lexical [N] specification is overcome in syntax only through movement to a head with a strong selective [N] feature, and the definite article is the only D head qualifying as such in the Romanian nominal system.
(6) [ILLUSTRATION OMITTED]
2.4 Conclusions on nominalization
Examples in section 2 show that deverbal nouns based on infinitive forms have a complete nominal categorial status, that is [+N][-V]. Nouns based on supine forms do not always reach this categorial definition. That is, some supine-based nouns exhibit deficiencies of nominalization (e.g. vopsitul `the painting'), while others behave as full-fledged nouns (e.g. apusul `the sunset'). The former but not the latter show restrictions related to definiteness, number/person, and word ordering within the DP and are relevant for the present investigation.
A fact worth mentioning is the intersection of the two classes, providing a gray area with some deficient nouns showing flexibility of word order (e.g. cantatul neasteptat/neasteptatul cantat `the unexpected singing'), while others may show insensitivity to number restrictions (e.g. mersul-sg./ mersurile-pl, trenului `train schedule' vs. mersul-sg./(* mersurile-pl.) pe jos `walking'). This mixed group indicates that a transition takes place toward full-fledged nominalization, and that is possible when the [N] feature is inherent, as opposed to acquired syntactically, within a functional domain.
3. Syntactic convergence
3.1. The status of de
This section approaches de-supine constructions in which the supine does not raise to a D -l. In these constructions, clues about the syntactic legitimation of supines come from the status of de. In Romanian, as in other Romance languages, de may be a prepositional determiner or a complementizer in nonfinite clauses. In both cases de has nominal features, as argued in Kayne (1994). (5) As a prepositional determiner, de embeds a DP with a predicational structure, such as is seen in possessive constructions in Romance, which also extend to Romanian, as shown in Stefanescu (1997). Along these lines, we would expect de to embed a supine-based DP. However, this is not the case when the de supine occupies a complement position, as shown in (2b), repeated for convenience.
(2) b. * E greu de culesul porumbului. is difficult of gathered-SUP-the maize-the-GEN
The grammatical version of (2b) must eliminate the definite article from the supine form and change the case of the object from genitive to accusative. Moreover, the supine stem cules `gathering' rules out noun-specific modifiers (i.e. adjectives), as shown in (7b).
(7) a. Culesul timpuriu al porumubului nu e recomandat. gathering-the early of maize-the-GEN not is recommended `The early gathering of the maize is not recommended.' b. * E greu de cules timpuriu porumbul. is difficult of gathering-SUP early maize-the
On the other hand, small adverbs may (although rarely) modify the supine form in de constructions, which is not possible in DP structures, as shown in (8).
(8) a. E greu de cules bine/iar porumbul cand is difficult of gathering-SUP well/again maize-the when ploua in fiecare zi. rains in every day `It is difficult to gather the maize properly when it rains every day.' b. * Culesul bine/iar (al) porumbului ... gathering-SUP-the well/again of maize-the-GEN
The facts in (2b), (7), and (8) show that in de constructions the supine form behaves like a verb: it rules out the definite article, it takes a direct object in the accusative, and it can be modified by adverbs. Accordingly, the element de does not embed a DP but a clausal structure and may, therefore, qualify as a complementizer in a nonfinite clause.
3.2. Internal structure: [V] feature checking in de supines
Since de constructions function as clausal units, the supine must be contained in a verbal phrase. As a transitive verb, the supine is merged with a direct object on the right. So the structure must be a VP shell in terms of Chomsky (1995), as in (9), where "light v" is the site for two checking tasks: the supine moves to "light v" to check the [V] feature of the latter, while the case feature of the DP-object is checked through feature matching, as in the configuration in (9).
(9) [ILLUSTRATION OMITTED]
A priori we can posit an analysis with supine raising to "light v," since all verbs raise overtly in Romanian. Empirical support for a movement analysis comes from the placement of small adverbs such as bine `well' and iar `again', which intervene between the supine and the direct object, as seen in (8a). These adverbs may adjoin to the embedded VP. However, the word order in (8a) may also result from verb movement to a functional domain outside vP, leaving behind several adjunction positions for adverbs. The following tests will show that the supine vP does not project an extended domain.
