Istro-Romanian Subjunctive Clauses

Abstract

This paper aims to define the featural composition of the complementizers that introduce subjunctive complements in Istro-Romanian, and to identify the internal organization of the subjunctive clause in terms of subject positions, verb movement, clitic placement and constituent fronting. In a nutshell, the observation is that the complementizer neca replaces se within the syntactic pattern of Old Romanian; that is, a pattern that displays intra- and inter-language variation with respect to the distribution of complementizers within the subjunctive CP. Tests of word order also indicate intra-language variation in the parametric settings for clitic placement (either high or low), for the argumental subject position (either in Spec,TP, yielding SVO, or in Spec,vP, yielding VSO) and for constituent movement under discourse triggers (either scrambling or fronting to CP).

1. Overview

Speakers of Istro-Romanian are bilingual, having Croatian as their native language as well. In their Istro-Romanian grammar, there are two subjunctive complementizers: neca, which is productive—see (1)—and se, as in (2), which is unproductive.1 Neca is borrowed from Croatian, whereas se is inherited from Old Romanian. In root clauses, these complementizers head imperative surrogates—see (1a). In embedded clauses, they head adjuncts (mostly of purpose) or conditionals—see (1b,c). They also head selected clauses that qualify as direct objects to verbs or to other lexical categories—see (1d).

(1)	a.	Neca	ste	srečen.			(SF, p. 90)
		sbjv	be.sbjv.3sg	happy
		‘Let him be happy.’
	b.	Atunče	l-a	zdrenit	din mašuna fåra …
		then	it=have.3pl	taken.out	from stable out…
		neca	påscu	pâr la sera.			(K 1971, p. 192)
		sbjv	feed.3pl	until at evening
		‘Then they took it [the horse] out of the stable so that it would feed until the evening.’
	c.	Neca ştiu	care-s… (SF, 151)
		sbjv know.1sg who are…
		‘If I know who they are…’
	d.	Si-l’	zițe	neca-s	țevå	cumpare.	(SI, p. 32)
		and=him.dat.3	says	sbjv-refl.3	something	buy.3sg
		‘And he tells him to buy something.’

The two complementizers are functionally equivalent. This is attested in coordination structures as in (2): clause coordination signals equivalent syntactic status for their complementizers.

(2)	Ie-vo		tunțe	rogę	neca-l	lå̯ie		cu sire
	he=her.acc.f.3.sg		then	ask.pres.3sg	sbjv = him.acc.f.3.sg	take.3sg		with self
	si	se	nu	lu nițur	spure.		(SI, p. 33)
	and	sbjv	neg	dat no-one	tell.3sg
	‘Then hei asks her, to take himj with her and not to tell anyone.’

The replacement of infinitive complements with subjunctives (which is a Balkan Sprachbund property—Mišeska-Tomić 2006) is incomplete in Istro-Romanian, to the extent that infinitive complementation remains the default option (Dragomirescu and Nicolae 2020). For Old Romanian, Frâncu (2009) shows that, statistically, the replacement of infinitives with subjunctive clauses begins with verba voluntatis/iussives, and advances gradually to verba dicendi, then to modals and, finally, to aspectuals. In this paper, we notice that, following the same axis, the replacement of infinitives in Istro-Romanian remains optional and restricted to verbs that require irrealis interpretation, with no attestation of subjunctive complementation for aspectual verbs. From a formal perspective, this paper will show that: (i) the subjunctive complements preserve the CP structure seen in Old Romanian; (ii) the subjunctive clauses display parametric variations in the word order, clitic placement and, possibly, verb movement; and (iii) Croatian does not significantly impact the syntax of Istro-Romanian subjunctive clauses.

2. Background

2.1. Sources and Methodology

The data for this paper come from an Istro-Romanian corpus that includes texts recorded after the 1900s: texts from Traian Cantemir, recorded between 1932 and 1933; Pușcariu (1906) and Sârbu and Frățilă (1998), the latter recorded between 1982 and 1996, both in the north (Žejane area) and in the south (Šušnjevica area). Some examples are also taken from Kovačec (1971) and from the published volumes of Neiescu (2011).

The Istro-Romanian data will be compared with the Old Romanian data for the following reason: Istro-Romanians split from the Daco-Romanian population sometime between the 10th and 13th centuries (Caragiu Marioțeanu et al. 1977; Puşcariu et al. 1926, a.o.). Hence, the data provide an opportunity to assess the path of change in the two branches of Romanian since that time. In particular, we are interested in the options for parametric settings that were unstable in Old Romanian (i.e., intra- and/or inter-language variation) even by the 16th century but became stabilized in modern Daco-Romanian. The question is: how did the changes fare in Istro-Romanian?

