Romance and Croatian in Contact: Non-Clitic Auxiliaries in Istro-Romanian

Abstract

This paper focuses on Istro-Romanian and argues that the TAM auxiliaries of this variety are not morphophonological clitics. This analysis is supported by the existence of several empirical phenomena (auxiliary-licensed VP-ellipsis, scrambling, and interpolation), some not found in modern Romance, others very rare in modern Romance. This property of Istro-Romanian auxiliary verbs accounts, in conjunction with other features of this variety (e.g., the availability of C-oriented and I-oriented pronominal clitics), for the massive variation in the word order of pronominal clitics, auxiliaries, and the lexical verb found in the Istro-Romanian sentential core. An endangered Romance variety spoken in Istria and in the diaspora, historically related to (Daco-)Romanian, Istro-Romanian has been in contact with Croatian since the settlement of Istro-Romanians in the Istrian peninsula. As some of the Istro-Romanian features and phenomena are found both in Croatian and in old Romanian, it appears that contact with Croatian acts as a catalyst of structural convergence engendering the retention of an archaic property of Istro-Romanian auxiliaries: a lower position on the grammaticalization cline, closer to the full word status of their etyma.

1. Introduction

1.1. Background

Istro-Romanian is an endangered Romance variety spoken in Istria, Croatia, with about 200 speakers at present (cf. Filipi 2002; Vrzić and Doričić 2014), plus émigré speakers in the New York area (Maiden 2016, p. 91). Traditionally considered to be one of the sub-Danubian “historical dialects” of Romanian (alongside Aromanian and Megleno-Romanian), Istro-Romanian1 displays two tendencies in its syntax: (i) on the one hand, Istro-Romanian displays a series of Romance archaic features, also found in old Romanian (which have been taken to mirror the shape of the language at the moment of the dialectal split, presumed to have finished around the 14th century); (ii) on the other, we observe the influence on the syntax of Istro-Romanian of Croatian2 (see Kovačec 1966, 1968) and, to a much more limited extent, of Northern Italian dialects, Croatian–Istro-Romanian bilingualism being the norm in the area. While a comprehensive grammar/syntax of Istro-Romanian is not available at present, there exist partial descriptions of this linguistic dialect (Popovici 1914; Pușcariu 1926; Coteanu 1957; Kovačec 1971, 1984, 1998; Caragiu Marioțeanu 1975; Sârbu and Frățilă 1998; Hurren [1981] 1999; Zegrean 2012; Giusti and Zegrean 2015; Maiden 2016) which offer significant details and insights into its structure. It is also important to mention that ‘old Romanian’ refers to the timespan comprised between the beginning of the 16th c. and the year 1780; the first written attestations of Romanian date back to the beginning of the 16th c.: Neacșu’s Letter (1521), the first non-translated text written in Romanian, and The Hurmuzaki Book of Psalms (1500/1510), the first translation of a foreign (Old Church Slavonic) text in Romanian. See Timotin (2016) for more details on the periodization of Romanian. Notably, the dialectal split took place prior to the earliest written attestations of Romanian.

1.2. Aim of the Paper and Methodology

The verbal nexus of Istro-Romanian is characterized by massive variation, displaying numerous distinct word order patterns of pronominal clitics, auxiliaries, and the lexical verb—all of them found across Romance, but not at the same time in the same variety. Furthermore, Istro-Romanian also shows instances of auxiliary-licensed verbal ellipsis (=VP-ellipsis), a phenomenon severely restricted in the Romance languages (see Nicolae 2019a: ch. IV for an overview). By using a standard minimalist generative grammar framework (see Section 1.3 below) and by working with the existing written corpora of Istro-Romanian3, we aim to demonstrate that the paradoxical properties briefly mentioned above (as well as many others, discussed throughout the paper) are best understood under the premises that Istro-Romanian auxiliaries are not clitics (see also Geană 2017). Therefore, in contrast to the auxiliaries of Daco-Romanian, Istro-Romanian auxiliaries are less grammaticalized (closer to full words on the Hopper and Traugott’s (2003) ‘grammaticalization cline’), and this morphophonological property has syntactic relevance. The non-clitic nature of Istro-Romanian auxiliaries will be shown to result from the combined effect of the conservative nature of this idiom (=retention of an archaic feature from an unattested stage of (proto-)Romanian) and language contact with Croatian4.

1.3. Conceptual Tools

In what follows, we adopt the current generative view on clause structure (1), according to which finite clauses are made up of three domains: the CP-domain, the IP(/TP)-domain and the vP-domain(/v-VoiceP)-domain. Research in this area has revealed that, cross-linguistically, any of the three domains can be targeted by V-raising/movement. It is well-known that these domains themselves are made up of a relevant number of projections—cf., for example, Rizzi’s (1997) split-CP analysis or Cinque’s (1999) cartographic approach; finer-grained structural details will be provided when necessary.

1.	CP	>	IP	>	vP
	(continental) Germanic V2		modern Romance		English
	old Romance V2

It is also necessary to introduce some minimal background on the phenomena relevant for our argumentation.

Consider first auxiliary-licensed VP-ellipsis; the main properties of this phenomenon5—which occurs rather rarely cross-linguistically (it productively occurs in English)—are the following:

• it is a type of elliptical construction which is dependent on the existence of a licenser (a TAM auxiliary or a modal auxiliary);
• it is insensitive to clause type (main/embedded clause) and other conditions (e.g., in contrast to ellipses such as gapping, it is not parasitic on coordination);
• there are two conditions which need to be met for the licensing of this type of ellipsis (see Thoms 2010; Authier 2012, i.a.):

  (i)

  lower verb movement of the lexical verb (movement below I);

  (ii)

  non-clitic auxiliary verbs.

