Exploring Microvariation in Verb-Movement Parameters within Daco-Romanian and across Daco-Romance

: This article reviews some of the principal patterns of morphosyntactic variation within Daco-Romanian and across Daco-Romance in support of a distinction between low vs high V-movement grammars variously distributed in accordance with diatopic variation (Daco-Romance: west vs east, Aromanian: north vs south), diachronic and diagenerational variation (Megleno-Romanian) and endogenous vs exogenous factors (Istro-Romanian). This approach, which builds on the insights of the Borer–Chomsky conjecture, assumes that the locus of parametric variation lies in the lexicon and the (PF-)lexicalization of specific formal feature values of individual functional projections, in our case the clausal heads T and v and the broad cartographic areas that they can be taken to represent. In this way, our analysis locates the relevant dimensions of (micro)variation among different Daco-Romance varieties in properties of T and v . In particular, we show that the feature values of these two heads are not set in isolation, inasmuch as parameters form an interrelated network of implicational relationships: the given value of a particular parameter entails the concomitant activation of associated lower-order parametric choices, whose potential surface effects may consequently become entirely predictable, or indeed render other parameters entirely irrelevant. In this way we can derive properties such as verb–adverb order, auxiliary selection, retention vs loss of the preterite, the availability of a dedicated preverbal subject position, the distribution of DOM, and the different stages of Jespersen’s Cycle across Daco-Romance quite transparently, based on the relevant strength of T and v in individual sub-branches and sub-dialects.


Introduction
Romance boasts an increasingly vast literature on V(erb)-movement, including detailed studies of individual varieties (Lois 1989;Kayne 1991;Cornilescu 2000b, pp. 89-92;Alboiu and Motapanyane 2000, pp. 22-24;Tortora 2002Tortora , 2010Tortora , 2015;;Zagona 2002, pp. 162-24, 168-70;Ledgeway and Lombardi 2005;Ledgeway 2009a;2009b, sct. 21.3.3)and comparative studies across different varieties (Cinque 1999, pp. 142-52;Ledgeway and Lombardi 2005, pp. 103-6;2014;Schifano 2015Schifano , 2018)).At the macro level, there are some broad similarities in the distribution of V-movement across Romance, inasmuch as all varieties appear to show some degree of raising (though for Brazilian Portuguese, see Tescari Neto 2013, 2019, 2020a, 2020b, 2022a, 2022b;Tescari Neto and Pataquiva 2020;Araújo-Adriano 2020, 2022, 2023), a view widely reported in the literature since Pollock's (1989) and Belletti's (1990) seminal contrastive studies of English, French and Italian (cf.also, Emonds 1978), witness the contrast between (1a) and (1b,c).Although in all three varieties exemplified in (2) the finite lexical verb invariably leaves its base position to vacate the v-VP complex, witness its position to the left of the very low completive adverb COMPLETELY immediately adjacent to the v-VP complex, it raises to different functional projections within the I-domain, as illustrated by its differential position with respect to different adverb classes.For example, in Spanish, the finite verb raises to the head position of the continuative aspectual projection immediately below the low adverb STILL, whereas in Italian it rises higher, to a head position above STILL but below the 'irrealis' modal projection lexicalized by the high adverb PERHAPS (for precise details, see Ledgeway and Lombardi 2005), and in French it rises to the highest available position above all adverb classes.Now, within this range of variation Romanian represents a somewhat special case, since opinions are divided about the correct classification of verb movement in the language.Following the north-south distinction developed in Ledgeway (2020), in southern varieties such as Spanish (3a) the finite verb is not probed by T, but remains within the lower vdomain, as shown by its position to the right of both higher and (many) lower adverbs, whereas in northern Romance varieties such as French (3b), the finite verb rises to a high position T within the I-domain from where it precedes adverbs contained both in the higher and lower adverb spaces.Our expectation then is that, as a southern variety (cf.La Fauci [1988] 1994, 1991, 1997, 1998;Zamboni 1998Zamboni , 2000;;Ledgeway 2012, chp. 7;2020), Romanian should pattern with Spanish (3a), and not with French (3b).However, in practice, both the lower (4a) and higher (4b) options are found.
( Putting to one side the question of whether the position of the finite verb is the result of head or phrasal movement, the low V-movement option illustrated in (4a) is reported, among others, by Cinque (1999, p. 152), Ledgeway and Lombardi (2005, p. 101) and Ledgeway (2012, pp. 140-50), whereas scholars such as Dobrovie-Sorin (1994), Alboiu and Motapanyane (2000, pp. 15f., 22-24), Cornilescu (2000b, pp. 84-92), Alboiu (2002), Schifano (2015Schifano ( , 2018, pp. 65-68, 192-94), pp. 65-68, 192-94), Nicolae (2015, chp. 3;2019, pp. 13-25) and Dragomirescu and Nicolae (2021, pp. 2f.) have argued for a high V-movement grammar (4b).Now, rather than treat these two positions as competing and incompatible theoretical analyses of the same phenomenon, Ledgeway (2020, p. 54;2022a, sct.2) hypothesizes, largely on the basis of Costea (2019, sct. 3.1), that both positions may be correct insofar as they characterize different idiolectal and diatopic varieties of Romanian, varieties no doubt even (partially) present within the same speaker, given widespread literacy and the inevitable exposure today through the media to different varieties of spoken Romanian.Consequently, our fieldwork has revealed, for example, that eastern speakers (e.g., from Muntenia) typically interpret the postverbal position of the low aspectual adverb mereu as pragmatically neutral in utterances such as (4b), whereas the preverbal position in (4a) is interpreted as occupying a marked focus position such that the subject copilul must necessarily be interpreted as leftdislocated.By contrast, our western informants (e.g., from Oltenia) typically interpret the preverbal position of the same low aspectual adverb in (4a) as pragmatically neutral, with the consequence that the subject copilul is also free to be interpreted as a corrective focus, and the adverb can also be preceded by higher modal adverbs such as cică 'apparently', probabil 'probably' and poate 'perhaps' (5).