2.2. Framework

The formal framework adopted in this paper assumes that clauses are built from the bottom up, according to a hierarchy organized over three fields (Chomsky 2008), as in (3a).

(3)

a.

CP < TP < vP

CP maps the discourse and pragmatic features, together with the clause typing features. TP is the field for inflectional features, comprising agreement and TAM (tense, aspect, mood), while vP maps the argument structure of the verb (basically, the thematic roles). Each field can be further split. Thus, roughly, vP maps the theta-roles over several projections that license the direct object (in VP), the indirect object (in vP) and the subject (in VoiceP), as in (3b).

(3)

b.

VoiceP < vP < VP

For the TP field, we adopt the cartography in Cinque (1999), consisting of the TAM features distributed as in (3c). The AspP and ModP fields can be further split according to the type of modal or aspectual features involved.

(3)

c.

TPpast/fut < ModP < AspP < TPant < AspP < VoiceP

For the CP field, we adopt the cartography in Rizzi (1997), shown in (3d). Force maps the clause typing feature, and Fin maps the features that select a certain type of verb inflection (tense, modality/mood, agreement). Constituents with topic or contrastive focus readings may be fronted to the middle of this field.

(3)

d.

ForceP < TopP< FocP < FinP

In this framework, clausal negation in Old Romanian occurs at the field border between CP and TP (Hill and Alboiu 2016), as in (3e).

(3)

e.

CP < (NegP) < TP < vP

Word order is the main criterion for assessing the hierarchical organization of the subjunctive clause. In this respect, it has been established for indicative clauses that Istro-Romanian is predominantly SVO and has unstable options for clitic placement and verb movement (Dragomirescu and Nicolae 2018, 2020, 2021). SVO arises either from subject movement to Spec,TP (argumental position) or to Spec,TopP (non-argumental position). VSO, which is the predominant word order in Old Romanian and modern Daco-Romanian, is not excluded either. VSO is typical for Balkan languages and arises from a derivation in which the subject remains in situ (Spec,VoiceP), whereas the verb moves to the TP field.

In this paper, subjunctive clauses will be analyzed according to the framework presented, with the understanding that their classification as such depends on the syntactic features, not on the morphology of their verbs. From this perspective, there is no difference between subjunctive and infinitive clauses as long as they occur in the same syntactic position (i.e., selected by a matrix verb as its complement). That is, they both have anaphoric tense and may encode irrealis modality. The anaphoric tense is encoded as a [-finite] feature (the term non-finiteness will refer to this feature), whereas modality is encoded as [-realis] in the C system.

3. Morphology

The morpho-syntactic analysis of Istro-Romanian subjunctives relies on a comparison with Old Romanian as a criterion for determining the direction of language change. In Old Romanian, the subjunctive forms are parasitic on the present indicative forms in the first and second persons, but they generally remain distinct in the third person. Istro-Romanian ceased to make that distinction, as the difference between present indicative and subjunctive forms was neutralized in favor of the indicative forms (Caragiu Marioțeanu et al. 1977, p. 225). Thus, the subjunctive verb paradigm, shown in

Table 1. Verb inflection.

	Indicative	Subjuctive
3sg	(ĭe) vire	(ĭe) neca vire //        (ĭe) se vire
	he comes	he sbjv come	he sbjv come
3pl	(ĭel’) pótu	(ĭel’) neca potu//        (ĭel’) se pot
	they can	they sbjv can	they sbjv can

, is systematically identical to the present indicative, irrespective of the subjunctive marker (which may be either neca or se).

One exception to this regularization concerns the verb ‘be’, which retains special subjunctive morphology for all persons, as was also the case in Old Romanian.

Despite the morphological identity between indicative and subjunctive verbs, the clauses they derive remain distinct due to different clause-typing devices. Indicative clauses have no complementizers in the root context and a che complementizer in the subordinate context, whereas subjunctive clauses have the complementizer se (Lat. si ‘if’) or neca (Croat. neka ‘let/may/sbjv’) in any syntactic context (Caragiu Marioţeanu 1975, p. 205; Kovačec 1971, pp. 123, 150; Hurren [1981] 1999, p. 105; Geană 2017, p. 210). Notably, che is also possible in combination with neca/se when the clause is selected, as shown in this paper. The complementizer se is inherited from Old Romanian and, with some phonetic variations, occurs in other Romanian languages in similar contexts (see Alboiu and Hill 2023). On the other hand, neca is idiosyncratic to Istro-Romanian, being borrowed from Croatian (Corbeanu and Geană 2022).