Secondly, a minimal background on verb morphosyntax in (Daco-)Romanian is also necessary; this provides a starting point for understanding the Istro-Romanian data. Up to the present point, research on old and modern (Daco-)Romanian has established the following:

• modern Romanian auxiliaries are clitics (Avram 1999); the same characteristic carries over to old Romanian too (Nicolae 2019b);
• modern Romanian synthetic finite verbs are characterized by V-to-I movement (Dobrovie-Sorin 1994; Cornilescu 2000; Ledgeway 2015, i.a.);

  ○

  in a framework that assumes a finer-grained cartographic structure of the IP in the tradition of Cinque (1999) (distinguishing a Mood field, a Tense field, and an Aspect field), synthetic finite verbs raise to the top-most I-field, Mood (Nicolae 2015; Schifano 2018);

  ○

  in periphrastic constructions, both the auxiliary/auxiliaries and the lexical verb reside in the IP-domain (Alboiu and Motapanyane 2000; Nicolae 2015);

• V-to-C movement is restricted to imperatives, conditional imprecations, and gerunds in modern Romanian—and it always takes the shape of Rivero’s (1992, 1994) Balkan ‘Long Head Movement’ (=LHM) rule, by which it is the lexical verb, not the auxiliary that moves to C;
• although statistically V-to-I movement is the dominant option (Nicolae 2015), V-to-C movement is more widespread in old Romanian (also occurring in finite non-imperative clauses), and (English-type) low(er) verb movement (on the surface manifested as scrambling and interpolation) is also found in old Romanian finite clauses (Nicolae 2019b);
• pronominal clitics are I-oriented in both old and modern Romanian (cf. Săvescu Ciucivara 2011; Hill and Alboiu 2016, i.a.)6, and occupy dedicated projections at the edge of the IP-domain; in correlation with other syntactic diagnostics, the proclisis/enclisis alternation is relevant for assessing the level of verb movement (V-to-I/V-to-C).

Some terminological clarifications are also necessary.

• Interpolation represents the separation of the clitic pronoun from the finite verb by (an) intervening XP(s), while scrambling refers to the separation of the auxiliary verb from the lexical verb by (an) intervening XP(s). It has been shown that in the Romance languages multiple processes may lie behind the derivation of interpolation and scrambling; thus, while on the surface, each of these phenomena, in its own right, may appear to be a unified phenomenon, from a structural point of view, they are based on different processes (see also Poole 2007, p. 188, n. 1, who makes similar comments in the analysis of old Spanish). Hence, in what follows, we use the terms interpolation and scrambling to refer to the word order patterns described above; their structural analysis is assessed on the basis of syntactic diagnostics for each of the varieties examined here (Istro-Romanian, old Romanian, and Croatian).
• We use the label ‘Long Head Movement’ (LHM) (Rivero 1992, 1994) conventionally, to describe an empirical phenomenon specific (but not limited) to the Balkan languages: the displacement of a non-finite lexical verb across the auxiliary. Technical details on the derivation of LHM structures will be provided below.

In what follows, we set out to bring evidence for the main claim of the paper, namely that Istro-Romanian auxiliaries are not clitics. Viewed from a diachronic, Latin-to-Romance perspective, this property translates as the preservation of a more archaic status of Istro-Romanian auxiliaries: while these verbs become functional elements showing the typical phenomena of auxiliation—(morpho)phonological reduction (in comparison to their etyma), loss of thematic structure, loss of morphosyntactic features (see, i.a., Roberts and Roussou 2003, ch. 3; Ledgeway 2016, p. 768)—from the perspective of Hopper and Traugott’s (2003) grammaticalization cline, their status is that of full words (neither clitics, nor affixes). We argue that contact with Croatian acts as a catalyst of structural convergence engendering the retention of an archaic feature. As will be seen, the hypothesis that this is an archaic feature is supported by old Romanian data, as early old Romanian shows a few clear instances of auxiliary-licensed VP-ellipsis and related phenomena (interpolation and scrambling).

2. Three Phenomena Showing the Non-Clitic Nature of Istro-Romanian Auxiliaries

2.1. Auxiliary-Licensed VP-Ellipis

The availability of VP-ellipsis in Istro-Romanian is a phenomenon which has not been discussed in the literature (Zegrean 2012, p. 122 briefly mentions it). From a comparative Romance perspective, this feature of Istro-Romanian is unexpected: in contrast to the other Romance languages, Istro-Romanian auxiliary-licensed VP-ellipsis brings this idiom closer to languages such as English. The following data, extracted from the corpora7, show that the phenomenon is frequent. In what follows, we present the distribution of auxiliary-licensed VP-ellipsis in Istro-Romanian.

(i) The auxiliary licenses ellipsis in main clauses. Ellipsis is found with the perfect auxiliary (2a–e), the future auxiliary (2f), and the conditional auxiliary (2g,h). The symbol √ marks the ellipsis site; in situations in which the antecedent is not immediately obvious from the immediate context, it is represented in brackets < >.

2.	a.	‒ Ŭai	dus?
		have-aux.perf.2sg	bring-ptcp
		‒ Ŭam √. (TC, 8)
		have-aux.perf.1sg
		‘‒ Did you bring/take away?/‒ I did.’
	b.	‒ […] ŭaț	vezut	iuva-i	cea mare	hrușve?
		have-aux.perf.2pl	see-ptcp	where-is	the big	pear.tree
		‒ Ŭam √. (TC, 46)
		have-aux.perf.1pl
		‘– Did you see where the big pear tree is?/– We did.’
	c.	‒ Ce,	ŭai		pogodit?
		q-prt	have-aux.perf.2sg		guess-PTCP
		‒ Ŭam √,	cum ŭai			zis. (TC, 51)
		have-aux.perf.1sg	how have-aux.perf.2sg			say.ptcp
		‘– Did you guess?/– I did, as you said.’
	d.	‒ Ce,	ŭai		specit	tot?
		q-prt	have-aux.perf.2sg		grill-ptcp	everything
		‒ Ŭam √. (TC, 52)
		have-aux.perf.1sg
		‘– Why, did you grill everything?/– Yes, I did.’
		‒ Ai	tu	cea	facut
		have-aux.perf.2sg	you-sg	that	do-ptcp
		ce-ai	zis?
		what-have-aux.perf.2sg	say-ptcp
		‒ Am √! (LM, 138)
		have-aux.perf.1sg
		‘– Did you do what you said?/– I did!’
	e.	‒ Ai	verit,		frate?
		have-aux.perf.2sg	come-ptcp		brother-voc
		‒ Am√! (LM, 169)
		have-aux.perf.1sg
		‘‒ Did you come, brother?/‒ I did.”