(5) Copilul [ HAS cică / probabil / poate [ LAS mereu plânge [ v-VP plânge.]]](Ro 1 ) child.DEF apparently probably perhaps always cry.3SG 'The child apparently/probably/perhaps always cries.' Largely similar facts are presented by Boioc Apintei (2021), who, based on a survey of 193 speakers, finds, for instance, that under pragmatically neutral readings preverbal mereu 'always' was preferred in 56.2% of cases, with even higher percentages for preverbal adverbs from the higher adverb space such as probabil 'probably' (71%) and poate 'perhaps' (94.3%).Data like these therefore raise the possibility that uniform characterizations of Romanian verb movement like those proposed, for instance, for French (high) and Spanish (low) are not possible.Rather, the situation is more akin to that found in Italian where, in line with the underlying dialects, northern regional varieties of Italian display high(er) V-movement and southern varieties typically show low(er) V-movement, with the (written) standard aligning more readily with northern usage.
In what follows we therefore explore further and expand on the preliminary evidence discussed in Costea (2019, sct. 3.1), Boioc Apintei (2021) and Ledgeway (2022a) to argue that we must recognize two diatopic varieties of Romanian, one with low (4a) and the other with high (4b) V-movement, henceforth referred to as Romanian 1 (west) and Romanian 2 (east), respectively.Given the general consensus in the literature (though see Hill and Alboiu 2016) that old Romanian was a low V-movement language (cf.Nicolae 2015, chp. IV, 2019, chp. 3;Costea 2019, sct. 2.1.1;Boioc Apintei 2021, chp. 5;Dragomirescu and Nicolae 2021, pp. 12f.), it is logical to take Romanian 1 to represent a more conservative southern Romance variety, with Romanian 2 interpreted as a more innovative grammar, quite possibly the result of Balkan areal contact, a conclusion which will be further substantiated below. 2 In this light, it is quite easy to see how a P-ambiguous string (cf.Clark and Roberts 1993) such as (4a), the output of the southern Romance original low V-movement grammar (6a), could be reanalysed, under the influence of areal contact, as the output of a high Balkan V-movement grammar (cf.Rivero 1994), with the preverbal adverb concomitantly reinterpreted as a focused constituent (6b).
(6) a. [ TopP (Copilul)  Indeed, in what follows we shall expand our investigation beyond different diatopic varieties of Daco-Romanian to explore the distribution of competing southern Romance and Balkan grammars more broadly across Daco-Romance to include comparative evidence from Aromanian, Megleno-Romanian and Istro-Romanian. 3In this way, not only do we offer an original descriptive account of the microvariation attested in the height of V-movement within Daco-Romanian in accordance with a broad west vs east split (cf.Section 3.1), but our analyses will also capture microvariation in the height of Vmovement across Daco-Romance (cf.Section 3.2), taking into account both its Balkan and southern Romance dimensions.Finally, we shall show how differences in the setting of the V-parameter across Daco-Romanian and Daco-Romance can be used to throw light on a whole series of apparently unrelated phenomena, including perfective auxiliary selection and the distribution of analytic and synthetic past-tense forms (Section 4), subject positions (Section 5), differential object marking (Section 6) and negation (Section 7).