Subjunctive complementizers constitute one of the Balkan Sprachbund properties, so it is also present in Balkan Slavic, including Croatian, with which Istro-Romanian is in language contact, in a situation of bilingualism.2 The texts of our corpus display free alternation between se and neca, as it was shown in (2). The difference between the two complementizers is one of frequency not of function: neca is much more productive than se (Corbeanu and Geană 2022).

Economy conditions in grammar avoid redundancies and free optionality. As such, the existence of two syntactically identical subjunctive complementizers in the same language requires a closer look. The analysis proposed in this paper is that the borrowing of neca arises as a replacement for se, as the latter is also used as a conditional complementizer, and this function became the preferred option for se over its subjunctive one. The analysis has to establish to what extent this replacement impacts the syntax of the subjunctive CP.

4. Old Romanian să

The oldest preserved texts written in Daco-Romanian date from the 16th century. The group of Istro-Romanian speakers had split from the speakers of Daco-Romanian at least two centuries before that (Puşcariu et al. 1926). Hence, whatever changes we see in the grammar of 16th-century Old Romanian must have been at less advanced stages at the time of the split.

When it comes to subjunctive clauses, Old Romanian grammar displays contexts that indicate five stages of reanalysis for the complementizer să (=IR se), as shown in the remainder of this section, by adopting the analysis developed in Hill and Alboiu (2016, chp. 8). Needless to say, these stages appear concurrently in texts (mostly inter-language variations), and the free alternation was gradually reduced to dominant tendencies.

Just for clarification, philologists argue that Old Romanian să has a different etymology in conditional versus subjunctive clauses, i.e., Lat. si-conditional versus Lat. adv. sic or Lat. verb sit (see Sava 2012 for overview and references). While we respect these findings, we also consider that this is irrelevant to the learners from the 10th century onward (i.e., since groups of Romanian speakers started to immigrate into the Balkan Peninsula). That is, children acquire language only based on the available input data, and the data around them provided a homophonous să with the same syntactic status (i.e., complementizer). The children had to figure out what features this complementizer checked (those in Force or Fin, with or without operators).

The information available in the 16th century for language acquisition was the following: a complementizer să (=IR se) introduces mostly imperative, conditional and adjunct clauses and, less frequently, complement (selected) clauses. In these constructions, să is not constrained to pairing with subjunctive verb forms but can also introduce clauses that have indicative and conditional verb forms. Imperative, conditional and adjunct clauses that display subjunctive verb forms have two common properties: an irrealis modality (mapped to Fin) and a clause typing operator (mapped to Force); să could check them both. This situation is the starting point for the diachronic assessment of the complementizer in Hill and Alboiu (2016, chp. 8), so it is considered here as Stage 1. Adopting a development by stages, we follow their analysis, which argues that:

In Stage 1, să merges in Fin to check [-realis] and moves to Force to check the clause-typing operator (be it imperative, conditional or adverbial).

• Stage 1
  Force/OP < să < Fin/[_-realis] <să> < TP

Notably, the checking mechanism in Stage 1 was unstable in the 16th century, as the data indicate intra- and inter-language variation: in the same text, and in the same type of clauses, să could also remain in Fin, in which case the operator is checked through long-distance Agree from Fin; hence, Stage 2. The telling sign was the possibility of fronting constituents to Topic, above să, in imperative and conditional clauses (i.e., ForceP < TopP < FocP < Fin- să)

• Stage 2
  Force/OP (să) < Fin/[_-realis] -să < TP

Less than a century later, să was stripped of the operator properties and was reanalyzed as exclusively a Fin complementizer, i.e., as irrealis only, by being left in situ. This reanalysis triggers the use of concurrent items that check the clause typing operator, so complex complementizers occur, such as cum să, ca să in adjunct clauses of purpose or consequence, and de să in conditionals, that is, Stage 3:3

• Stage 3
  Force/OP-cum < Fin/[_-realis] -să < TP

By the 17th century, the bulk of clauses headed by să was the subjunctive complements selected by verba voluntatis and iussives (Frâncu 2009). This indicates a specialization of să for irrealis Fin and its independence from Force features. However, the subjunctive Fin also has a [-finite] feature (on a par with the replaced infinitivals), while să had been used with [+finite] Fin in root and adjunct clauses. Hence, the ambiguous clues for the finiteness of să are resolved in Stage 4 by splitting Fin so that să checks only the irrealis feature, unambiguously associated with it, whereas another complementizer checks [-finite]. Thus, for a relatively short time, complex complementizers, such as de să or ca să head subjunctive complements in contexts with or without obligatory control, that is, Stage 4.4