	f.	‒ Grofu	av		urdineit	che io me	mores
		count	have-aux.perf.3sg		order-ptcp that I cl.refl.1sg		have.to
		ănsura c-alui (…)		fil’e,	e io	săm ănsurat;	nu
		marry with-his		daughter	and I	am marry-ptcp	not
		poc.		Ver √		te	tu,
		can-ind.pres.1sg		will-aux.fut.2sg		cl.acc.2sg	you-sg
		oucearule?
		shepherd-voc
		‒ Io	voi √,		io! (LM, 123)
		I	will-aux.fut.1sg		I
		‘‒ The count ordered that I have to marry his daughter, but I am married; I can’t. Will you, shepherd?/‒ I will.’

	g.	Ke	l’a		zís	–ręi̯
		that	cl.acc.3sg=have-aux.perf.3sg		say-ptcp	would-aux.cond.2sg
		te	tu	maritå	dupa míre?
		cl.refl.2sg	you-sg	marry-inf	after me
		i̯a	zíče:
		she	says
		‒ Reș √. (AK, 338)
		would-aux.cond.1sg
		‘and he said—would you marry me? She says: ‒ I would.’
	h.	‒ Rei		borebit	pocini?
		would-aux.cond.2sg		maybe	rest-inf
		‒ Reș √. (TC, 57)
		would-aux.cond.1sg
		‘– Would you rest?/– I would.’

(ii) The auxiliary licenses ellipsis in embedded clauses. In our corpus survey, we have identified the perfect auxiliary (3a) and the future auxiliary (3b) in these contexts.

3.	a.	‒ Cuheit-ŭai?
		cook-ptcp=have-aux.perf.2sg
		Ie zis-a	che a √,			ma
		he say-ptcp=have-aux.perf.3sg	that have-aux.perf.3sg			but
		c-a	zadurmit		și
		that=have-aux.perf.3sg	fell.asleep-ptcp		and
		nușcarle… (TC, 71)
		someone
		‘‒ Did you cook?/He said that he had, but he fell asleep and someone…’
	b.	Zis-au	cela tireru	che	va √
		say.ptcp=have-aux.perf.3sg	that young.def	that	will-aux.fut.3sg
		<însura> (LM, 137)
		marry
		‘The young one said that he would <get married>.’

(iii) Auxiliary-licensed VP-ellipsis is also available in negative clauses:

4.	a.	‒ Vezi,	tu-i		zadurmit.
		see-ind.pres.2sg	you-2sg=have-aux.perf.2sg		fall.asleep-ptcp
		‒ Nu	ŭam √. (TC, 36)
		neg	have-aux.perf.1sg
		‘‒ See, you fell asleep./‒ No, I did not.’
	b.	Ma ur	fost-a	ănsurat,	e	ur
		but one	be-ptcp=have-aux.perf.3sg	marry-ptcp	and	one
		n-av √. (LM, 77)
		not=have-aux.perf.3sg
		‘But one got married, and the other did not.’

(iv) In other cases (5), the auxiliary licenser is also followed by remnants:

5.	a.	‒ Cum	ŭai			sta nopte durmit?
		how	have-aux.perf.2sg			this night sleep-ptcp
		‒ Ŭam √		bire. (TC, 57)
		have-aux.perf.1sg		well
		‘‒ How did you sleep last night?/‒ I did <sleep> well.’
	b.	Se-i		tu		utat,
		if=have-aux.perf.2sg		you-2sg		forget-ptcp
		n-ŭam	√		io. (TC, 39)
		not=have-aux.perf.1sg			me
		‘Even if you have forgotten, I have not.’
	c.	Ali ia	cunoscut-a				ie.	Ie ia
		but she	recognize.ptcp=have-aux.perf.3sg				him	he her
		n-av √. (TC, 45)
		not=have-aux.perf.3sg
		‘She recognized him, but he did not <recognize> her.’

The features in (i)–(iv) show that the auxiliary-licensed VP-ellipsis in Istro-Romanian has the same properties as its English counterpart (see Section 1.3 above).

2.2. Discontiguous Verbal Clusters as Evidence for Low Verb Movement

In previous work, we have brought evidence in favour of the hypothesis that, in contrast to modern Daco-Romanian, Istro-Romanian shows numerous instances of low verb movement (Dragomirescu and Nicolae 2018). Low(er) verb movement is also found in old Romanian (Nicolae 2019b) (see also Section 3.2.1 below). The tools for diagnosing low(er) verb movement are the diagnostic adverbs for V-movement (Section 2.2.1) in Cinque’s (1999) hierarchy (see Schifano 2018 for an update and revised hierarchy) and other patterns of discontiguity in the verbal cluster—scrambling for periphrastic structures and interpolation for synthetic structures (Section 2.2.2).

The existence of low(er) verb movement is directly relevant to our argumentation: it is a property which directly correlates with the availability of auxiliary-licensed VP-ellipsis, which, in its turn, acts in correlation with non-clitic auxiliaries.

