On the origin of inflected “non-finite” forms: the infinitive vs the supine

This paper deals with two verbal forms which, despite being traditionally labelled as “non-finite”, display inflection/agreement. We will focus on the behaviour and origin of the inflected infinitive attested in Romance and in languages from other families, against which we analyse the novel inflected supine found in the north-eastern area where Romanian is spoken (comprising the Republic of Moldova, Ukraine and the north-eastern part of the Romanian province of Moldova). The goal of the paper is to identify the common paths of diachronic change of these verbal forms and to put forward a formal account of the observed diachronic changes. From a diachronic perspective, our analysis shows that the functional structure of non-finite forms may become more enriched, a conclusion that is at odds with traditional findings, which generally argue for simplification, not enrichment of functional structure. At the same time, the proposed analysis also offers some insights into the diachrony of the supine marker de.


Introduction
In this paper we present the most significant linguistic data related to the inflected infinitive found in Romance and in languages from other families, focusing our attention on the origin and evolution of these forms as they are presented in the literature.We then turn to a series of recent special usages of the Romanian supine in the Moldavian variety employed in The Republic of Moldova, Ukraine and northeastern Romania1 , which bring the Romanian supine closer to the Romance inflected infinitive.Finally, we put forward a scenario which accounts both for the innovations in the morphosyntax of the Romanian supine (the enrichment of its functional structure) and for the diachronic recategorization of the supine marker de.
On the basis of the data presented below, our goal is to answer the following questions: (i) to what degree is the inclusion of these inflected forms in the 'non-finite' category justified?; (ii) what is the relevance of the origin and evolution of these forms for their morphosyntactic behaviour?; (iii) what are the definitional features of the infinitive (and of non-finite forms in general), including control properties, ability to accept a subject and morphology?; (iv) do the diachronic changes that affect these forms also lead to processes of grammaticalization?
We have chosen to compare the Romanian supine with the inflected infinitive from other languages because, of the Romanian non-finite forms, only the supine appears to favour the attachment of inflectional markers, while the infinitive remains non-inflected for person and number.The analysis of the Romanian supine will be shown to contribute to a better understanding of the category of 'finiteness' , a concept insufficiently understood in current syntactic theorizing, and of the diachrony of this category (see Ledgeway, 2007;Vincent, 1998).