3.3. Internal structure: [D] feature checking in de supines
In all the Romance languages de is the complementizer that brings "nominal" properties to the clause. We assume here that all clauses are tense phrases (TP), as proposed since Chomsky (1995), that may be embedded under a CP. Thus, the [N] feature of C de is checked against the [N] feature of the nonfinite TP. This is not possible with other complementizers, such as Fr. que, Rom. ca, that do not have [N] features. In the case of supine clauses, a new factor arises: the supine has a lexically deficient [N] feature that needs further checking in syntax. In this case, the [N] features of de and of the supine must be checked against each other, without the mediation of a T head. The following tests show that this analysis is on the right track.
Signs of a functional domain come from the presence of clitic pronouns, auxiliaries, negation, or adverbs that merge with functional (aspectual) heads (Cinque 1997). No such elements may intervene between de and the supine or between the supine and the direct object, as shown in (10)-(14). Examples of infinitive clauses containing all these elements are offered as a comparative term.
(10) a. E greu [de (* i) adunat (* i).] is difficult of them gathering-SUP them b. Ne permite [de a- i aduna.] us permits of to them gather-INF `It permits us to gather them.'
(11) a. E greu [de (* fi) adunat materialul.] is difficult of have-INF gathering-SUP material-the b. Permisiunea [de a fi adunat permission-the of to have-INF gathered-PAST.PART materialul.] material-the `The permission to have gathered the material ...'
(12) a. E greu [de (* nu/ne) adunat materialul.] is difficult of not gathering-SUP material-the b. Ne permite [de a nu aduna materialul.] us permits of to not gather-INF material-the `It permits us not to gather the material.'
(13) a. E greu [de (* mereu/des/totdeauna) adunat is difficult of all the time/often/always gathering-SUP materialul.] material-the b. Ne permite [de (* mereu/des/totdeauna) a aduna us permits of all the time/often/always to gather-INF materialul.] material-the `It permits us to always/often gather the material.'
--Short inflectional adverbs:
(14) a. E greu [de (* mai) adunat material.] is difficult of again gathering-SUP material b. Ne permite [de a mai aduna material.] us permits of to again gather-INF material `It permits us to gather material again.'
Short adverbs such as mai `again' do not have modifiers and remain adjacent to the left of the verbal form irrespective of the degree of verb movement to inflection. They have been analyzed as part of the inflectional domain in Romanian (Cornilescu 2000 among others).
The examples in (10) to (14) show that the supine verb does not project an extended domain, so a supine clause is unable to accommodate clitics, auxiliaries, and negations. The situation of adverbs in (13) confirms the adjacency requirement between de and the supine. In infinitives, the adjacency concerns the highest inflectional head, that is, the mood marker a, not the infinitive stem. Adverbial clitics like mai in (14) point to the absence of a functional domain. Now, the presence of short (TP-independent) adverbs after the supine in (8a) can safely be considered as adjoined to the VP embedded in the vP. Therefore the representation of the supine vP in (9) is correct, with supine movement to "light v," and the complete clausal structure will have the configuration in (15).
(15) [ILLUSTRATION OMITTED]
The vP projects a subject position as its specifier (PRO) and ensures checking of the [V] feature through supine movement through "light v." De attracts the supine form to check the [N] feature, and thus it produces the obligatory adjacency between the two elements.
3.4. Internal structure: final analysis
De supines as sentential complements behave differently from de supines as prepositional adjunct phrases (PP) in that the former cannot embed a DP (see example [2b]). It was shown that supine forms in these constructions behave like verbs embedded in a CP: they take direct objects in the accusative and are modified by adverbs (see examples  and ). Further tests on the internal structure of CP de constructions showed that the supine does not project an inflectional domain, and the complementizer de attracts the supine to C, as in (15).
Checking constraints on the supine as a category with [+N], [+V] features, where [N] is deficient in the lexicon, explain why de is necessary in the supine clause: de is the only element that can check the [N] feature in a clausal configuration. However, one may wonder why de is obligatory, as shown in (2a), repeated below, whereas it may be optional in infinitive clauses, as in (16).
(2) a. E greu [* (de) cules porumbul.] is difficult of gathered-SUP maize-the (16) Ne permite [(de) a aduna materialul.] us permits of to gather-INF material-the `It permits us to gather the material.'