• Stage 4
  (Force) < Fin1-de < Fin2-să -< TP

Eventually, in Stage 5, să is reanalyzed as [-finite] as well (by moving să from Fin2 to Fin1). Fin remerges, and its exclusivity for subjunctive clauses is stabilized, leading to the extension of the subjunctive complementation to all the classes of relevant verbs and to the complete replacement of the infinitive complements in the standard register.5 That is:

• Stage 5
  (Force) < Fin- să < TP

This summary of the development of să subjunctives in Old Romanian provides a starting point for assessing the status of se in Istro-Romanian in this paper. More precisely, this indicates that at the time of the dialectal split, the use of să in Old Romanian was very unstable, preceding Stage 1 above, and the direction of reanalysis for să is expected to display feature stripping and specialization in the sense of systematic use for checking only a certain feature, as it is attested for Old Romanian. The specialization brought stabilization in Daco-Romanian (i.e., no more placement variations), so the distribution of să changed from being an operator in Force plus modal and finite to being just a non-finite complementizer. Thus, the grammar eliminated the use of să in conditional clauses and/or with verbal mood forms other than the subjunctive.

5. Istro-Romanian se

Within 100 pages of Cantemir (1959), there were 92 tokens of the complementizer se distributed as follows: 68 occurrences as a conditional complementizer; 13 occurrences as a complementizer in adjunct clauses (mostly of purpose), and 11 occurrences as a complementizer of selected clauses, as subjunctive complements. This distribution indicates that Stage 1 is dominating the analysis of se, that is, it is strongly associated with the clause typing operator in Force, while its use in complement clauses is minimal. We did not find evidence for Stage 2 in the Istro-Romanian data. That is, there were no root clauses (imperatives) headed by se that would display constituents higher than se in TopP or FocP. Hence, se is systematically in Force.

The tendency of analyzing se as Force rather than just Fin extends to selected clauses. This is attested by its cooccurrence with che ‘that’, as in (4a,b). Che ‘that’ is the non-operator complementizer (in declaratives) that merges in Force (equivalent to Romance che, que). In Istro-Romanian, it may also appear in Fin in the doubly filled COMP of clauses with indicative verb forms. For example, Corbeanu and Geană (2022) provide data with wh-phrases (in Spec,FocP), where, under selection, Istro-Romanian che may occur either in Force, above the wh-phrase, or in Fin, lower than the wh-phrase. Crucially, che may also cooccur with se in subjunctive complements with subject obviation (hence, ForcePs), and the result is a flipping order, as in (4a,b).

(4)	a.	Verit-a		rumunu		ca și voi	și	zis-a
		come-has		Romanian		like too you.pl	and	told-has
		se	che	va	ii	cu ie.			(SF, p. 63)
		sbjv	that	will	go.3sg	with him
		‘This Romanianj came just like you and said hek should go with himi.’
	b.	Cl’ęme	gospodåru	che se	rę	fi	bur	scula-se	(DDI-I, p. 38)
		called	master.the	that sbjv	would	be.inf	good	wake.inf-refl.3
		‘He called to his master that it would be good for him to wake up.’

The flipping order means that either complementizer can check the features of Force or of Fin, taking turns. It follows that se is reanalyzed as a possible Fin complementizer, not exclusively Force, as in root and adjunct clauses, but not exclusively Fin either. The Fin reanalysis of se provides evidence for Stage 3: when in Force (4b), che is a clue for the learner that se may be independent of Force in complement clauses; that is, se can be stripped of the clause-typing operator feature and left with the irrealis feature in Fin. This is the Old Romanian pattern. On the other hand, the se che sequence does not occur in Old Romanian. Old Romanian does not use ca if să is in Force, because ca in Fin1 would block the movement of să from Fin2 to Force. Thus, se che is another confirmation that Istro-Romanian continues to treat se as merged directly in Force, so the status of this complementizer is not stabilized at Fin.6

Furthermore, when we look at the classes of verbs that may select se-subjunctives, we notice that all of them require an irrealis Fin, as shown again in (5). There is no evidence that se subjunctives reached the aspectual verbs, which would require a realis interpretation. Aspectuals select only infinitive clauses in Istro-Romanian.

(5)	Cårle	va	pre lume-mnå,	trebe se	segav	fi̯ie.	(SI, p. 44)
	who	will	on world=go.inf	must sbjv	smart	be.sbjv.3sg
	‘Whoever will travel the world has to be smart’

Against this background, we must conclude that the reanalysis of se followed the same direction as Old Romanian să but has not touched all the stages and has not finished the replacement of the infinitive competition. In fact, the distribution of se in 20th century Istro-Romanian displays a fluctuation of its featural make up comparable to the 16th–17th-century Old Romanian să, being either directly merged in Force or directly merged in Fin2, with or without movement to Fin1 and Force.