The inflected infinitive
Before delving into the analysis of the inflected infinitive data, it is necessary to make a terminological clarification (following Ledgeway, 1998): we distinguish between the 'personal' infinitive particular to many languages including Romanian (see Mensching, 2000), a form which may take its own lexical subject in the nominative (1), and the 'inflected'/'conjugated' infinitive limited to a more limited number of languages and which displays person and number agreement with the subject (2).
In what follows, we exclusively deal with the 'inflected'/'conjugated' infinitive, which is attested in languages distinct from a genealogical and typological point of view: Portuguese, Galician, Sardinian, Southern Italian Dialects from South Calabria, Old Neapolitan, Old Leonese (of the Romance varieties), Hungarian, Welsh, West Greenlandic, Evenki (a Manchu-Tungusic language) and, probably, Greek.Miller (2003) formulated the hypothesis that the emergence of the inflected infinitive is determined by the presence of the lexical subject (allowed by the inflected infinitive only in very restricted contexts - Ledgeway, 2000Ledgeway, , 2007)), which has inherent subject agreement features (i.e., phi-features).
(3) a. Depois de chegarem, fugimos after of arrive.inf.3plfled.1pl'After they arrived, we fled' (Willis, 1971, p. 338, apud Bentley, 2014, p. 96) b. despois de eles chegarem viram as ruínas after of they.nomarrive.inf.3plsaw.3pl the ruins 'after they arrived, they saw the ruins' (Willis, 1971, p. 338, apud Ledgeway, 1998, p. 7) Groothuis (2015) shows that the structure in which the European Portuguese inflected infinitive is employed is biclausal, on the basis of two diagnostics: the matrix verb and the embedded infinitive may have distinct lexical subjects (4a), and both predicates may be simultaneously modified by the same adverbial (4b): (4) a. Nós lamentamos terem eles recebido pouco dinheiro we regret have.inf.3plthey received little money 'We regret that they have received little money' (Raposo, 1987, p. 97 (adapted from Raposo, 1987, p. 97) There are significant differences between Old and Modern Portuguese with respect to the distribution of the inflected infinitive.First, while in Modern Portuguese this form is available only in embedded clauses (Raposo, 1987, p. 86), in Old Portuguese it could appear in both main and embedded clauses (Martins, 2006, p. 342).Second, there are significant differences between the two stages of Portuguese with respect to the infinitive selected by causative verbs and direct perception verbs: while the canonical infinitive is selected by these classes of verbs in both stages of Portuguese (5a), the inflected infinitive is not attested in Old Portuguese in these contexts, but it is perfectly grammatical in present-day Portuguese (5b) (see also Sheehan, 2015, where a series of less categorical findings are presented); thus, examples like (5b) are not attested in pre-15 th c.Portuguese, but are perfectly grammatical starting with the 15 th c. (Martins, 2006).
(5) a. Mandei/ Vi os polícias prender o ladrão sent/ saw the corps arrest.inf the thief b.Mandei/ Vi os polícias prenderem o ladrão sent/ saw the corps arrest.inf.3pl the thief 'I made/I saw the policemen detain the thief ' (Martins, 2006, p. 327) It appears that this distribution correlates with other features of the canonical infinitive in control or raising configurations.Thus, while in Old Portuguese verbal negation did not precede the infinitive in these structures, but rather occurred on the main predicate, in Modern Portuguese both the embedded infinitive and the selecting verb may be independently negated.(Martins, 2006, p. 328) Martins (2006) shows that these phenomena are indicative of a syntactic change, from a more reduced to a more extended functional structure, in the history of the Portuguese infinitive.For the analysis of the Romanian dialectal supine, we will keep in mind the idea that the functional structure of non-finite forms undergoes a diachronic process of enrichment.

Sardinian
Relevant data on the Sardinian inflected infinitive (8) are given especially by Jones (1992Jones ( , 1993Jones ( , 2003) ) and Miller (2003).In Sardinian, the inflected infinitive is devoid of temporal autonomy and does not have an inflectionally specific profile, being syncretic with a finite form, the imperfect subjunctive.This syncretism results from the common origin of these forms, the Latin imperfect subjunctive (Jones, 1993, p. 278).
(8) 'sing' Singular Plural 1 cantárepo cantáremus 2 cantares cantáredzis 3 cantaret cantaren (Jones, 1992, p. 298) The difference between the two verbal forms is apparently given by the type of complementizer which introduces them and by the position of the subject.In (9a) the imperfect subjunctive appears in clause headed by the finite complementizer ki, and the subject is preverbal, while in (9b) the inflected infinitive is headed by a, and the subject is postverbal (Jones, 1993, p. 279).Word order differences of this type usually correlate with different V-raising options in the functional structure of the clause (V-to-I / V-to-C), a further potential difference between these two forms.
(9) a. non credío ki Juanne ésseret inoke not thought.1sgthat John be.subj.imperfhere 'I did not think that John was here' (Jones, 1993, p. 279) b. non kelio a cantares tue not want1.sgto sing.inf.2sgyou.nom'I don't want you to sing' (Jones, 1992, p. 297) Furthermore, the distribution of these two forms is not identical.The inflected infinitive is employed when its subject has independent reference (10a), but is usually excluded in obligatory control configurations (10b).
In obligatory control configurations, the subject of the infinitive is obligatorily shared with the selecting predicate, and the infinitive is devoid of temporal independence (11c).In both distributional contexts, it is preceded by a or de, which have been analysed as complementizers.The inflected infinitive has specific forms only in the plural; its singular forms are syncretic with the canonical infinitive-see the Table in (2) above.According to Ledgeway (2007, p. 340-341), the paradigmatic pressure of the singular, which contained only forms syncretic with the canonical infinitive, determined the loss of the plural inflection of the infinitive.
(11) a. ave plazuto a li nuostri Diey de nuy esseremo in questa parte has pleased dat the our Gods of us be.inf.1pl in these parts 'it pleased our Gods for us to be in these parts' (Ledgeway, 2009, p. 600) b. per nuy averemo ordene for we have.inf.1plorders 'so that we may receive orders' (Ledgeway, 2009, p. 922) c. se nui avertevamo de le andarimo appriesso if we considered of cl.acc.3pl=go.inf.1pl after non ne scapava nissciu(n)o not cl.dat.1pl=escapednot.one'if we had intended to go after them, not one of them would have got away' (Ledgeway, 2007, p. 338)