It appears that in (16) the nominal feature of the infinitive clause undergoes checking in the absence of de. It is important to notice, however, that the [N] feature of the infinitive is not lexical, but it belongs to the functional domain, as a property of T. Thus, [N] feature checking in (16) is not implemented by the null C (if it exists at all), but through the checking of the infinitive T, anaphorically related to the matrix T. When matrix T checks infinitive T for temporal features, it checks the entire feature set under T, including [N], if this has not been done by then. So [N] checking in (16) may free-ride on T checking. This is not possible in (2a) because the supine lacks a functional domain, and T checking cannot apply. Hence, de is obligatory, as the only checking device for the deficient [N] feature of the supine.
3.5. Conclusion on the sentential use of the supine
The supine stem has a lexical [V] feature, which makes it compatible with VP and clausal configurations. Due to the presence of a second lexical feature with nominal properties, the supine stem may not extend its verbal projection to a TP, so that the clause has the reduced structure CP-VP. In this structure, de is obligatory to ensure convergence of the derivation through checking of deficient [N]. This configuration applies in all the environments with sentential use of the supine, such as sentential complements, imperatives, exclamatives, relatives, and tough constructions. (7) The latter will be discussed in the next section.
A desirable consequence of this analysis is that past participles may now be understood as inflectional derivations from supine stems. The difference between the supine and the past participle is that the latter exhibits [PHI]-features, which in syntactic configurations are related to the presence of object agreement features within the extended domain of the verb. It is then reasonable to say that whenever the supine occurs in a checking relation with an object agreement head (e.g. when selected by `be'), the supine becomes a past participle.
4. Tough constructions
Romance languages form tough constructions with infinitive complements, as in French, (17a), whereas Balkan languages use subjunctive complements, as in Bulgarian, (17b). Although both infinitive and subjunctive complementation exist in Romanian, the choice for tough constructions is the de supine, as in (2c), repeated for convenience.
(17) a. Ces photos sont amusantes a regarder. these pictures are fun-fem.pl. to look-at-INF `These pictures are fun to look at.' b. Tezi problemi sa trudni da gi resis. these problems are tough-fem.pl. to them solve-SUBJ.2sg `These problems are tough to solve.' (2) c. Problemele sunt greu [de rezolvat.] problems-the are difficult of solve-SUP `The problems are difficult to solve.'
This section will show that the presence of supines in tough constructions as in (2) follows naturally from the analysis of de supines proposed in section 3. Crucially, the reduced de-VP structure is compatible with a type of clause union in which the embedded object moves to the subject position in the matrix clause. Thus, the tough constructions do not contain gaps, null operators, or reconstructed chains, which are postulated for constructions with full-fledged TP complements as in (17).
Tough constructions are a special case of clause union, in which the matrix predicative adjective undergoes "restructuring" with a selected TP. While adopting this perspective, we must note, however, that the clause-union mechanism established for other Romance languages does not account for Romanian equivalent constructions (i.e. T-to-T raising, as in Kayne 1991). So, any analysis of de supines in tough constructions must take into consideration the pattern of verb restructuring in the language.
4.1. Verb restructuring with infinitives
Motapanyane and Avram (2000) show that the modal putea `can' undergoes verb restructuring with a bare infinitive. That is, clause-union constructions such as (18) rule out any TP-related elements in the left periphery of the infinitive segment, as in (18a). So clitics, auxiliaries, negation, and short adverbs precede the matrix modal but not the infinitive, as in (18b).
(18) a. Maria putea [(* nu/le/mai/fi) citi Maria could not/them-DAT/more/have read-INF orice.] anything b. Maria nu le- ar mai fi putut [citi Maria not them-DAT would more have could read-INF orice.] anything `Mary would not have been able to continue reading them whatever.'
Notably, the infinitive marker a `to' must be omitted in this construction, as in (18c), showing further proof of the reduced inflectional domain.
(18) c. Maria putea [(* a) citi orice.] Maria could to read-INF anything
However, postverbal subjects (considered to be in situ in Spec, VP as in Cornilescu 2000) and VP-adjoined adverbs may follow the infinitive verb, as in (18d).
(18) d. (Maria) putea [citi (Maria) mereu anuntul.] Maria could read-INF Maria anytime ad-the `Maria could read the ad any time.'