North-South Romance Divide
In contrast to traditional Romance classifications in terms of a west-east axis (cf.Wartburg 1950), more recent work (La Fauci [1988] 1994, 1991, 1997, 1998;Zamboni 1998Zamboni , 2000;;Ledgeway 2012, chp. 7) has argued within a wider Romance typology for a northsouth axis which distinguishes between varieties of the northern Romània (langue d'oïl and langue d'oc varieties, Francoprovençal, Raeto-Romance, northern Italian dialects, and (Tuscan)-Italian) characterized by an active-stative alignment, and varieties of the southern Romània (central-southern Italian dialects, Sardinian, Ibero-Romance varieties, Dalmatian, and Daco-Romance) which display a nominative-accusative alignment.Following ideas first developed in Ledgeway (2020), we argue here that this broad areal difference can be derived from the relative strength and often complementary featural composition and roles of the clausal functional heads T and v such that a major dimension of variation between northern and southern Romance concerns the ability of T to probe V.For example, on a par with the French example in (3b), in the Gallo-Romance varieties of Wallon (7a), Provençal Occitan (7b) and Milanese (7c), the finite verb (in bold) raises to a high position within the T-domain, from where it precedes adverbs (underlined) contained both in the higher and lower adverb spaces.In southern varieties, by contrast, T does not probe the finite verb which remains within the v-domain, as shown by its position to the right of both higher and (many) lower adverbs in the Sicilian (8a), Valencian Catalan (8b) and European Portuguese (8c) examples, which replicate the pattern already seen above for Spanish in (3a).Our approach, which locates the relevant dimensions of variation in the properties of individual functional heads, builds on the insights of the Borer-Chomsky conjecture (Baker 2008, p. 353), placing the locus of parametric variation in the lexicon and, in particular, in the PF-lexicalization of formal feature values of individual functional heads (Borer 1984;Chomsky 1995).These feature values are not, however, set in isolation, inasmuch as parameters form an interrelated network of implicational relationships, whereby the value of a particular parameter may entail the concomitant activation or deactivation of associated lower-order parametric choices.Consequently, the setting of a given parameter often entails quite considerable knock-on effects on other functional heads whose force and content either become predictable or are rendered irrelevant.Such is the case, for instance, with T, which Ledgeway (2020Ledgeway ( , 2022a) ) shows probes the verb in northern Romance languages and dialects which exhibit high V-movement, a parametric setting that, in turn, accounts for V-Adv orders, the availability of verb-subject inversion, the licensing of Stage II and III negation, the presence of subject clitics and a dedicated preverbal subject-position, and active-stative perfective auxiliary selection.In the varieties of southern Romance, by contrast, all these same options are absent since T fails to attract the verb, which is probed instead by v, yielding low V-movement, a parametric choice which explains Adv-V orders, the absence of V-to-C movement, Stage I negation, the use of generalized or person-driven perfective auxiliation, active participle agreement with in situ objects, and the availability of differential object marking (DOM).
Building on this approach to Romance variation, in the following sections we shall demonstrate how the differential setting of the V-parameter, in terms of the relative strength of T and v, can also offer a principled account of the variation observed not only within Daco-Romanian, but also more broadly across Daco-Romance.In particular, our detailed comparative examination of the V-parameter will not only enable us to identify and formalize some quite substantial similarities and differences between the four principal subbranches of Daco-Romance, but it will also allow us to discern more subtle differences within Daco-Romanian which have either been entirely overlooked or dismissed in the literature.Thus, rather than assuming Daco-Romance to be characterized by a single homogeneous grammar, the evidence to be reviewed in the following sections will reveal some significant microvariation among different areas and speakers, which will allow us, among other things, to distinguish, on the one hand, a low(er) V-movement grammar typical of western Daco-Romanian varieties and a high V-movement grammar, on the other, typical of eastern Daco-Romanian varieties and the written standard.In a similar fashion, our broader investigations of Daco-Romance will highlight similar low vs high V-movement distinctions across other sub-branches of Daco-Romance in accordance with diatopic varia-tion (e.g., Aromanian), diachronic and diagenerational variation (e.g., Megleno-Romanian) and endogenous vs exogenous factors (e.g., Istro-Romanian).

Daco-Romanian
As noted in Section 1, views in the literature about Romanian V-movement are divided between low and high V-movement analyses, although the actual details and, in particular, microvariational differences between individual speakers and regions have to date largely been overlooked.By contrast, our fieldwork inquiries highlight a striking difference, not only in the placement of the verb, but also in the lexicalizations of different adverb classes employed in western and eastern Romania, roughly corresponding, respectively, to (pockets of) Maramures , , Banat, Cris , ana and Oltenia on the one hand and Muntenia (Greater Wallachia), Dobrogea, Moldova and Bucovina on the other.While diatopic variation regarding the lexicalization of different adverb classes has been previously noted in the literature (cf.Chircu 2006), it has never been taken into account when evaluating V-movement.Significantly, we have found that when we test V-movement in relation to adverbs taken from the standard language, e.g., deja 'already', mereu 'always' and probabil 'probably', native speakers of Daco-Romanian tend to display analogous patterns of V-movement: oversimplifying somewhat, the tendency is to raise the verb above TP Anterior lexicalized by deja 'already' but below MoodP Irrealis lexicalized by poate 'maybe' and ModP Epistemic lexicalized by probabil 'probably'.However, when we test the same speakers using their local adverbial forms, there emerges a clear east-west divide, ultimately parallel to the broader north-south divide proposed by Ledgeway (2020), with western varieties displaying a lower level of V-movement than eastern varieties.