Southern Italian Dialects
Although traditional scholarship does not record the existence of the inflected infinitive in southern Calabrian varieties, Ledgeway (1998) has shown that the clauses headed by mu/ma/mi (< Lat.modo) are infinitival (see also Miller, 2003, Ledgeway, 2007), in opposition to those headed by ca-/chi-, which are finite.The following arguments have been invoked for granting infinitival status to the modern reflexes of Lat.modo (Ledgeway, 1998(Ledgeway, , 2007; see also Taylor, 2016 for a discussion of modo in Nicoterese): (i) the only possible word order is subject > modo (12a), which indicates that modo is not a complementizer, as complementizers are higher than the subject in the clausal hierarchy; (ii) modo can co-occur with another complementizer, namely ca, especially in hortative sentences (in (12b), chimmu < ca + mu); (iii) clauses headed by modo are devoid of temporal independence; their tense is anaphoric, i.e. fully determined by the tense specification of the higher selecting predicate.

Non-Romance languages
The data on the inflected infinitive in non-Romance idioms are rather scant, hence probably incomplete.Forms considered as matching the Romance inflected infinitive have been recorded in Standard Greek and Romeyka (/Pontic Greek, a dialect of Greek spoken in north-eastern Turkey), Hungarian, Evenki, and Welsh.

Greek
With respect to Standard Modern Greek, Miller (2003) considers that the verbal form inflected for person and number agreement with the subject and preceded by na is quasi-infinitival, on a par with what was noted above for southern Calabrian.However, most scholars have analysed the na-structures as subjunctives which have diachronically replaced the infinitive.Miller (2003) supports his proposal with the following arguments: (i) these structures appear in obligatory control configurations ( 13 Romeyka presents two features which distinguish its infinitival system from that of Standard Modern Greek (Sitaridou, 2014): (i) the canonical and the personal infinitive (i.e. the infinitive that may take its own subject) have been preserved, and (ii) there emerged a novel form, the inflected infinitive (15).

Hungarian
As shown by Miller (2003), Hungarian possesses a canonical infinitive, whose ending is -ni, and an inflected infinitive, in the structure of which the morpheme -ni is followed by person and number agreement markers.The inflected infinitive has been attested since Old Hungarian.While the canonical infinitive has a controlled PRO subject (16a), the subject of the inflected infinitive is either overtly realised or null (i.e.pro) (16b): (  Miller (2003) identifies two general sources for the inflected infinitive: the purpose subjunctive (in Romance and Balkan languages) and the adjunction of pronominal elements on nominalizations (in Hungarian, West Greenlandic, and Welsh).However, the analysis of a bigger number of languages shows that the sources of the inflected infinitive are more diverse.