Then the word order shows that the infinitive moves out of VP. This may follow if the infinitive stem moves overtly to matrix V for restructuring.
The relevant aspect of the syntactic behavior in (18) is that verb restructuring takes place between a modal and an infinitive stem and not between the modal inflection and an infinitive TP, as in Romance. The Romanian constructions in (18) show, then, that verb restructuring does not depend on the projection of T. It rather depends on the availability of head-to-head movement in a local domain, so that the selective modal features of matrix V get checked against the infinitive V, which has similar scope-bearing properties. In other words, the Romanian facts support a view of verb restructuring that posits the trigger in the modal properties of the matrix verb, rather than in the properties of its extended domain. Clause union may then be arrived at through different configurations as long as a local checking relation is obtained; both [V-[T.sub.INF]] movement to matrix T in Romance and [V.sub.INF] movement to matrix V in Romanian qualify since they ensure checking through head adjunction.
4.2. Restructuring with de supines
The case of verb restructuring with infinitive stems shows that Romanian adopts a derivational pattern that avoids an inflectional domain. This is important for de supines, since they also lack an inflectional domain, so it is expected that the mechanism for clause union will be different from the one proposed for other Romance languages (e.g. in Roberts 1997). From the beginning, we exclude a clause union mediated through T in tough constructions such as (2c); hence, we have to identify the trigger for the reanalysis and restructuring between the predicative adjective and the supine, as well as the mechanism for their implementation.
Adjectives qualify as mixed [+N], [+V] categories that modify nominals. The modifying property is manifested in syntax through an agreement relation between the adjective and a nominal element, which is translated through checking of the [N] feature of the adjective. In the case of tough constructions (where object raising applies), as in French, (17a), and Bulgarian, (17b), the object DP is merged in the matrix Spec,TP and thus it enters into an agreement relation with the verb in T. The predicative adjective has its [N] feature checked against the DP-T nominal complex, and thus it shares the agreement relation. This agreement may or may not be lexically visible, depending on the language. Spec,TP is a nonthematic position, so a DP occupying this position needs to be chain-related to a theta position, which in tough constructions corresponds to the empty object slot in the embedded VP. We suppose that the mechanism for chain formation is the same as has been proposed for equivalent English constructions since Chomsky (1977).
Romanian tough constructions present the outstanding peculiarity of an invariable adjective, as in (19a), even when object raising applies. Thus, (19) contrast with sentences such as those in (17a) and (17b), where the adjective shows obligatory agreement with the object DP.
(19) a. Problemele sunt greu de rezolvat. problems-the-fem.pl are-3pl hard-(default) of solving-SUP `The problems are hard to solve.' b. * Problemele sunt grele de rezolvat. problems-the-fem.pl are-3pl hard-fem.pl of solving-SUP
The contrast between (19a) and (17) indicates that the raised DP cannot merge in Spec,TP because it does not have access to the agreement relation with the predicative adjective. The next possibility is that the DP moves to Spec,TP, across de (which was shown in Kayne 1994, among others, to be transparent to movement in clause-union configurations) and obtains an agreement relation with the verb, as in any regular construction with a raising verb. Therefore, a DP-T complex with nominal properties emerges, but this complex does not check the [NJ feature of the adjective. So what blocks the checking relation between the DP-T complex and the adjective?
An analysis of (19a) must account simultaneously for the [N] checking on A and for the composition of the A-chain created by DP raising. This is achieved under the hypothesis that AP in (19) does not select the de supine, but it actually merges in Spec,de. Thus, the `be' verb selects the de phrase, as in (20), where the AP in Spec, de has the [N], [V] features matched against the [N], [V] features of the de supine, under a local Spec-head relation. The presence of de is necessary to make up for the lexical poverty of the supine [N]. The checking configuration is given in (20). In (20) the adjective remains invariable because its [N] feature is checked against the de-supine nominal complex instead of the DP-T complex. The object DP raises from one argumental Spec to another, bypassing Spec,CP, which does not have argumental status.