The local lexicalizations of each of the adverb classes making up the higher and lower adverb spaces in western varieties is given in (9a,b; cf. also ALR (V), 1323,1384,1470).Testing of each of the various classes in relation to V-movement reveals that the finite verb remains in the lower adverb space, reaching a position immediately above AspP Proximative .This is exemplified in (10) and (11) from the western variety spoken in Sacos , u Mare (Timis , , Banat), where we observe that the finite verb follows not only higher adverbs such as the irrealis puaci 'perhaps' (10a), the temporal dăl ȗoc (10b) and volitional într-a dins 'deliberately' (10c), but also those adverbs situated in the highest portion of the LAS, namely perfective tădăuna 'always' (11a) and retrospective bas , acuma 'just' (11b), but obligatorily precedes the proximative dăluoc 'soon' (11c) and the completive dă tuot 'completely' (11d). 4From this we can conclude that in western Romanian varieties the finite verb raises at least as high as the Asp Retrospective head, but not any higher (12).( 9 In the east, by contrast, the verb raises higher, reaching the higher adverb space.The local lexicalizations of each of the adverb classes across both spaces are given in (13a-b; cf. also ALR (V), 1323, 1384, 1470), with relevant examples from Grint , ies , (Neamt , , Moldova) in ( 14) and ( 15) illustrating the position of the finite verb in relation to higher and lower adverbs, respectively.The contrast between (14a) and (14b,c) shows that the verb raises to a position below the higher epistemic adverb poati 'probably', but above the temporal and volitional adverbs amu' 'now' and ispre 'intentionally', and consequently above all the lower adverbs exemplified in (15a-d).We can therefore conclude that in eastern varieties the finite verb targets the head of MoodP Epistemic in the higher adverb space, as illustrated in ( 16).Turning now to analytic past-tense forms, contrasts in the height of V-movement are on the whole less clear-cut, although here too, the east appears to display a higher level of V-movement.For instance, in eastern varieties the verbal complex consisting of auxiliary and participle must occur below the epistemic modal adverb cre' c' 'probably' (17a) but to the left of higher temporal adverbs such as amu' 'now' (17b), which unambiguously situates it in the head of MoodP Epistemic , whereas in the west the verbal complex must occur below these same temporal adverbs (17c).Indeed, the position of the analytic verbal complex is much lower in western varieties, in that it not only surfaces to the right of the volitional adverb într-a dins 'intentionally' (18a) situated at the bottom of the higher adverb space, and also below the perfective aspectual adverb tădăuna 'always' (18b) situated towards the top of the lower adverb space, but it occurs above the proximative dăluoc 'soon' (18c) and the completive dă t ȗot 'completely' (18d Thus, on a par with simplex finite verbs, in the east the auxiliary + participle complex raises at least to the Mood Epistemic head in the higher adverb space (cf.16), from where it precedes temporal adverbs.In the west, by contrast, the analytic complex surfaces in the lower adverb space where it raises at least to the Asp Restrospective head (cf.12), from where it follows all higher adverbs as well as the lower perfective adverb ALWAYS, but precedes proximative and completive adverbs.More precise evidence regarding the position of the verbal complex in the west can be adduced from minimal pairs such as the Oltenian examples from Valea Mare (Vâlcea) in (19a,b).As previously suggested in the literature (cf.Schifano 2018, p. 78), in low(er) V-movement varieties, postverbal adverbs typically receive a focalized reading, inasmuch as they are forced to occur in the lower left periphery (Belletti 2004).Such is the case in (19a), where our informants reported that the only possible reading of the simplex finite verb preceding the negative adverb neam 'not (at all)' is that in which the adverb is focalized.This follows straightforwardly from our observation above, that the finite lexical verb in the west raises to a position below NegP2, viz.Asp Retrospective (cf.12).Consequently, an example such as (19a') with a focalized postverbal subject proves ungrammatical, since there can only be one focus per sentence, either the negative adverb or the subject, but not both.By contrast, examples such as (19b) containing a compound verb preceding neam were judged pragmatically unmarked by the same informants.This must imply that neam in such cases occurs in its base position in the clausal spine with the verbal complex moving above it, presumably raising to the head of AspP Perfective , since complex verbs follow adverbs such as ALWAYS (cf.18b).Indeed, this is confirmed by the observation that in such cases neam can now be followed by a focalized subject (19b'), since the negative adverb occurs in situ and not in the lower focus position, which is free to host the postverbal subject.We conclude therefore that, whereas in the east both simplex and compound finite verbs invariably raise to the head of MoodP Epistemic (20a), in the west, V-movement proves sensitive to the distinction between simplex and compound verb forms (20b); the former raise to the head of AspP Retrospective but the latter raise slightly higher, to the head of AspP Perfective .

Daco-Romance
Variation in the height of V-movement is by no means limited to Daco-Romanian, but equally characterizes other Daco-Romance varieties.Beginning with Aromanian, we can observe a split between southern and northern varieties (henceforth Aromanian 1 and Aromanian 2 ), with the former displaying lower V-movement than the latter.For example, in southern dialects lexical verbs surface below the perfective aspectual adverb tută dzua/dzua tută 'always', lexicalizing one of the lower positions within the lower adverb space (21a).In northern Aromanian varieties, by contrast, this same class of adverbs usually occur after the lexical verb in pragmatically unmarked sentences (cf.21b), 6 highlighting how the lexical verb raises to one of the higher positions within the lower adverb space, a behaviour which we interpret as a syntactic borrowing from Greek, which displays high(er) V-movement (Mavrogiorgos 2010, p. 182) (Papahagi 1905, p. 71) A similar development is also found in Megleno-Romanian varieties.Although it has not been possible to identify any clear generalizations about diatopic variation (Megleno-Romanian is today mainly spoken in the Meglen region split between Northern Macedonia and Greece), there do seem to be consistent idiolectal options in Theodor Capidan's (1928) corpus and in present-day recordings (e.g., the VLACH project, https://www.oeaw.ac.at/ vlach/collections/meglen-vlach (accessed on 28 December 2023)).In particular, speakers consistently display either high V-movement to the left of TP Past/Future , lexicalized by temporal adverbs such as cmo 'now' ( 22), or lower V-movement to the right of the same functional projection and its associated adverb (23).(Capidan 1928, p. 123) This is the traditional situation evidenced in older materials (henceforth, MRo. 1 ).However, an examination of contemporary data such as the VLACH recordings from Os , ani/Ossiani made in 2015 reveals a diachronic development whereby the variation between high and lower V-movement observed in the past (cf.22-23) is today increasingly replaced by a more consistent low(er) V-movement grammar across all speakers (henceforth MRo. 2 ).This can be seen in ( 24) below, taken from 2015 VLACH recordings from Os , ani/Ossiani, where the lexical verb typically occupies a low position below, for example, tucu 'always' (viz.SpecAspP Perfect ).