2.3.1.
The inflected infinitive originating from other verb forms 2.3.1.1.At least three hypotheses have been put forward for the origin of the Portuguese inflected infinitive (see Pires, 2002;Miller, 2003;Scida, 2004;Carvalho, 2015): (i) José Maria Rodrigues (1913) claims that the inflected infinitive directly descends from the Latin imperfect subjunctive, both with respect to its form (17), and with respect to some of its functions; this "subjunctive" remained in use until the 15 th -16 th centuries (Ledgeway, 1998, p. 6).
(17) Lat.amaremus > Ptg.amar-mos love.inf.1pl'love' (ii) Theodoro Henrique Maurer (1984) claims that the origin of the inflected infinitive is the personal infinitive; in other words, the inflected infinitive emerged analogically and spontaneously from the personal infinitive with a nominative subject.(iii) Gamillscheg (1970), also followed by Miller (2003), puts forward a mixed hypothesis, according to which the inflected infinitive resulted from the overlapping between the Latin imperfect subjunctive and the Romance infinitive in specific contexts: (18) a. Lat.Placuit (nōbis) vende-re pleased.3sgwe.dat sell.inf'it was agreeable (to us) to sell' b.VLat.Placuit nobis ut venderemus pleased.3sgwe.dat comp sell.conj.imperf.1pl'it was agreed that we sell' (Roberts, 1953(Roberts, -1957, p. 30, p. 30 As shown by Miller (2003), the Latin infinitive (18a) and imperfect subjunctive (18b) were functionally equivalent and occurred in free variation in many texts.Since Early Latin, the complementizer ut ('to, in order that/to') was frequently omitted as in (18c) (see Pires, 2002, p. 145).Accordingly, at the basis of the Portuguese inflected infinitive lie structures like (18c), the diachronic change being favoured by the formal resemblance of the Latin and Portuguese canonical active infinitive (Lat.vendere, Ptg.vender) and by other common contexts of occurrence.

2.3.1.2.
The source of the Old Neapolitan inflected infinitive is the Latin pluperfect indicative (Loporcaro, 1986;Ledgeway, 1998, p. 6;Martin Maiden, p.c., expresses serious doubts about the accuracy of this hypothesis, arguing that the inflected infinitive is built with the imperfective stem and not with the perfective stem, specific to the pluperfect).Interestingly, in Old Neapolitan all non-finite forms also had inflected variants, albeit not used with the same frequency for all persons (Loporcaro, 1986, p. 173-174).( 19) ama(ve)ramu(s) > amàramo 'we had loved' > amare-mo /a'marəmə/ 'love.inf.1pl' The origin of the Romeyka inflected infinitive is the canonical infinitive.Sitaridou (2014) puts forward the following scenario depicting this process: the contrafactive form ixa ('I had') plus the infinitive is attested in the medieval period; appearing in strict adjacency to the verb ixa, the infinitive develops analogic inflection, identical to the aorist of ixa (20).Analogy was also favoured by the formal identity between the canonical infinitive and the third person form of the aorist.Subsequently, the inflected infinitive spread as a complement of modal verbs and, due to the fact that Romeyka does not possess complementizers, in other contexts as well (as a complement of volitional, causative and perception verbs).
(20) a. ixe ipina had.3sg say.aor.inf.1sg'If I had said' b. ixe ipines had.3sg say.aor.inf.2sg'If you had said' 2.3.2.The inflected infinitive originating from other classes of words 2.3.2.1.The Welsh inflected infinitive is the result of the reanalysis of some inflecting prepositions as infinitive agreement markers (21b) (Miller, 2004).Welsh does not have an infinitive form with dedicated inflectional marking, but it does have verbal nouns which are sometimes accompanied by the preposition i ('to' , 'for').The inflected infinitive thus resulted from the merger between the marker i, agreement markers and the verbal noun; the entire complex possesses person inflection and the preposition i is reanalysed as an infinitival marker ( 22). ( 21) a. i-daw 'to him' (Middle Welsh) b. i-ddo 'to-3sg.m'(Modern Welsh) (22) death y dyn [i-ddynt ei gweld hi] came the man to.3pl 3sg.f see.nonfin her 'the man came so they could see her' (Tallerman, 1998, p. 119) 2.3.2.2.Finally, the source of the Evenki inflected infinitive (23b) is the marker of nominal possession (23a), which was reanalysed as an agreement marker (Miller, 2003).