2.2.1. Scrambled and Interpolated Adverbs

The most significant argument for low verb movement is the position of Cinque’s diagnostic adverbs for V-movement. In particular, we observe that the lexical verb occurs to the right of the adverbs ‘well’, ‘already’, ‘then’ (6a-e), themselves preceded by the auxiliary; it is well known that these adverbs lexicalize various specifiers in Cinque’s (1999) hierarchy: ‘well’ marks the boundary between the IP-domain and the vP-domain, ‘already’ and ‘then’ are T-oriented adverbs. Other (circumstantial) adverbs (6f-g) may also occur in between the auxiliary and the lexical verb.

6.	a.	Åm			bíre	zis? (SF, 91)
		have-aux.perf.1sg			well	say-ptcp
		‘Did I say it well?’
	b.	io	n-åm				bíre-nțelés (SF, 168)
		I	not=have-aux.perf.1sg				well understand-ptcp
		‘I did not understand well.’
	c.	ke	i̯é s-a				vet”	opít (AK, 338)
		that	he cl.refl.3sg=have-aux.perf.3sg				already	get.drunk-ptcp
		‘that he already got drunk’
	d.	Iel’i	s-åu				atúnce	ustrașít (SF, 187)
		they	cl.refl.3pl=have-aux.perf.3pl				then	get.scared-ptcp
		‘and then they got scared’
	e.	vor			atúnče	cuvintå (SF, 158)
		will-aux.fut.3pl			then	speak-inf
		‘they will speak then’
	f.	si	i̯å		l-av				musåt
		and	she		cl.acc.m.3sg=have-aux.perf.3sg				beautifully
		zahvalít (SP, 13)
		thank-ptcp
		‘and she thanked them a lot’
	g.	Și	ie	che	s-a			chiaro	pričudit,
		and	he	that	cl.refl.3sg=have-aux.perf.3sg			very	wonder-ptcp
		che	svoini		n-a		șavo		vezut,
		that	never		not=have-aux.perf.3sg		like.this		see-ptcp
		ni	avzit		pre		tota	luma (TC, 7)
		nor	hear-ptcp		in		all	world
		‘And he was very astonished, because he had never seen something like this, anywhere in the world.’
	h.	io	n-åm				colę	fos (SF, 155)
		I	not=have-aux.perf.1sg				there	be-ptcp
		‘I wasn’t there.’

Moreover, in structures without auxiliaries, these adverbs occur in preverbal position; notice (7a) in particular, where the ‘well’-adverbial is trapped between the subject and the synthetic lexical verb, clearly indicating a low position of the lexical verb on the spine8.

7.	a.	ke	voi	bíre	știț (SF, 177)
		that	you-pl	well	know-ind.pres.2pl
		‘that you know well’
	b.	și	mușat		zehvales (TC, 11)
		and	beautifully		thank-ind.pres.1sg
		‘and I thank you a lot’

The position of the lexical verb with respect to adverbials is an indicator for a lower position of the lexical verb on the clausal spine9.

2.2.2. Other Patterns of Scrambling

The scrambling of several types of constituents in between the auxiliary and the lexical verb has a double significance: on the one hand, it is a supplementary argument for low verb movement; on the other hand, it shows that Istro-Romanian auxiliaries show a lower degree of grammaticalization being self-standing verbal elements, at least when compared to Daco-Romanian (Dragomirescu and Nicolae 2016, 2018). The constituents that can occur in between the auxiliary and the lexical verb are: the subject realized as a simple nominal phrase (8a) or as a complex one (8b), the prepositional object (8c), the temporal adverbial and the direct object (8d), the pronominal subject and the manner adverbial (8e), etc.

8.	a.	Cănd a	cesaru	dozneit (TC, 23)
		when have-aux.perf.3sg	emperor-nom	find.out-ptcp
		‘when the emperor found out’

	b.	Cănd	au	mul’era	lu	cesta	ŭom	verit
		when	have-aux.perf.3sg	wife	gen	this	man	come-ptcp
		cu	merinda (TI, 146)
		with	lunch
		‘When this man’s wife brought lunch’

	c.	Iel’		a		zehvalit
		they.nom		have-aux.perf.3pl		thank-ptcp
		și	s-a			cu ie	pozdravit (TC, 12)
		and	cl.refl.3pl=have.aux.perf.3pl			with her-acc	greet-ptcp
		‘They thanked her and greeted her.’
	d.	și	inche	n-ŭam		nigdar	ŭom
		and	yet	not=have-aux.perf.1sg		nowhere	human-acc
		vezut (TC, 26)
		see-ptcp
		‘and I did not see any human anywhere’
	e.	cum	ŭam		io		bire zis (TC, 45)
		as	have-aux.perf.1sg		I-nom		well say-ptcp
		‘as I said well’

Scrambling is also found in old Romanian, alongside interpolation (see Dragomirescu 2013; Nicolae 2019b, ch. 3). As shown in Nicolae (2019b), there is no evidence for Aux-to-C movement in old Romanian; therefore, strings of the form Aux—(scrambled) XP(s)—V (in conjunction with other properties of old Romanian) constitute convincing evidence in favour of a lower position of the verb on the clausal spine. By contrast, as argued in previous work (Dragomirescu and Nicolae 2020) and as will also be shown below, Aux-to-C movement is available in Istro-Romanian; from this perspective, scrambling is compatible in Istro-Romanian with two structural analyses (being an instance of parametric ambiguity in the sense of Clark and Roberts 1993)—one in which the Aux resides in the IP and another with Aux-to-C movement—hence, it only represents partial evidence for low verb movement. However, due to the particular word order of the scrambled elements, there are examples such as (8e) which clearly favour an analysis in which the verb occupies a lower position on the clausal spine. In this example, the nominative subject io (‘I’) precedes the vP-edge adverbial bine (‘well’), which in its turn precedes the lexical verb; the most plausible analysis takes the subject to be in the canonical Spec,TP position and the adverbial in its base position, thus with the lexical verb occupying a lower position (presumably on the edge of the vP-domain), irrespective of the position of the auxiliary.