(24) Tucu ne ra frica.(MRo. 2 , Os , ani/Ossiani) alwaysus= be.PST.IPFV.3SGfear.DEF 'We were always wary' (VLACH, The snake and the flute, 04.09) Superficial oscillations in the level of V-movement apparently occur also in Istro-Romanian, witness the position of the finite lexical verb to the left and to the right of the manner adverb bire 'well' in (25a-b), respectively.However, in this case, optionality is only apparent, inasmuch as speakers, who are all bilingual with Croatian, seem to be able to access both a conservative, Daco-Romance-style higher V-movement grammar (henceforth IRo. 1 ) in which the verb raises to the left of lower adverbs such as bire 'well' (25a), 7 as well as an innovative Croatian-style lower V-movement grammar (henceforth IRo. 2 ) in which the verb follows lower adverbs such as bire 'well' (25b; cf.Dragomirescu and Nicolae 2021).The difference between (25a) and ( 25b) is therefore to be captured in terms of language contact, with verb-raising above and below VoiceP in accordance with competing Daco-Romance and Croatian parametric options available to all speakers.While synthetic verbal forms such as the present, past imperfective and the preterite allow us to make specific generalizations about V-movement, the same cannot be said about analytic past tenses.Deliberately oversimplifying, both Aromanian and Megleno-Romanian display an analytic past tense consisting of the auxiliary HAVE and an invariable form of the past participle to encode iterative aspect and/or events with present relevance (Capidan 1925, pp. 167, 205;1932, p. 463).However, the default past-tense form in both varieties is the preterite, with which the analytic past is in free variation, even in contexts such as (26a-b) (Tomić 2006, pp. 377-79, 384-86).Thus, given the sporadic attestations of the analytic past in Aromanian and Megleno-Romanian in both oral and written corpora, it has not been possible to reach any significant conclusions regarding the default position of auxiliary and participle in relation to adverbs.
(26) a. Me-am dusă tu aistă politie multe or.(Aro.)me.ACC=have.1SGlead.PTCP in this city many times 'I've been to this city more than once.' (Tomić 2006, p. 385) b.Mi-am dus prin toati cătunili.(MRo.)me.ACC=have.1SGlead.PTCP through all villages.DEF 'I've been through all the villages.'(Atanasov 2011, p. 485) Istro-Romanian, on the other hand, has generalized an analytic past tense consisting of the auxiliary HAVE and the invariable past participle (Kovačec 1971, p. 147;Sârbu and Frăt , ilă 1998, p. 30).Interestingly, Istro-Romanian speakers seem once again to permit a dual syntactic interpretation of the auxiliary: on one hand it may pattern with Croatian and occupy the second position in the clause, 8 but on the other it can also replicate the Daco-Romance pattern and lexicalize various I-related positions. 9Now, for some speakers the default position of past-tense auxiliaries is immediately below (Spec)TP Anterior , lexicalized by vet" 'already' (cf.27a) in the lower adverb space.However, for most speakers who today have Croatian as their dominant language, they can also move the auxiliary to the second position in the clause, as we see in (27b) where raising of the finite auxiliary to a very high position-variously interpreted in the literature as involving movement to a high position within the I-domain (Bošković 1997;Migdalski 2006, pp. 71, 180f.) or to the C-domain (Dragomirescu and Nicolae 2021)-places it to the left of the pronominal subject voi 'you' and immediately left-adjacent to the deictic adverbial as , 'thus'.