(20) [ILLUSTRATION OMITTED]
Clause union and transparency of de in (20) would predict the possibility of clitic climbing, which was shown to be agrammatical in these contexts, as in (21b). (8)
(21) a. Fotografiile sunt imposibil de aratat copiilor. pictures-the are impossible of shown-SUP children-DAT `It is impossible to show the pictures to the children.' b. * Fotografiile le sunt imposibil de aratat. pictures-the them are impossible of shown-SUP
The ban on clitic climbing across the complementizer de even in clause-union configurations has been discussed at length in the literature (see Martineau 1990 for a review on this topic). However, in (20), the transparency of de is increased by its function in nominal feature checking, to the extent where NP-movement may cross it. It would then be expected that a clitic could cross it as well.
The explanation proposed here for the lack of clitic climbing in (20) will be independent of the degree of clausal-border transparency. More precisely, the argument is that clitics cannot escape from de + V sequences on the same grounds as they cannot escape from sa + V sequences in subjunctive clauses. Consider the examples in (22).
(22) a. Maria le putea [citi]. Mary them could read-INF b. Maria (*le) putea [sa (le) citeasca]. Mary them could SA them read-SUBJ `Mary could read them.'
Both (22a) and (22b) have the same interpretation and undergo verb restructuring and clause union. However, clause union with a bare infinitive (as in [22a] and ) requires clitic climbing, whereas clause union with a subjunctive clause (as in [22b]) disallows it. The explanation for the restriction in (22b) is well accepted: the mood morpheme sa behaves as a prefix (or "discontinuous morpheme" in Dobrovie-Sorin 1994), attracting the verbal complex (that is, the sequence clitic pronoun-auxiliary clitic-verb) to its right. No clitic can be extracted from the emerging cluster. Thus, the ban on clitic climbing in (22b) follows from the morphologic properties of the subjunctive verb, irrespective of clausal-border transparencies. Similarly, de in (20) attracts the supine verb to its right, for head-to-head checking, and forms a barrier for clitic extraction on a par with sa complexes.
The empirical evidence supporting the analysis in (20) shows that extraction across de from the supine clause yields acceptable results, while it is infelicitous with infinitive complements, as in (23).
(23) a. Arhivele sunt [usor de cercetat cu aparate archives-the are easy of researching-SUP with devices speciale.] special `The archives are easy to research with special devices. b. Cu ce aparate sunt arhivele [usor de with what devices are archives-the easy of cercetat?] researching-SUP `What devices are the archives easy to research with?' c. Tehnologia ne permite [de a cerceta archivele technology-the us permits of to research-INF archives-the cu aparate speciale.] with devices special `Technology permits us to research the archives with special devices.' d. ?Cu ce aparate ne permite tehnologia [(* de) a with what devices us permits technology-the of to cerceta arhivele?] research-INF archives-the `What devices does technology permit us to research the archives with?'
Extraction from an infinitive complement with de in C is ungrammatical, unless de is deleted, as in (23d). This result contrasts with (23b), where de in C does not block the same type of extraction. This contrast proves that the construction in (23b) relies on a clause-union configuration, which cancels the barrierhood of de, whereas the sentence in (23d) maintains distinct CP domains. (9)
To conclude, lack of agreement marks on the predicative adjective in de supines indicates that the adjective merges in a position that is not accessible to checking from matrix T. This position was identified as the Spec of de. In this structure, object raising may take place through NP movement to the matrix subject position, across Spec, de. (10) This is possible because the intervening CP becomes transparent to argumental movement through the clause-union process. Thus, tough constructions with de supines differ from tough constructions in Romance and Balkan, (17), with regard to the mechanism for clause union and the location of the predicative adjective. As a result, the de-supine version of tough constructions achieves a more economical derivation, similar to clause union in constructions with raising verbs. (11)
5. Summary and conclusion
This paper has shown that the Romanian supine comes as a separate verbal stem with nominal features and qualifies as a [+V], [+N] category. The [N] feature is lexically weak insofar as it lacks [PHI]-features. Hence, the [N] feature of supines becomes uninterpretable and needs syntactic checking. The combination of [V], [N] features and the obligatory overt checking on [N] interact in a way that presents only two successful configurations for syntactic convergence: the supine merges either within a DP, which allows overt supine movement to D, or it merges within a VP, which ensures [N] checking against C. The two elements that meet the selective criteria in Romanian are the default definite article -l and the nonfinite complementizer de. So supines may occur only in DP with D -l or in CP with C de. It is interesting to notice that although the structural equivalence between DP and CP has long been established on theoretical grounds (Kayne 1994 among others), this is a clear empirical proof for their equivalence, since they both embed the same supine phrase.