In sum, the clues for se as a subjunctive Fin complementizer were and are weak in the primary linguistic data. From this perspective, the incomplete replacement of the infinitive with the subjunctive receives a language-internal motivation (in addition to language contact or language isolation): Fin-se was not a strong enough competition for the Fin checked by an infinitive verb in selected clauses.

6. Istro-Romanian neca

The reanalysis of se proceeded in a bilingual situation, where the dominant language also uses a syntactically similar complementizer, i.e., Croatian neka, whose features and distribution resemble those of se (i.e., it serves for checking the features of Force/Fin with clause typing operators and irrealis modality). This similarity is unexceptional within the Balkan Sprachbund, and so is the possibility of transferring this complementizer to Istro-Romanian. (Vrzić and Doričić 2014, p. 110) show that neca is a free morpheme that qualifies as a complementizer in Istro-Romanian.7

The transfer of the complementizer must be motivated by more than bilingual opportunity. In this particular set up, the use of se as a complementizer in selected subjunctives remained unproductive for centuries (compared with its use in root clauses), whereas neka is productive in Croatian subjunctive complements. Neka is transplanted in all the environments compatible with se, since it is a clause-typing operator as well as a marker of irrealis modality (Pavešić and Finka 1967–1976, p. 701). Thus, in Istro-Romanian, neca can introduces imperative, conditional and adjunct clauses, and it is compatible with the subjunctive mood in irrealis contexts (see examples in Corbeanu 2023).

The data indicate adjustments of neca to the Istro-Romanian syntax. First, there is no evidence of person restriction. In Croatian, neka can be used only in the first person (expressing an exhortation for the joint completion of an action in which the speaker will participate) and the third person (expressing permission, tolerance, concession, request), making it the most restrictive modal particle (Kramer 1986, p. 74). In Istro-Romanian, it is generalized over the person paradigm, and it is found with verbs inflected for the second person, as in (6).

(6)	Ti-ŭam	zis	neca	ștepți.	(TC, p. 17)
	you.2sg.dat=aux	say.pple	sbjv	wait.2sg
	‘I told you to wait.’

Second, the use of neca with selected subjunctive clauses is extended to constructions with obligatory control in Istro-Romanian, as in (7), whereas obligatory control allows only infinitive complementation (no neka) in Croatian (Mišeska-Tomić 2002–2003, pp. 355–56).

Furthermore, the features of neca are readjusted or stripped in selected clauses: In root and adjunct clauses, this complementizer selects finite TPs. However, subjunctive complements are non-finite, which requires the recognition that neca is compatible with [-finite] Fin. This is similar to the reanalysis of să in Old Romanian (Stage 4): as a clue for the learner that să is compatible with the subjunctive Fin (despite the fact that subjunctive Fin has a [−finite] value), Fin was split so that a different complementizer checks [-finite], while să checks only [-realis]. This pattern also applies to neca, as in (7), where che is used as a prop to solve the ambiguity (i.e., that neca here is in Fin not in Force).

(7)	Ie-l’	ganę	che neca-l’	dåie
	they.nom=him.dat	tell.ps.3.pl	that sbjv=him.dat	give.3sg
	corner lir.			(TC, p. 84)
	forty pounds
	‘They told him to give him forty pounds.’

In (7), obligatory control entails that the subjunctive complement does not project a ForceP, so only one analysis is available for the sequence che neca, that is, a split Fin: che checks [-finite], whereas neca checks [-realis]. As predictable, neca is concurrently reanalyzed from [+finite] to being underspecified for finiteness (it can check Fin with either + or [-finite]), so Fin remerges and neca checks it with no further props, as seen in (6) and further in (8).

(8)	Domnu	fåțe	neca	vire	åpa	ăntru.	(SI, p. 26)
	God.def	make.3sg.	sbjv	come.3sg	water.def	inside
	‘God makes the water come inside.’

The single occurrence of neca with subjunctive complements, as in (8), is the most productive option in the corpus.

Nevertheless, the analysis of neca as a Fin complementizer in subjunctive complements is not systematic. Hence, examples as in (9) are still found, where neca is a Force complementizer preceding the wh-element in Spec, FocP.

(9)	A	lui	zapisęit	neca če-m	då
	has cl.3sg.dat		wrote	sbjv what-me.1sg.dat give.2sg
	‘He wrote to him what you would give to me’ (TC, p. 143)

Formally, the replacement of se with neca has no syntactic impact on the properties of subjunctive CP, and the rate of replacing infinitive complements with subjunctives has not increased. The intra-language and inter-language variation in the featural makeup of neca is similar to what was observed for se and it is still strong. However, one tendency transpires: while neca replaces se in subjunctive embedded clauses, se remains strong as a conditional complementizer. Thus, se is not eliminated from the grammar but reassigned to a particular syntactic environment where it is systematically in Force.