Interim Summary
Our investigations of V-movement have revealed a picture involving some quite significant microvariation, which makes it impossible to talk of a single setting for the V-parameter across Daco-Romance.In particular, we have seen that there is notable diatopic variation within Daco-Romanian, with varieties from the west (Romanian 1 ) displaying lower placement of the finite lexical verb and the analytic compound past than those spoken in the east (Romanian 2 ).Similar degrees of variation have also been observed for the other major branches of Daco-Romance, with Aromanian showing a low vs high opposition in accordance with a respective south vs north distinction (viz.Aromanian 1 vs Aromanian 2 ); Megleno-Romanian traditionally displaying free variation between different speakers (Megleno-Romanian 1 ), though with a tendency in more recent times for all speakers to converge towards a low V-movement grammar (Megleno-Romanian 2 ); and Istro-Romanian apparently alternating between a conservative inherited higher low V-movement grammar (Istro-Romanian 1 ) and a Croatian-style very low V-movement grammar (Istro-Romanian 2 ).This is the situation with finite lexical verbs, whereas the picture with the analytic past has been less clear: the periphrasis proves so infrequent in Aromanian and Megleno-Romanian that reliable data are not available, but the frequency of the periphrasis in Istro-Romanian allowed us to identify once again two competing patterns, a Daco-Romance-style placement in the higher portion of the lower adverb space (Istro-Romanian 1 ) and a Croatian-style movement of the auxiliary to the high I-domain or the C-domain (Istro-Romanian 2 ).These facts are summarized in Table 1.Finally, we must note that our investigations have highlighted that, across Daco-Romance in general and in Daco-Romanian in particular, it is not always possible to test Vmovement in relation to particular adverb classes of the standard language.This is because the adverbs in question are simply not lexically available in some of the non-standard varieties and the relevant modal, temporal and aspectual categories have to be licensed, not through the internal merge of an adverb in the specifier of the relevant functional projection, but rather through raising of a particular verb form to the head of the same functional projection.This is the case, for example, in some of the western Daco-Romanian varieties spoken in Oltenia, Cris , ana and (Serbian) Banat, which lack a local retrospective aspectual adverb corresponding to the standard tocmai 'just', but mark this aspectual category with the preterite, which is raised to the head of AspP Retrospective (28a).That the verb sits in the head of AspP Retrospective in examples such as (28a) is clearly demonstrated by examples where the verb obligatorily follows the immediately higher continuative încă 'still' (28b) and perfective to' timpu' 'always' (28c) adverbs, but precedes the completive adverb dă tot 'completely'.Similarly, in the Boyash variety of Daco-Romanian currently spoken in Hungary there is no local habitual aspect adverb corresponding to standard de obicei 'usually' (cf.Kahl and Nechiti 2019); by the same token, we can therefore assume that habitual aspect is licensed in this variety by the erstwhile pluperfect raising to the head of AspP Habitual (29).Another notable example comes from Megleno-Romanian where evidential modality is lexicalized, not through the external merge of an adverb in SpecMoodP Evidential (cf.Ro. aparent 'apparently'), but by raising of the so-called 'inverted analytic past' to the head of MoodP Evidential (30a), 10 as partially indicated by examples such as (30b), where the verbal complex is correctly predicted to precede adverbs of the lower adverb space such as the repetitive aspectual adverbs de cătiva or and ară 'again'.(Capidan 1928, p. 102) In all three cases in ( 28)-( 30) we see a mismatch between the traditional value of the relevant morphological verb forms (viz., preterite, pluperfect and present perfect) and the syntactic positions in which they surface, highlighting how the aspectual and modal values of each are transparently licensed, not by their form, but by the functional projections they lexicalize.

Daco-Romanian
As hinted in Section 2, following Ledgeway (2020, sct. 3.4, sct. 4.1), a further correlation deriving from the variability of V-movement concerns perfective auxiliary selection. 11 In particular, in most northern Romance varieties, which uniformly display a high Vmovement grammar, we see the continuation of an original active-stative split, whereby predicates with AGENT subjects select auxiliary HAVE (31a) and predicates with UNDER-GOER subjects select auxiliary BE (31b In southern Romance, by contrast, this active-stative distribution has in most cases been replaced by a nominative-accusative alignment variously involving the generalization of a single auxiliary, whether HAVE (32a)   Perlmutter 1978;Burzio 1986), it has been assumed that auxiliary BE represents the superficial reflex of a co-indexation relation between T and V in accordance with the idea that unaccusative structures involve raising of the object to the surface subject position, as formalized in (33).
(33) Auxiliary BE is selected whenever (Spec)T is indexed with V(,DP) In northern Romance, verbs overtly raise to T, an operation that automatically results in the co-indexation of V and T which, in accordance with (33), produces the observed Now, in traditional work on active-stative patterns of auxiliary selection within the Unaccusativity Hypothesis (cf.Perlmutter 1978;Burzio 1986), it has been assumed that auxiliary BE represents the superficial reflex of a co-indexation relation between T and V in accordance with the idea that unaccusative structures involve raising of the object to the surface subject position, as formalized in (33).
(33) Auxiliary BE is selected whenever (Spec)T is indexed with V(,DP) In northern Romance, verbs overtly raise to T, an operation that automatically results in the co-indexation of V and T which, in accordance with (33), produces the observed sensitivity of the perfective auxiliary to the active-stative distinction.By the same token, we now also have a natural and principled explanation for the typical absence of active-stative auxiliation patterns in southern Romance, since verbs do not raise to T in the south but, rather, remain within the v-domain. 12It follows from the Phase Impenetrability Condition (PIC) that T and V will never be co-indexed and auxiliary BE will never therefore surface as the result of an unaccusative structure.