To sum up, Istro-Romanian scrambling also provides indirect evidence for low verb movement.

2.3. Aux-to-C and Other Word Order Patterns with Pronominal Clitics and Auxiliaries

As extensively shown in Dragomirescu and Nicolae (2020), Istro-Romanian also shows instances of Aux-to-C movement, reflected by the Aux-Cl ordering. It is highly likely that Aux-to-C is, in part, an effect of language contact, as it is well known that Croatian displays Wackernagel cliticization (Čamdžić and Hudson 2007; Schütze 1994; Migdalski 2009). However, there are examples such as (9b) that show that Aux-to-C also occurs when the clause-initial position is occupied by another prosodically integrated constituent.

9.	a.	åu	se	dus	la ea (SF, 96)
		have-aux.perf.3pl	cl.refl.pass	take-ptcp	at her
		‘they were taken to her’
	b.	måi̯a	av-o		sprocâlnit (AK, 327)
		mother	have-aux.perf.3sg=cl.acc.f.3sg		curse-ptcp
		‘Her mother cursed her’

	c.	și	ŭam	le	dus	ăn iame (TC, 47)
		and	have-aux.perf.1sg	cl.acc.f.3pl	bring-ptcp	in pit
		‘and I brought them in the pit’

	d.	Veț-âl			cunóște?		(SF, 55)
		will-aux.fut.2pl=cl.acc.m.3sg			know-inf
		‘Will you know him?’
	e.	Se nu,	voi		te		ucide (TC, 90)
		if not	will-aux.fut.1sg		cl.acc.2sg		kill-inf
		‘If not, I will kill you’
	f.	voi		o	lå		cu mire (TC, 40)
		will-aux.fut.1sg		cl.acc.f.3sg	take-inf		with me
		‘I will take her with me’
	g.	iel’	vor-o			țere (SP, 8)
		they	will-aux.fut.3pl=cl.acc.f.3sg			ask-inf
		‘they will look for her’
	h.	Cine	va	o		porta? (TC, 47)
		who	will-aux.fut.3sg cl.acc.f.3sg			bring-inf
		‘Who will bring this?’

The fact that Aux-to-C in Istro-Romanian may be partially determined by Wackernagel’s law (either imported from Croatian as an effect of language contact, or preserved in Istro-Romanian from a medieval Romance phase in the form of the Tobler-Mussafia law10, “ultimately a Romance-specific reanalysis of the Wackernagel Law”, as characterized by Ledgeway 2012, p. 161) is in itself a supplementary piece of evidence in favour of the non-clitic nature of Istro-Romanian auxiliaries: auxiliaries are good candidates for occupying a clause-initial position; hence, they are not clitics.

Moreover, in clusters with pronominal clitics, auxiliaries, and lexical verbs, virtually every word order possibility is attested: besides Aux-Cl-V (9), we find Aux-V-Cl (10a), Cl-Aux-V (10b); Cl-V-Aux (10c); V-Cl-Aux (10d); V-Aux-Cl (10e). This flexibility shows that Istro-Romanian patterns both with Daco-Romanian and the Balkan languages more generally—in displaying LHM patterns (Rivero 1994) such as (10d) and (10e)—and with the (older stages of the) Western Romance languages11—through the availability of Aux-to-C movement (9), (10a). Thus, both the lexical verb and the auxiliary are equally good candidates for movement to C.

10.	a.	pac	av	cârstít-o (SF, 136)
		then	have-aux.perf.3pl	christen-ptcp=cl.acc.f.3sg
		‘then they christened her’
	b.	Cesaru	l-a			cl’emat	ăn case. (TC, 15)
		emperor	cl.acc.m.3sg=have-aux.perf.3sg			call-ptcp	in house
		‘The emperor called him in the house.’
	c.	Cum	vo	scapul’eit-a? (TC, 91)
		how	cl.acc.f.3sg	save-ptcp=have-aux.perf.3sg
		‘How did he save her?’
	d.	låt-a			i̯él’	båstele	dispre åsiri=
		take-ptcp=have-aux.perf.3pl		t	hey	saddle-pl	out.of donkey-pl
		si	legat	l ’a		de marún (AK, 331)
		and	tie-ptcp	cl.acc.m.3pl- have.aux.perf.3pl			of chestnut.tree
		‘they took the saddles of the donkeys and they tied them up to the chestnut tree’
	e.	și	pus -a			vo
		and	put-ptcp=have-aux.perf.3sg			cl.acc.f.3sg
		ăn cadere		cuhei (TC, 81)
		in pot		cook-inf
		‘she put it in the pot to cook’

Relevance for the status of Istro-Romanian auxiliaries. We believe that this word order flexibility is highly relevant for the status of Istro-Romanian auxiliaries.

In previous work, Nicolae (2015, 2019b) has argued that lexical verbs, not auxiliaries, may undergo movement to C in old Romanian, and that this is determined by their morphophonological nature. To be more precise, it is shown that Rivero’s (1994) LHM pattern is derived by Vicente’s (2007) (cf. also Kayne 1991) V⁰-to-Spec analysis (an analysis also adopted by (Harizanov 2016) in the derivation of Bulgarian participle fronting). Old Romanian auxiliaries (which are clitics) remain in situ, while lexical verbs move to C; evidence that clitics themselves cannot move to C, while non-clitic elements may undergo movement to C comes from the incorporation of clitic adverbs (clitic adverbials incorporate into the structure of the lexical verb and move along with it, but cannot themselves move) and from the blocking effects of phrasal negation on V-to-C movement12. Turning now to Istro-Romanian, from the sets of examples in (9) and (10), it appears that both lexical verbs and auxiliaries may undergo movement to C. Applying the same logic as for old Romanian, it follows that in Istro-Romanian auxiliaries too may be categorized as non-clitic elements, this ensuring their eligibility for movement to C13,14.