7. Clause Structure

This section focuses on the internal structure of subjunctive complements. The main criterion for assessment is the word order, insofar as it indicates the hierarchical structure conforming to (3). No discrimination is made between clauses headed by se and neca complementizers since the derivational pattern is the same.

7.1. Clitic Pronouns

The verbs of subjunctive clauses are inflected for phi-features (number and person). Since phi-features are a property of T (the [phi] set is transferred from C-to-T; Chomsky 2008), it is expected that the verb is moved to the TP field. In subjunctive clauses, there is no clear evidence for the degree of verb movement within the TP field: constructions as in (10) suggest V-to-T, because of the obligatory preverbal clitic adjacency (10a), or at least V-to-Asp, since the verb may occur higher than subjects in VoiceP, as in (10b) below.

(10)	a.	Rugat	lu Martin	neca	l’i	lase	chia.	(TC, p. 18)
		asked	dat Martin	sbjv	him.dat.3sg	leave.3sg	key.def
		‘He asked Martin to leave him the key.’
	b.	Neca	vire	cåşu	γust.			(SF, p. 66)
		sbjv	become.3sg	cheese.def	dense
		‘The cheese should become dense.’

When it comes to clitic pronouns, their surface location varies, as shown in (11).

(11)	a.	Se	nu-l	cadu			(SF, p. 26)
		sbjv	neg=him.acc.3sg	fall.3sg
		‘Lest it should fall on him.’
	b.	Rogę	neca-l	țevå	neputę.		(SI, p. 35)
		ask.3sg	sbjv=him.acc.3sg	something	suggest.3sg
		‘He asks him to suggest something to him.’
	c.	Lui	a fost	vol’a	neca	ănsurå-se.	(TC, p. 60)
		he.dat	aux be.pple	wish	sbjv	marry.inf=refl.3sg
		‘He wanted to get married.’

In (11a), the clitic is T-oriented: the negation marks the clause border between Fin and TP, and the clitic is lower than the negation but higher than the verb. This clitic placement is unexceptional for Romanian languages (and for most Romance languages). On the other hand, in (11b), the clitic is separated from the verb by the scrambled DP țevå and takes as its host the complementizer on its left. So, this seems to be a C-oriented clitic. C-oriented clitics generally go hand in hand with Wackernagel law, which is not observed in Istro-Romanian, but it is in Croatian (Mišeska-Tomić 1996; Stjepanović 1998). Instances of clitics at C have been signaled in various contexts in Istro-Romanian (see data in Dragomirescu and Nicolae 2020), and the subjunctive clauses make no exception. Finally, in (11c), the clitic is lower than the inflected verb, which indicates a location nearing vP. Ledgeway (2018) and Dragomirescu and Nicolae (2020) point out the exploitation of this location in Old Romanian, as well as in other types of Istro-Romanian clauses. In the framework of (3c), this would be T_anterior: since clitic pronouns are agreement markers in Romanian and phi-features are mapped to T, nothing can theoretically prevent the spellout of clitics in either of the T heads or even in both of them (see the multiple clitic spellout in Ledgeway 2018).8

In sum, the default placement for clitic pronouns in Istro-Romanian is at T and preverbal. Post-verbal clitics are also productive. This is also a particularity of Old Romanian (Hill and Alboiu 2016; Nicolae 2019). The least productive is the placement of clitics at C, which we attribute to the unsystematic negative transfer from Croatian. Crucially, the variation in clitic placement indicates instability in the setting of this micro-parameter even with the same speaker.

7.2. Subjects

Istro-Romanian is a null subject language across the board, including the subjunctive complements, as in (12), with subject obviation for non-lexical subjects.

(12)	Mai bire	ra	fi	neca	mejeț	din	a me
	more well	aux.cond.3sg	be	sbjv	go.2pl	from	pos.1sg
	iåpa.						(TC, p. 123)
	mare
	‘It would be better for you to go after my mare.’

Various clause types display genuine preverbal subjects in Istro-Romanian (Spec,TP; Dragomirescu and Nicolae 2021). In subjunctive clauses, both SV and VS are attested, as shown in (13a) and (13b), respectively.