The combination of [+V], [+N] features as well as the checking configuration in CP make the supines an optimal complement in environments with restructuring between a verb and an adjective of the tough class. Specifically, the de-supine complex is a lexical match for the adjective, which also has [+V], [+N] features and needs a form of checking on [N]. Checking takes place in a Spec-head local relation, which has as a side effect the implementation of clause union between the matrix domain and the supine. This configuration is equivalent to a monoclausal structure in which the embedded arguments freely move to the matrix domain. In this sense, the definition of the verb-restructuring mechanism leaves room for a higher degree of generalization. In Romanian, this process seems to occur whenever a local relation (head-head or Spec-head) can be established between the trigger and the candidate for restructuring, or between the extended domains of the two elements.
Table 1. Romanian supines Infin. stem Infinitive Past participle Supine -a lauda `to praise' laudat m.sg./laudata laudat f.sg./laudati `praising'/ m.pl./laudate f.pl. `to praise' `praised' -ea vedea `to see' vazut m.sg./vazut vazut f.sg./vazuti `seeing'/ m.pl./vazute f.pl. `to see' `seen' -e drege `to fix' dres m.sg./dreasa dres f.sg./dresi `fixing'/ m.pl./drese f.pl. `to fix' `fixed' -i povesti `to tell' povestit m.sg./ povestit povestita f.sg./ `telling'/ povestiti `to tell' m.pl./povestite f.pl. `told'
Received 11 May 2001
Revised version received
13 December 2001
* I am grateful to two anonymous reviewers for their helpful suggestions, to Gabriela Alboiu for comments on an earlier draft, and to Rod Hill for editorial assistance. Correspondence address: University of New Brunswick, P.O. Box 5050, Saint John, NB, Canada E2L 4L5. E-mail: email@example.com.
(1) This contrast arises between past participles used independently, as in small clauses like (i), versus supines, like (ii).
(i) O aveam [rezolvata] her had-1sg solved-PP.fem.sg `I had it solved.' (ii) O aveam [de rezolvat/(* rezolvata)] her had-1sg of solved-SUP `I had to solve it.'
Use of past participles in a verbal formation (i.e. complex tenses with auxiliary `have') is expected to change their syntactic behavior, since a different feature distribution will apply among the components of the verbal complex.
(2.) De + DP-supine occurs freely as an adjunct, with a clear PP structure; e.g.
(i) Spunea ceva apropos de culesul porumbului. said-3sg something regarding of gathering-SUP-the maize-the-GEN `She said something with regard to the gathering of the maize.'
(3.) Texts in Old Romanian attest the concomitant use of-re and bare-stem infinitives. In both cases an inflectional morpheme for mood, a `to', precedes the infinitive verb. E.g. a purtare//a purta `to carry'.
(4.) "Exception" to the rule: the nouns that follow the pattern in (5) have a supine stem that enters de structures as in (1) and (2). However, there are a number of supine-based nouns that behave like regular nouns; in general, their stem rules out de structures. E.g. Sup. apus `setting' [right arrow] apusul `sunset' (un apus `a sunset'; apusurile neuter, pl. `sunsets'); * ?Soarele e gata de apus `The sun is ready to set'. Unergative supines form most of this class, but other classes are concerned as well. Ambiguities exist when the supine may look like a verb, e.g. Anul scolar se apropie de sfarsit `The academic year nears (DE) the end-SUP'. However, the supine has a nominal value here, since it can take the form Anul scolar se apropie de sfarsitul mult dorit `The academic year nears the so much expected end', where the supine noun is modified by an adjective.
(5.) De occurs in the same environment when it selects a DP, as in (i), or a supine, as in (ii).
(i) o fata [de-a Mariei] a girl of/to Mary/`one of Mary's girls' (ii) o fata [de maritat] a girl of wedded-SUP/`a girl to be wed'
It is expected that de has the same (nominal) features in (i) and (ii), and checking applies.