Results and problems
From this brief survey of the inflected infinitive and of its origin, we can draw a few more general observations: (i) not only the infinitive, but also other non-finite verbal forms have inflected/(subject) agreeing variants; in what follows, we will show that the Romanian supine is such a form with a non-inflected and an inflected variant; (ii) the source of the Romance inflected infinitive is not the canonical infinitive, but finite forms (the subjunctive, the indicative); the inclusion of these forms in the class of 'infinitives' is due to their distributional (syntactic) rather than inflectional properties; (iii) of the analysed languages, it appears that the inflected infinitive developed out of the canonical infinitive only in Romeyka and Hungarian (the label 'inflected infinitive' thus seems most appropriate when applied to these languages); we will show that the Romanian inflected supine is an extension of the canonical supine, a fact which draws Romanian closer to Greek varieties or Hungarian rather than to the Romance languages.
Furthermore, in view of the data analysed, one may wonder what a non-finite form is and how an 'infinitive' can be defined.How is it possible to distinguish an inflected infinitive from a subjunctive in view of the fact that both forms are inflected and may assign nominative case to the subject?It is true that the canonical infinitive is 'non-finite' in the morphological sense (i.e. it does not display number and person agreement with the subject) (see Ledgeway, 2007, p. 336 for the concept of 'morphological finiteness'), but the inflected infinitive is morphologically finite.Potential answers to these questions may be given if we adopt a scalar view on finiteness, as proposed by Ledgeway (1998, p. 8) In this representation, (a) and (d) represent the unmarked options available in many languages, while (b) and (c) represent points of significant cross-linguistic variation (see, for details, Ledgeway, 2000Ledgeway, , 2007)).
In the next section, we will also identify the position of the Romanian supine in this typology.

The Romanian supine: standard vs dialectal
It is a well-known fact that the Romanian supine is a non-finite form stricto sensu, i.e. it does not display morphological variation and, in general, does not accept a lexical subject, with a few exceptions (Pană Dindelegan, 2011;Dragomirescu, 2011).Its distribution is limited to a few well delimited syntactic contexts (see Pană Dindelegan, 2008Dindelegan, , 2013;;Dragomirescu, 2013a,b).The Romanian verbal supine has the following relevant syntactic properties: it may take an accusative direct object (25a), and it cannot combine with pronominal clitics (25b), clausal negation (25c), and clitic adverbials (25d).Pronominal clitics, negation and clitic adverbs obligatorily undergo raising to the higher selecting predicate (25e).
(25) a.In some varieties of Romanian spoken in The Republic of Moldova, Ukraine and north-eastern Romania, the supine behaves differently: (i) it has a wider distribution than in Standard Romanian; (ii) it may combine with pronominal clitics, clausal negation, and clitic adverbials in certain contexts; (iii) it may reflect agreement with the subject when it is selected by the verb trebui ('must, have to') or when it appears in tough-constructions.

Combination with pronominal clitics
The ability of the supine to host pronominal clitics in these Romanian varieties has been recorded by Gabinschi (2010) and analysed by Dragomirescu & Hill (2014) and Dragomirescu (2015).This property manifests itself in the following contexts: (i) after modal verbs such as a avea ('have') (with a personal paradigm), a trebui ('must, have to'), a putea ('can, be able to') (used as impersonal verbs): ( (30) ca aceste răspunsuri să merite that these answers să subj worth.subj.3plde încercat de le căutat de sup try.sup de sup cl.acc.3pllook.for.sup'in order for these answers to be worth looking for' (www.opinii.md)(iv) after e greu de (i.e.'tough-constructions'): (31) a. eu cred că aestea mici e mai greu I believe that these little is more tough de le făcut de sup cl.acc.3plmake.pple'I think that these little ones are tougher to make' (www.torrentsmd.com)b. bune sfaturi, greu de le urmat good advice.pltough de sup cl.acc.3plfollow.sup'good advice, tough to follow' (blogs.fanbox.com)(v) after adjectives taking complements introduced by de: (32) demne de le urmat worthy.f.pl de sup cl.acc.3plfollow.sup'worthy of following them' (ro-ro.facebook.com)(vi) in topic predicate fronting constructions: (33) că de le pus în geantă, tot vreau that de sup cl.acc.3plput.pple in purse still want.ind.pres.1sgsă le pun de vreo săptămînă să subj cl.acc.f.3pl put.pple for a week 'because as for putting them in my purse, I wanted to put them for a week' (www.miresici.ro)These examples also bear witness to another important feature of the supine in the varieties under investigation, namely the extension of its distribution in contrast to Standard Romanian.The supine may be selected by the modal verb a putea ('can, be able to') (used impersonally) and by the conative verb a încerca ('to try').Furthermore, the modal verb a trebui ('must, have to') always selects a de-supine, in contrast to the standard language, in which trebui selects a bare supine (this representing the only context in which the verbal supine is not introduced by de) or a participle (see, for details, Sandfeld & Olsen, 1936, p. 281;Rosetti, 1968, p. 258;Lombard, 1974, p. 301;Neamțu, 1980, p. 512-513;Pană Dindelegan, 2007, p. 170-171, 2011, p. 121;Dragomirescu, 2013aDragomirescu, , p. 36-38, 2015)).