(13)	a.	Iel’i	måi vişe	a nostre	jeiånske cânte		zahtevuis
		they	more	our	Žejane songs		ask.pres.3pl
		neca	noi	cântåm.				(SF, p. 301)
		sbjv	we	sing.1pl
		‘They ask more that we sang our local songs.’
	b.	Zițe	Domnul	neca	vire	̯ie	fåre.	(SI, p. 37)
		say.3sg	God	sbjv	come.3sg	he	outside
		‘God tells him to come outside.’

Again, intra-language variation is at work, this time for the word order settings (e.g., the speaker of (13b) also produces subjunctives with SVO). Free alternation in the argumental position for subjects (i.e., either Spec,VoiceP or Spec,TP) is unproductive in the Old Romanian texts of the 16th century. By that time, the Balkan VSO was generalized (Alboiu and Hill 2017). However, the existence of the examples with free alternation for subjects indicates the tail end of a Romance parametric setting that must have been productive in the previous centuries, possibly at the time of the Istro-Romanian emigration. From this perspective, Istro-Romanian has preserved the unstable parameter for the subject position up to now, while Old Romanian and the other Romanian languages generalized VSO.

7.3. Constituent Movement

Istro-Romanian syntax intensively maps discourse features, on a par with other Balkan languages and old Romance languages. Discourse features trigger constituent movement to designated slots, either within CP or within TP. Constituent movement under discourse triggers in Istro-Romanian is illustrated in (14): In (14a), the direct object is postverbal, and the reading is neutral, whereas in (14b), the direct object is moved to a preverbal position, and the reading places it in the spotlight.

(14)	a.	Rogę	neca-l	neputę	țevå.			(SI, p. 34)
		ask.3sg	sbjv=him.acc.3sg	suggest.3sg	something
		‘He asks him to suggest something to him.’
	b.	Rogę	neca-l	țevå	neputę.			(SI, p. 35)
		ask.3sg	sbjv=him.acc.3sg	something	suggest3sg
		‘He asks him to suggest something to him.’
	c.	Grofu	urdineit-a	lu pechi	neca	fornu	zarescu.	(TC, p. 53)
		count.def	ordered=has	dat bakers	sbjv	oven	heat.3pl
		‘The count ordered the bakers to turn on the oven.’

Constituent movement, as in (14b,c), targets a discourse-motivated position within TP, amounting to scrambling. There is no evidence for an OV parametric setting in Istro-Romanian (i.e., leaving the direct object postverbal does not trigger ungrammaticality); this is a discourse-oriented language, similar to the rest of Balkan languages. As for the syntactic structure, a subjunctive clause, as in (14c), with obligatory control, is a FinP (versus ForceP), so the complementizer neca is necessarily in Fin. Hence, the moved constituent is lower than Fin, within TP. Scrambling, as in (14b,c), is productive in subjunctive clauses in the corpus.

However, movement to the CP field is also possible, as in (15), produced by the same speaker who produced (14a,b).

(15)	Trebe	se	segav	fiĭe!	(SI, p. 44)
	must.3sg	sbjv	smart	be.sbjv.3sg
	‘It is necessary that he be smart.’

In (15), the adjective has a contrastive focus reading that signals constituent movement to FocP, within CP, not scrambling. This is plausible because the subjunctive complement displays subject obviation, so it is a ForceP, and se can be in Force. Hence, the adjective phrase can target Spec,FocP.

Again, there is intra-language variation, this time concerning the field for mapping the discourse features. Scrambling is attested in Old Romanian texts as a tail end of a type of grammar that also had non-clitic auxiliary verbs and preverbal (Spec,TP) subjects (Alboiu and Hill 2017). These properties were unproductive in 16th-century texts, as the CP field was the most exploited option for discourse-motivated movement. The productivity of scrambling in Istro-Romanian is another sign that this is a very conservative language, since the source of scrambling seems to be the pre-attested Old Romanian. On the other hand, Croatian also displays the possibility of scrambling, which raises the question of a language contact transfer to Istro-Romanian (Dragomirescu and Nicolae 2021). Nevertheless, the scrambled constituents in Croatian are mostly adverbs, whereas in Old Romanian, the scrambling of direct objects is as productive as the adverbs, and so it is in Istro-Croatian.

Summing up the information so far, the structure of a subjunctive complement in Istro-Romanian has the representations in (16a) and (16b).

(16)	a. Subject obviation
	b. Obligatory control

In (16a), the complementizers merge in Fin and move to Force. The subject merges in Spec,VoiceP, where it may remain at Spell Out, or it moves to Spec, TP. Notably, these are both argumental positions, so movement applies only if licensing is not available in Spec, VoiceP. It seems that sometimes it is and sometimes it is not. If the DP is licensed in Spec, VoiceP for theta-role and Case, then the subject remains in situ and the parametric setting is VS, i.e., Balkan. If the DP checks the theta-role in Spec,VoiceP but has to move to Spec,TP to receive Nominative Case, then the parametric setting is SV, i.e., Romance.9 Similar alternatives seem to have been available in Old Romanian previously to the attested 16th-century texts (Alboiu and Hill 2017; Nicolae 2019).