As a side note, it is important to specify that, up to the present movement, we have not been able to identify a clear trigger of verb movement to C (be it an auxiliary or the lexical verb) in Istro-Romanian. On the Croatian pattern, Wackernagel requirements might be one of the causes, but this does not hold for all the situations where we find verb movement to C. This state of affairs strongly resembles old Romanian, where a clear trigger for V-to-C movement could not be identified either, a fact which lead to interpreting old Romanian V-to-C movement as an instance of old Romance “relaxed” V215, whereby movement to C is triggered by a formal feature in the C-domain, which does not have a clear functional correlation.

To sum up, Aux-to-C movement is relevant for assessing the status of Istro-Romanian auxiliaries: if we agree that, as in old Romanian, only non-clitic elements many undergo movement to C, then the fact that Istro-Romanian auxiliaries may undergo movement to C indicates that they are not clitics.

2.4. Summary: Non-Clitic Auxiliaries in Istro-Romanian

In this section, we have reviewed the evidence in favour of analysing Istro-Romanian auxiliaries as full words (non-clitic elements); this consists in the following:

• ability of auxiliaries to license VP-ellipsis;
• scrambling and interpolation, formally analysed as instances of low(er) verb movement on the clausal spine;
• Aux-to-C movement; Istro-Romanian auxiliaries may undergo movement to C just as the lexical verb of Daco-Romanian and other Balkan languages that display LHM.

3. Retention of an Archaic Feature vs. Contact: Old Romanian and Croatian

In this section, we examine the phenomena discussed above with respect to old Romanian and Croatian, in order to formulate a proper analysis of the Istro-Romanian data. One of the important goals of this paper is to determine the weight played by heritage vs. language contact in the syntax of Istro-Romanian, and VP-ellipsis and the other word order phenomena discussed here offer the perfect setting to carry out this investigation.

3.1. VP-Ellipsis

3.1.1. Old Romanian

In old Romanian, VP-ellipsis is very rarely attested (the five examples in (11)—see Nicolae 2019b, p. 85; Zamfir 2007, pp. 163, 320; Carabulea 2007, p. 45). In four of these examples, (11a), (11b), (11d), (11e), the ellipsis is co-licenced: the auxiliary is accompanied by functional material with which it makes up a cluster. Co-licensing is a form of phonological enhancement that bypasses the morphophonological licensing problem of auxiliaries: it indicates that old Romanian auxiliaries are transitioning from a full word status to a clitic status.

11.	a.	de să	va			cunoaște carii
		if cl.refl.pass	will-aux.fut.3sg			know-inf who
		l-au			rănit		și
		cl.acc.3sg.m=have-aux.perf.3pl			hurt-ptcp		and
		carii nu	l-au √ (Prav.1646: 168)
		who neg	cl.acc.3sg.m=have-aux.perf.sg
		‘if it will be known who hurt him and who did not’
	b.	cine	l-au			rănit		și
		who	cl.acc.m.3sg=have-aux.perf.3sg			hurt-ptcp		and
		cine	nu	l-au	√
		who	not	cl.acc.m.3sg=have-aux.perf.3sg
		‘who hurt him and who did not’(Prav.1646:168)
	c.	De	voiu	face	aceasta	de voe,	plată
		if	will-aux.fut.1sg	do-inf	this	by will	payment
		am;		iară să	voiu √,		fără	de voe,
		have-ind.pres.1sg		but if	will-aux.fut.1sg		without	by will
		vistiernicie	mi-e			data.
		treasury	cl.dat.1sg=be-ind.pres.3sg			give-ptcp
		‘If I do this willingly, I will be paid; but if I (do it) unwillingly, I will be given the treasury (department).’ (NT.1648:231r)
	d.	de-l	va	fi		mutat		sau
		if=cl.acc.m.3sg	will-aux.fut.3sg be-irr16			move-ptcp		or
		de	nu-l	va		fi √
		if	not=cl.acc.m.3sg	will-aux.fut.3sg		be-irr
		‘whether he has moved it (=the border) or has not’ (Prav.1646:78)
	e.	de	va	fi	greșit	cu	înșelăciune
		if	will-aux.fut.3sg	be-irr	err-inf	with	fraud
		sau	de nu	va		fi √
		or	if not	will-aux.fut.3sg be-irr
		‘if he has deceitfully committed a mistake or not’ (Prav.1646:160)

As will be shown below (see Section 3.2.1), the other condition on VP-ellipsis licensing is met: old Romanian shows evidence for low(er) verb movement on the clausal spine.

3.1.2. Croatian

Auxiliary-licensed VP-ellipsis is also attested in BCS (12) (Stjepanović 1998), as shown in (12). Arguing that clitics are located in different projections in syntax, Stjepanović (1998) shows that the size of the ellipsis site is variable: in (12b), ellipsis targets both the dative and the accusative clitic (along with the lexical verb), while in (12c), in the same configuration under VP-ellipsis, the accusative clitic is elided along with the lexical verb and the dative clitic is preserved. However, (12d) shows that ellipsis cannot operate discontinuously by deleting, along with lexical verb, the dative clitic and preserving the accusative clitic.