Combination with verbal negation
While the Standard Romanian supine combines only with the prefixal negator ne-, but not with the freestanding negator nu (see also Cornilescu & Cosma, 2010 for a different interpretation of prefixal supine negation), in the north-eastern varieties the supine is compatible with the freestanding negator nu: (34) a. Sînt multe de spus, multe de nu le spus are many de sup say.supmany de sup not cl.acc.3plsay.sup'There are many things to be said and many things not to be said' (sorinels.blogspot.ro)b. trebuia de nu le pus pampers need.imperf.3sgde sup not cl.acc.3plput.pple pampers 'It was necessary for us not to put them in pampers' (hainutebebe.com)

Combination with clitic adverbs
In contrast to the standard variety, in the dialects analysed the supine may combine with aspectual clitic adverbs: (35) Așa că trebuie de le lăsat poarta deschisă so that must de sup cl.acc.3plleave.supdoor open spre plecare și de nu-i mai netezit pe bășcălie for leaving and de sup not=cl.acc.3plmore protect.supon mockingly 'One must leave the gate open for them to leave and not to mockingly protect them anymore' (m.publica.md)

The agreeing supine
Although in Standard Romanian the supine does not display agreement in any context, in the analysed varieties after the verb a trebui ('must, have to') the supine displays gender and number agreement with 3 rd person subjects (the 1 st and 2 nd person subjects are excluded from the agreeing configurations), irrespective of the position of the subject, i.e. raised to the main clause (36a-b) or in post-supine position (36c-e); gender and number agreement is also attested in tough-constructions (37).

Results and problems
As we have seen in this section, the supine of the north-eastern varieties fundamentally contrasts with the Standard Romanian supine on the following points: (i) the extension of its distribution, (ii) its compatibility with pronominal clitics, the verbal negator nu and the aspectual clitic adverb mai, and (iii) gender and number agreement.In the next section, we set the evolution of the Romanian supine against the evolution of the inflected infinitive in other languages, and we put forward a formal analysis which accounts for the present-day behaviour of the supine in the north-eastern varieties of Romanian.

Analysis
In the analysis that follows, we adopt the currently accepted generative clausal structure, according to which clausal structure is split into three layers: the CP-layer (the functional domain of the complementizer, which accommodates complementizers, wh-phrases and other elements displaced to the left periphery), the IP-layer (the functional domain which hosts mood-tense-aspect projections and the pronominal clitic field; the projection of this domain is also responsible for nominative assignment), and the vP-layer (the lexical domain, where the verb and the core arguments are merged).Internal arguments are merged and case-marked in the lexical domain.By contrast, the external argument (the subject), although generated in the lexical domain (Koopman & Sportiche, 1991), is dependent on the projection of the TP phrase (included in the IP-domain) for nominative assignment (Chomsky, 1981;see Cornilescu, 2000 andStan, 2005 for Romanian).

Standard Romanian supine
The syntactic diagnostics reviewed above (absence of subject, incompatibility with verbal negation, pronominal clitics and clitic adverbs) indicate that the supine has a reduced functional structure-cf.many proposals about structure of imperatives in many languages-, as proposed in (38): According to this representation, in the functional structure of the supine the C-domain is not projected at all, and the I-domain (which, in Romanian, accommodates the negation phrase, the clitic-hosting person phrases, and the mood, tense, and aspect phrases ordered as such-see, for details, Nicolae, 2015) is defective: the NegP, the TP (responsible for nominative assignment) and the pronominal-clitic-hosting PersP are absent, the entire I-domain of the supine being occupied by the marker de.By contrast, the lexical domain vP is fully projected, as shown by the fact that the supine may take accusative direct objects (generated as sisters to V) and dative indirect objects (S-a apucat [de trimis cadouri copiilor dative ] supine 'He started sending gifts to the children'); indirect objects are also generated in the vP-domain, as specifiers of an Applicative phrase (ApplP) (i.e.v > Appl).The absence of the subject in (non-raising) supine configurations follows from the absence of the TP-projection in the functional structure of the supine: without a TP-projection, nominative assignment is not available, and the existence of a caseless nominal violates The Case Filter (Chomsky, 1981).