7.4. Taking Stock

Verb movement seems to be unstable, judging by the linear order in Istro-Romanian subjunctives. That is why two possible landing sites are assigned in (16a,b): T, when no scrambling takes place, or Asp, in the presence of scrambling. However, it is remarkable that we could find no example that would contain a preverbal subject in the presence of scrambling in subjunctive clauses (negative evidence). Hence, it is possible that Spec, TP is used alternatively as either argumental, for subjects, or non-argumental, for scrambling (Alboiu 2002 argues that, in modern Daco-Romanian, Spec,TP is available as a non-argumental position for constituent movement under discourse triggers). In that case, verb movement may be stabilized to T in subjunctives; i.e., scrambling to Spec,TP situates the constituent higher than the verb in T.

The location of clitic pronouns also displays variability, occurring at C, T and T-anterior in (16a,b). Clitics at C is a parametric setting found in Balkan Slavic, where it goes by the Wackernagel law. This is not the case in Istro-Romanian, where clitics are rather distributed in lower positions. The most productive option in subjunctive clauses is for clitics at T, generally pre-verbally, but post-verbal clitics are also found (shown in T-anterior in (16a,b)).

Constituent movement for discourse purposes is productive, especially by scrambling, rather than by fronting to the CP area. This is also attested in Old Romanian, as a tail end of what may have been a productive construction sometime before the 16th century (together with non-clitic auxiliaries; Alboiu and Hill 2017; Nicolae 2019). The disappearance of scrambling from Old Romanian (and from Daco-Romanian) went hand in hand with the increased productivity of constituent fronting to the CP area (Alboiu and Hill 2017). Scrambling under discourse triggers appears well preserved in Istro-Romanian, not only in root declaratives and the presence of non-clitic auxiliaries (Dragomirescu and Nicolae 2020, 2021) but also in selected clauses, such as the subjunctives discussed here. The presence of scrambling in Croatian did not challenge this parametric setting of Istro-Romanian.

In sum, Istro-Romanian subjunctive complements display unstable word order options for all the criteria of this assessment: clitic placement, the argument position for subjects, constituent movement under discourse triggers and, possibly, verb movement. The variation is seen at intra- and inter-linguistic levels, amounting to generalized instability in parametric settings.

8. Conclusions

Subjunctive clauses in Istro-Romanian display a distribution consistent with their counterpart in other Romanian languages: they are imperative surrogates in root clauses, direct complements when selected, and adverbial adjuncts when not selected. There are two complementizers introducing the subjunctive clause (neca and se), both displaying a similar variation in their syntactic behavior: they are compatible only with irrealis modality, they may directly merge in either Fin or Force, and they may or may not be reanalyzed as [-finite]. In addition, the same complementizers head conditional clauses, where the verb is not in a subjunctive mood. The variation in the feature set associated with neca/se yields derivations whose CP is not always equivalent to infinitive complements, so the replacement of infinitives with subjunctives is possible but not systematic.

When it comes to the internal structure of subjunctive clauses, the general observation is that, in contrast to other Romanian languages, Istro-Romanian displays fluid parametric settings: there is free alternation between genuine SVO and VSO, the clitic pronouns may be spelled out either at C, T-pr/future or T-anterior, while the degree of verb movement within TP is unclear. This is a general observation arising from the available written corpus.

In terms of diachronic change, the impact of Croatian is minimal for the structure of subjunctive clauses: lexical borrowing of neca; some clitics at C; and convergence for scrambling (Dragomirescu and Nicolae 2021). Most properties of these constructions are inherited from Old Romanian. Notably, the instability of parametric settings noticed in Istro-Romanian matches the observations for 16th-century Old Romanian, especially in constructions that illustrate the tail end of a grammar with more Romance characteristics that preceded the preserved documents: the disappearing genuine SVO, the predominant analysis of să as a clause-typing operator and non-specialization for subjunctive clauses, the extended use of infinitive complements (i.e., with subject obviation) and constituent scrambling for discourse effects. In Old Romanian, these properties were replaced through balkanization by the 16th century: genuine VSO, complete reanalysis of să as an irrealis Fin complementizer, complete replacement of infinitive complements with subjunctives and constituent movement to CP for discourse effects. In other words, these parametric settings for Istro-Romanian remained unchanged for more than 600 years, with the addition of the occasional clitic at C as a transfer from Croatian.