12.	a.	Oni		su		kupili	novine,	a	i
		they.mpl		be-3pl		buy-ptcp	newspaper	and	also
		vi	ste √		(takodje)
		you	be-2pl		too
		‘They bought newspaper and you did so, too.’
	b.	Mi smo	mu		ga	dali,	a	i	vi
		we are	him cl.dat.3sg		cl.acc.3sg given		and	also	you
		ste	<mu	ga	dali>	(takodje).
		be-2pl	cl.dat.3sg cl.acc.3sg given too
		‘We gave it to him, and you did so, too.’
	c.	Mi smo	mu		ga	dali,	a	i	vi
		we be-2pl	cl.dat.3sg		cl.acc.3sg	give-ptcp	and	also	you
		ste	mu		<ga	dali> (takodje).
		be-2pl	cl.dat.3sg		cl.acc.3sg	Given too
		‘We gave it to him, and you did so, too.’
	d.	*Mi smo	mu		ga	dali,	a	i	vi
		we are	him-dat		it-acc	given	and	also	you
		ste	<mu>		ga	<dali> (takodje).
		be-2pl	cl.dat.3sg		cl.acc.3sg	given too
						(in Stjepanović 1998, pp. 529–30)

While it is not entirely clear whether the same formal analysis can be given to this type of ellipsis in Croatian and Istro-Romanian (and old Romanian, see Section 3.1.1 above), what is relevant is the existence of surface string identity in the two idioms in contact: just as in Croatian, in Istro-Romanian, an auxiliary can also license the ellipsis of the lexical verb with which it is associated.

3.2. Discontiguous Verbal Clusters and Their Relevance

3.2.1. Old Romanian Interpolation and Scrambling: Low Verb Movement

Low(er) verb movement on the clausal spine is also found in old Romanian. The different patterns of scrambling and interpolation, with and without Cinque’s diagnostic adverbs for V-movement, represent empirical evidence in favour of this analysis. For the limited purposes of this paper, we illustrate the phenomenon with a few relevant examples, but the reader is referred to Nicolae (2019b: ch. 3) for an extensive discussion which brings to the fore more phenomena and data, more than those presented here.

In synthetic structures, adverbs (13) are among the constituents interpolated; note that pururea (‘always’) (13a) and iară (‘again’) (13b) are part of the set of Cinquean diagnostic adverbs for V-movement (merging in the lower pre-VP area and lexicalizing an Asp_perfect specifier and, respectively, an Asp_repetitive specifier). (Floating) quantifier subjects may also be interpolated (13c).

13.	a.	ce		se	pururea			pomeneaşte
		which		cl.refl.pass	always			mention.ind.pres.3sg
		‘which is always mentioned’ (Prav.1581:258r)
	b.	Să	nu	te	iară	în		patŭ		răstorni
		să.subj	not	cl.refl.2sg	again	in		bed		return-subj.2sg
		‘you should not go back to bed again’ (CC2.1581:146)
	c.	Și	aşa	fu	a		se		toţi	mântui
		and	so	be-ind.ps.3sg	a.inf		cl.refl.3pl		all	save-inf
		‘and this is how they all found salvation’ (CV.1563-83:48r)

In analytic structures, the interposition of adverbials in between the auxiliary and the lexical verb is well represented. Thus, the following V-movement diagnostic adverbials are found: low adverbs such as bine (‘well’) (14), high adverbs such as amu (‘now’) (lexicalizing a T-related specifier) (15), and adverbial expressions such as acumuși, de sârg (‘quicky’) (16), which are ambiguous between an Asp-event reading and Asp-process reading, occupying different positions in Cinque’s hierarchy (the high adverb space and the low adverb space, respectively). (Floating) quantifier subjects may also break the auxiliary-lexical verb cluster (17).

14.	şi	va		bine	orând<u>i		lucrurile	tale
	and	will-aux.fut.3sg		well	organize-inf		things.def	your
	and he will organize your things well’ (DVT.1679-99:323v)
15.	a.	de	vămŭ		amu	pune	științele		noastre
		if	will-aux.fut.1pl		now	put-inf	knowledge.pl.def		our
		înaintea	lu	Dumnezeu
		before	gen	God
		‘if we now put our knowledge before God’ (CC2.1581:19–20)
	b.	De	veri	amu	vărsa	lacrămi	pre	fața	ta
		if	will-aux.fut.2sg	now	shed.inf	tears	on	face.def	your
		‘if you now have tears shedding on your face’ (CC2.1581:52)

16.	a.	că	vei		acmuşu	muri
		that	will-aux.fut.2sg		quicky	die-inf
		‘that you will quickly die’ (FT.1571-5:1v)
	b.	de	s-au			de sârgu		lăţit
		so.that	cl.refl.3sg=have.aux.perf.3sg			quickly		expand-ptcp
		‘so that it (=Moldova) quickly expanded’ (ULM.~1725:A-1f.2v)

17.	din	a	cui	sfinție	ne		vrom
	from	al.f.sg	whose	holiness	cl.refl.pass		will-aux.fut.1pl
	toți	sfinți
	all	hallow-inf
	‘from whose holiness we will all be hallowed’ (CM.1567:241r)

Multiple interpolated and scrambled constituents are possible:

18.	a.	se	nu	se	cumva		fără	socotinţă
		să.subj	not	cl.refl.3pl somehow			without	judgement
		meastece	fără		leage
		mix.up-subj.3pl	without		judgement
		‘they should not recklessly interfere without judgement’ (Prav.1581:243v)
	b.	cum	au			Cristus	noauă	dzis
		as	have-aux.perf.3sg			Christ	us.dat	tell-ptcp
		as Christ told us’ (FT.1571-5:1v)

Nicolae (2019b: ch. 3) documents the entire range of interpolated and scrambled constituents; for the purposes of the present argument, the interpolation and scrambling of adverbials and of quantifiers represent good empirical evidence for the existence of the low(er) verb movement option in the grammar of old Romanian.

Furthermore, the statistical examination of interpolation and scrambling (see Nicolae 2019b, ch. 3) has shown that, in old Romanian, these phenomena preponderantly occur in embedded clauses. From a diachronic comparative Romance perspective, this distribution confirms the formal analysis based on low verb movement: it is expected for low verb movement to mostly occur in embedded clauses where V-to-C movement (i.e., V2) is generally blocked by the lexicalization of complementizers or wh-elements.