Dialectal supine
In contrast to the Standard Romanian supine, the dialectal supine has developed a richer functional structure, as shown in ( 39

M-T-A făcut
The functional structure of this supine is thus more similar to that of the finite clause, in which all functional layers, CP, IP and vP, are projected.The marker de has been pushed upwards in the structure, becoming a C-head; thus, in these varieties, we witness the reanalysis of this inflectional element as a complementizer.Furthermore, it appears that the C-domain has developed a left periphery which may accommodate focused and topicalized constituents (40) (Hill & Dragomirescu, 2014).However, these examples are ambiguous, because the boldfaced constituents can be also interpreted as modifying the main verb.
( The I-domain is richer than in the standard language, accommodating at least the following phrases (specific to finite clauses with a fully articulated functional domain): the NegP, which hosts the clausal negator, the PersP (/Person-field), which accommodates pronominal clitics, and the AspP (/an Asp-field, à la Cinque, 1999), responsible for the merger of aspectual clitic adverbs.In the data examined so far, we have not encountered supines with their own nominative subject distinct from the subject of their selecting verb; thus, there is no significant empirical evidence for a nominative-assigning TP.In conclusion, while the functional structure of the dialectal supine is obviously richer, it is, however, impoverished in certain respects when compared to the structure of fully articulated finite clauses.
The full CP-status of the dialectal supine is further confirmed by the fact that it may undergo CPtopicalization: (41) a. dar [de le adus] i trebuie t i but de sup cl.acc.f.3pl bring.pplemust 'but bring them, one must' (www.realizat.com)b. [De le scris] i nu e greu t i , de sup cl.acc.f.3pl write.pplenot is hard am adăugat una aux.perf.1sgwrite.ppleone 'writing them isn't hard, I added one' (forum.softpedia.com)

The supine on the finiteness scale
The analysis of the dialectal supine from the perspective of the finiteness scale (24) put forward by Ledgeway (1998 and ssq.) indicates that this form behaves exactly like the Romance inflected infinitive, in that it displays all the features of a finite clause, except for the ability to take its own lexical subject.Consequently, the dialectal supine is characterized as [´Tense, `Agreement], which indicates that its Idomain, although featuring at least a NegP, a PersP (/Person-field), an AspP (/Asp-field), is still defective, i.e. its nominative-assigning TP is absent.In other words, the dialectal supine is morphologically finite (i.e., it displays morphological agreement), but syntactically non-finite (i.e., it is unable assign nominative case) (see Ledgeway, 2007).

Conclusions
This paper has examined the behaviour of the inflected supine attested in the north-eastern varieties of Romanian against the behaviour of the inflected infinitive attested in various Romance and non-Romance languages.
The analysis presented above allows us to formulate a few conclusions which concern, on the one hand, the category of 'finiteness' and its diachronic behaviour, and, on the other hand, the diachronic changes affecting the Romanian supine.
(i) Owing to its array of heterogeneous morphological, syntactic and semantic features, 'finiteness' is one of the least understood linguistic concepts (Ledgeway, 2007).Furthermore, the diachronic changes affecting it are also incongruous.For example, in the passage form Old to Modern Neapolitan, the inflected infinitive has been replaced by the personal and canonical infinitives, while the Romanian supine, which has been morphologically invariable and had a reduced functional structure, has become an inflected 'non-finite' form with an enriched functional structure in the dialects examined.Thus, diachronically, the degree of finiteness of a certain verbal form may become enriched or impoverished.