These facts find further confirmation in our dialectal investigations, in which the low Vmovement grammars of western Daco-Romanian varieties align with a more conservative (Balkan-style) nominative-accusative pattern, in contrast to eastern high-V-movement varieties within which we witness a partial move away from the traditional nominativeaccusative alignment towards an innovative active-stative split in the perfective auxiliaries.Beginning with western varieties, these operate a nominative-accusative split, reminiscent of (older) Balkan varieties, whereby a single auxiliary, usually avea 'have', is selected with both transitives (34a) and intransitives (34b) (cf.also ALR (VII) 1972, map 1801, 1997), and the participle remains invariant.Now, alongside this widespread pattern, there is also a further pattern with auxiliary fi 'be', which is increasingly obsolescent today, although it is residually preserved in some frozen structures.On a par with many Balkan (Slavonic) languages (Lambova 2003, p. 4;Migdalski 2006, p. 82), this obsolescent use of fi 'be' in western varieties also occurs with transitives (35), where, crucially, the participle always agrees with the subject, a further syntactic reflex of a nominative-accusative alignment.By contrast, perfective auxiliation in eastern varieties seems to display the early stages of an emerging active-stative alignment, as evidenced by an increasing use of the resultative periphrasis with auxiliary fi 'be' in conjunction with unaccusative participles (36a; cf.Dragomirescu and Nicolae 2013, p. 343).Marin (1985, p. 462), too, discusses the phenomenon, remarking that it is frequent in all Daco-Romanian dialects; however, all her examples of resultative fi 'be' come from the area that we labelled above as 'eastern', namely from Buzău, Teleorman and Mures , .Furthermore, our own corpus study of dialectal texts (e.g., Cazacu 1967Cazacu , 1975;;Cohut , and Vulpe 1973) also reveals an undeniably greater preference for this construction in the east than in the west, witness the eastern example in (36b) from Vrancea county.Analogous conclusions also emerged from our interviews with informants from western and eastern areas.More specifically, western speakers from Banat and Oltenia, for example, described examples such as (36a-b) as 'Greater Wallachian' or 'literary', failing to recognize them as part of their own dialect, 14 whereas eastern speakers from Muntenia such as Giurgiu county, for instance, immediately recognized them as part of their own colloquial registers.The emergence of a stative fi 'be' periphrasis in conjunction with UNDERGOER subjects is paralleled in the west by the concomitant emergence of an active resultative periphrasis with avea 'have'.Marin (1985, p. 464) discusses this type of structure as well, but does not offer any further details with respect to dialectal variation.However, all of the examples that she provides come again from the east, namely, Arges , and Suceava.Oversimplifying somewhat, it seems to be the case that speakers who display the fi 'be' periphrasis are more prone to accept, and indeed, to actively employ, the avea 'have' resultative periphrasis.For instance, structures such as (37) were readily accepted by our eastern informants from Giurgiu county, while the same structures were described as, at best, marginal by western speakers from Olt county (e.g., Slatina).In contrast to western varieties, we thus see in eastern varieties the presence of an active-stative auxiliary split in the resultative construction with HAVE selected in conjunction with AGENT subjects and BE with UNDERGOER subjects.This development towards an active-stative split in the auxiliary system is arguably also reflected in the loss of the preterite in the east (cf.ALR (VII) 1972, map 1981) since the latter is a reflex of a nominative-accusative alignment. 15It is therefore unsurprising that the preterite is preserved in western varieties spoken in Oltenia, Banat and Cris , ana (Rosetti 1955, pp. 69-73;Moise 1977, pp. 91-93;Pană-Boroianu 1982, pp. 423-24;Neagoe 1985, pp. 172-76;Havu and S , tirbu 2015, pp. 144-47; cf. also ALRR (Banat) 2005, map 674), although its values and distribution are not uniform across western varieties (cf.Chitez 2010).For example, in Cris , ana (or more generally 'Transylvania' in Chitez's terms) the preterite is used to refer to both events that took place in a distant past (38a), 16 and events that took place during the same day (38b), 17

Daco-Romance
Above, we noted that Aromanian and Megleno-Romanian display internal variation with respect to the height of V-movement, the former in accordance with a south-north split (viz.Aromanian 1/2 ) and the latter traditionally in accordance with inter-speaker variation (viz.Megleno-Romanian 1 ), despite a more recent tendency towards a consistent low V-movement grammar across all speakers (viz.Megleno-Romanian 2 ).As already briefly noted above, we interpret the high V-movement option in these two varieties, at least in our corpora, as a direct consequence of contact with Greek.Now, given our assumptions, low V-movement varieties are predicted to consistently display a nominative-accusative alignment in their verb system, hence their use of a single generalized auxiliary for past-tense forms and the retention of the preterite.And this is indeed what we find in both Aromanian and Megleno-Romanian, which appear to pattern largely with western Daco-Romanian in this respect: auxiliary HAVE (viz.Aro.av(e)are and MRo.veari) invariably occurs with both transitives (40a-b) and unaccusatives (cf.26a-b), and the preterite is currently employed to refer to all kinds of past events, with or without present relevance (41a-b).( 40 On a par with lexical verbs that can target higher positions, auxiliary verbs may also occupy similar positions.While this development cannot be tested by comparing the position of analytic past-tense forms with respect to the relevant adverbs (cf.Section 3.2), it can be indirectly verified by analysing the availability or otherwise of (emergent) active-stative tendencies in auxiliary selection which we independently know to be correlated with higher V-movement (cf.33).Now, in both Aromanian and Megleno-Romanian, speakers often make use, albeit inconsistently, of auxiliary BE (Aro.h'ire, MRo.iri) with unaccusatives. 19 For example, Megleno-Romanian speakers from Capidan's (1928) corpus coming from Os , ani/Ossiani (Greece) employ iri 'be' significantly more frequently with unaccusative participles (42) than speakers from Uma/Huma (present-day Northern Macedonia). 20In the case of Aromanian, the split seems to involve, as expected, the distinction between high-V-movement northern varieties and low-V-movement southern varieties (cf.also Capidan 1932, p. 541): the former (viz.Aromanian 2 ) optionally, though by no means exclusively, use h'ire 'be' with unaccusative participles, whereas such structures are absent in Papahagi's corpus, which predominantly includes southern varieties (viz.Aromanian 1 ) from, for example, Avela/Avdella, Met , ova/Metsovo, Samarina and Perivole/Perivoli.