3.2.2. Croatian Interpolation and Scrambling, and Their Relevance

Defined as above (Section 1.3) (i.e., the separation of the clitic pronoun/auxiliary verb from the (finite) lexical verb), interpolation and scrambling are also found in BCS. Thus, in (19a–b), we see the adverbial ‘always’ separating the clitic pronoun from the synthetic verb, while in (19c–d), we see temporal and manner adverbs intervening between the auxiliary and the lexical verb.

19.	a.	Jelena	me	stalno	nešto	zapitkuje.
		Jelena	cl.acc.1sg	always	something	ask-ind.3sg
		‘Jelena is always asking me something.’
	b.	Jelena	me	stalno	zapitkuje	nešto.
		Jelena	cl.acc.1sg	always	asks-ind.3sg	something
		‘Jelena is always asking me something.’
	c.	Marko je		tek	juče	nešto	saznao.
		Marko be-3sg		only	yesterday	something	find-out-ptcp
		‘Marko found out something only yesterday.’
	d.	Javi	mi		ako	se	budeš
		let-know	cl.dat.1sg		if	refl	be-pres.2sg
		‘Marko found out something only yesterday.’
		nečega		iznenada	setio.
		something-gen		suddenly	remember-ptcp
		‘Let me know if you suddenly remember anything.’
		(Stojanović 1997, pp. 307, 313–14)

Note that more than one constituent can be interpolated (19a) or scrambled ((19c), (20)). Interpolation and scrambling occur both in main clauses (20a–b) and in embedded clauses (20c).

20.	a.	Juče	mi	je	Marija	nešto		ispričala.
		yesterday	cl.dat.1sg	be-3sg	Marija	something		tell-ptcp
		‘Marija told me something yesterday.’
	b.	Marko je		tek	juče		Jelenu upoznao.
		Marko be-3sg		only	yesterday		Jelena meet-ptcp
		‘Marko met Jelena only yesterday.’
	c.	Jasna	mi	je		rekla		da
		Jasna	cl.dat.1sg	be-3sg		tell-ptcp		that
		je	Petar	nešto		kupio.
		be-3sg	Peter	something		buy-ptcp
		‘Jasna told me that Peter had bought something.’
		(Stojanović 1997, pp. 307–8)

While it is clear that, on the surface, the phenomena in (18) to (20) are very similar to their Istro-Romanian and old Romanian counterparts, it is hard to determine whether these data can be analysed relying on the same rich clausal structure as that employed for Istro-Romanian and old Romanian, two Romance varieties. Given the C-orientation (and Wackernagel nature) of Croatian pronominal clitics (Mišeska Tomić 1996; Rivero 1997; Migdalski 2009), a formal analysis of interpolation based on low(er) verb movement is uncertain—at most speculative. On the other hand, Stojanović (1997, p. 310) observes that BCS main (i.e., lexical) verbs in periphrastic tenses do not raise out of the VP; thus, low(er) verb movement can be invoked, at least in the analysis of scrambling.

What is important for the purposes of the present analysis is the existence of these surface string identities in Croatian, Istro-Romanian, and old Romanian: this supports the idea that convergence obtains, even though it is limited to surface phenomena, not to deep structural properties.

4. Analysis and Conclusions

Putting aside the issue of what the correct formal analysis of the Croatian data is, what we unequivocally observe is the existence of surface string-identity: identical word order phenomena are found in old Romanian, Istro-Romanian (which shares the same ancestry as old Romanian), and Croatian/BCS (the contact language). From this perspective, language contact with Croatian appears to have acted as a catalyst: several word order patterns of the contact language overlap with similar patterns of the target language, even though they might not have the same syntactic structure. This overlap generates convergence (cf. Hickey 2010), ensuring the preservation of several archaic properties: in the particular case of auxiliaries, they retain the full word status, i.e., they do not move further up on Hopper and Traugott’s (2003) grammaticalization cline. The representation below (Figure 1) summarizes the relation between Istro-Romanian, old Romanian, and Croatian:

The existence of auxiliary-licensed VPE along with low verb movement represents convincing evidence for the non-clitic nature of auxiliaries. We believe that this is not a trivial property, limited to (morpho)phonological component. Since it occurs along with correlation with other properties, the non-clitic nature of auxiliaries is a feature that genuinely reflects some central properties of Istro-Romanian syntax. Secondly, it explains the massive variation found in the Istro-Romanian verbal nexus: being categorised as non-clitic elements, auxiliaries show the same displacement properties like full lexical verbs. Thus, along with the double setting of pronominal clitics, I-oriented (the Romance parameter setting) and C-oriented/Wackernagel (a setting resulting from contact with Croatian) (see Dragomirescu and Nicolae 2018, 2020), the word order freedom of auxiliary verbs, lexical verbs, and pronominal clitics described in Section 2.3 above derives from the fact that auxiliaries are equally good candidates for movement to C; Rivero’s LHM rule may therefore affect auxiliaries too.

Our findings are consonant with recent scholarship on the effects of language contact in syntax. In the analysis of the Judeo-Spanish spoken in Bulgaria, besides confirming the fact that in language contact settings structures and elements belonging to narrow syntax are stable, while interface phenomena are vulnerable, Fischer (forthcoming)17 also shows that equivalence relations are important: “in order for these structures to be retained or innovated equivalence relations between languages in contact are relevant”. The non-clitic auxiliaries of Istro-Romanian are a perfect illustration of this idea: a morphophonological property which has syntactic effects is retained as it instantiates an equivalence relation between the two languages—particularly, features associated with non-clitic auxiliaries are found both in Croatian and Istro-Romanian.