(6.) Note that some supines may come in a negative form, e.g. necrezut `unbelieving', nepotolit `unrelenting'. These forms have the negative derivational prefix ne-, similar to negative nouns such as nebunul `un-good-the/the madman', necuratul `un-clean-the/the devil', and do not follow from an inflectional process. Their number is restricted.
(7.) Sentential complements:
(i) Am terminat [de cosit iarba.] have-1sg finished of mowing-SUP grass-the `I finished mowing the grass.'
(ii) De tinut la rece! of keeping-SUP at cold `To be kept at cold temperature.'
(iii) De necrezut! of unbelieving-SUP `It is unbelievable!'
(iv) spray [de curatat mobila] spray of cleaning-SUP furniture-the `a cleaning spray for the furniture'
(8.) I am grateful to an anonymous reviewer for pointing out the issue of clitic climbing and for suggesting these examples.
(9.) Wh extraction may apply over recursively embedded supines (see Grosu and Horvath 1987):
(i) Ce va fi greu de terminat de rezolvat pana vineri? what will be difficult of finished-SUP of solved-SUP until Friday `What will be difficult to finish solving before Friday?'
(10.) Argumental NP-movement is adopted here as the simplest way to account for nominative case and subject verb agreement in the matrix. However, if one would like to argue for a structure of Romanian as an exclusively VS language (preverbal subjects being internal topics), then NP-movement will have to be converted to whatever mechanism would account in that framework for subject-verb agreement.
(11.) Unlike other adjectives of the tough class, bun `good' exhibits agreement marks in constructions with de supines, as in (i).
(i) Legumele sunt bune de murat cu sare. veggies-the are good-fem.pl, of preserving-SUP with salt `The veggies are good for preserving in salt.'
Bun constructions accept subjunctive complementation, behaving thus differently from other tough constructions. Actually, bun opts for the Bulgarian pattern of tough constructions as seen in (17), where the DP merges into the matrix clause.
Cinque, Guglielmo (1997). Adverbs and Functional Heads. New York: Oxford University Press.
Chomsky, Noam (1977). On wh movement. In Formal Syntax, P. W. Culicover, T. Wasow, and A. Akmajian (eds.), 71-132. New York: Academic Press.
--(1995). The Minimalist Program. Cambridge, MA: MIT Press.
Cornilescu, Alexandra (2000). The double subject construction in Romanian. In Comparative Studies in Romanian Syntax, V. Motapanyane (ed.), 83-134. Oxford: Elsevier.
Dobrovie-Sorin, Carmen (1994). The Syntax of Romanian. Berlin: Mouton de Gruyter.
Grosu, Alexandru: and Horvath, Julia (1987). On non-finiteness in extraction constructions. Natural Language and Linguistic Theory 5, 181 196.
Kayne, Richard (1991). Romance clitics, verb movement, and PRO. Linguistic Inquiry 22, 647-686.
--(1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Martineau, France (1990). La montee du clitique en moyen franxais: une etude de la syntaxe des constructions infinitives. Unpublished Ph.D. dissertation, University of Ottawa.
--(2000). Les prepositions a et de de l'ancien francais au francais classique. In Le traitement de texte (edition, apparat critique, glossaire, traitement electronique), Claude Buridant (ed.), 165-183. Strasbourg: Presses de l'Universite de Strasbourg.
Motapanyane, Virginia; and Avram, Larisa (2000). The syntax of putea and its mixed typology. Belgian Journal of Linguistics 14, 145-162.
Raposo, Eduardo (1987). Case theory and Infl-to-Comp: the inflected infinitive in European Portuguese. Linguistic Inquiry 18, 85-110.
Roberts, Ian (1997). Restructuring, head movement and locality. Linguistic Inquiry 28, 423-46O.
Stefanescu, Ioana (1997). The Syntax of Agreement in Romanian. MIT Working Papers in Linguistics 14. Cambridge, MA: Department of Linguistics, MIT.
University of New Brunswick--Saint John
|Printer friendly Cite/link Email Feedback|
|Publication:||Linguistics: an interdisciplinary journal of the language sciences|
|Date:||May 1, 2002|
|Previous Article:||Anaphora: lexico-textual structure, or means for utterance integration within a discourse? A critique of the functional-grammar account. (1).|
|Next Article:||The role of morphology in gender determination: evidence from Modern Greek *.|