(42) Io sam vinit di un lucru.(MRo. 2 , Os , ani/Ossiani) I be.1SGcome.PTCP of a thing 'I've come for a thing.' (Capidan 1928, p. 80) Somewhat differently, Istro-Romanian has lost the preterite and generalized the analytic past formed from auxiliary (a)vę 'have' and an invariant form of the participle to refer to all types of perfective past events, whether with or without present relevance.Moreover, similarly to Croatian (Sârbu and Frăt , ilă 1998, p. 33) and old Daco-Romanian (Dragomirescu and Nicolae 2021), Istro-Romanian allows for various constituents to intervene between the auxiliary and past participle.The tendency among speakers to increasingly treat the auxiliary as a Wackernagel clitic, in line with what is observed in Croatian, is a consequence of generalized bilingualism.However, the fact that the auxiliary seems to display a dual behaviour, sometimes being treated as a Wackernagel element but at other times not, points towards the fact, already noted above (cf.Section 3.2), that Istro-Romanian speakers have access to two grammatical systems: one which is closer to the rest of Daco-Romance (cf. the lack of Wackernagel auxiliaries in Daco-Romanian, Aromanian and Megleno-Romanian), and one which is closer to Croatian.Thus, what may appear an optional syntactic mechanism should, rather, be interpreted as a case of exogenous 'PAT(tern)' replication (cf.Matras and Sakel 2007, p. 847) from Croatian, enriching the grammar.Nevertheless, it must be noted that Istro-Romanian speakers have adapted the original Croatian pattern, since, unlike what happens in Croatian (cf.Migdalski 2016, pp. 152f.), auxiliaries come to occupy a high (T-related) position in the clausal spine, which, in turn, allows for the co-indexation of T and V (cf.33).As a result, we can recognize two auxiliary patterns in Istro-Romanian: one following a typical Daco-Romance nominative-accusative alignment, with lower placement of the auxiliary yielding generalized HAVE, and the other in accordance with an active-stative alignment, where higher placement of the auxiliary according to a reanalysed Croatian model licenses auxiliary fi 'be' with unaccusatives (43; cf.Geană 2017, p. 213). 21  (43) Fil'a-i cu ie verit (IRo.)daughter.DEF=be.3SGwith him come.PTCP 'My daughter has come with him' (Geană 2017, p. 213, apud Sârbu andFrăt , ilă 1998, p. 173)

Daco-Romanian
The west-east split at the level of V-movement is also correlated with variation in the availability of different subject positions.On a par with southern Romance varieties which also display low V-movement, in western Daco-Romanian varieties, the T-domain is inactive, since the verb is not probed by T but by v (cf.Ledgeway 2020, pp. 41f., 54-64).As a consequence, T in western varieties not only lacks a V-feature, but also a D-feature, such that it fails to attract either the verb or the clausal subject, with the result that SpecTP is not available as a default subject-position in western Daco-Romanian.It follows that subjects occupy either a v-related position (viz., SpecvP), which is overwhelmingly preferred in pragmatically unmarked sentences yielding thetic VSO orders, or a marked left-peripheral topic or focus position whenever preverbal.A direct structural consequence of the absence of an EPP-type SpecTP position is that orders such as Focus + Subject + Verb are predicted to prove ungrammatical in western Daco-Romanian, inasmuch as a focus, the lowest available left-peripheral position in the C-domain (Rizzi 1997), preceding the subject would require the subject to lexicalize SpecTP.Indeed, this prediction is borne out by our inquiries, with western speakers from Banat, Cris , ana and Oltenia rejecting sentences such as (44a) in which the preverbal subject is preceded by a focused constituent, rephrasing it as in (44b), in which the subject surfaces in a v-related postverbal position.By contrast, eastern varieties which display high V-movement are characterized by an active T-domain.In these varieties, as we have seen (cf.Section 3.1), T therefore comes with a strong V-feature which attracts the verb, but it also carries a corresponding strong D-feature spelt out in the licensing of a preverbal SpecTP subject position. 22Indeed, for eastern Daco-Romanian speakers, there seem to be as many as three subject positions: a T-related EPP-type preverbal position (45a), the most frequent when the subject is overt and licensing pragmatically unmarked SVO thetic orders; a C-related position for marked topicalized or focalized subjects (45b); and a v-related postverbal position (45c Now, a crucial difference between eastern and western Daco-Romanian varieties lies in the fact that the former, but not the latter, display a so-called 'double subject construction' (Cornilescu 2000b, pp. 98-113), which on the surface appears to license two postverbal subjects.Thus, in a sentence such as (46), which is considered grammatical by eastern speakers, the first pronominal subject, ei 'they', occupies the dedicated preverbal subject position SpecTP, while the second lexical subject, copiii 'the children', occurs in SpecvP.Note that in this case, the verb vin 'will come' occupies a higher position, presumably a MOOD-related projection given its future meaning, 24 and thus occurring above both subject-positions.
(46) Las' că vin ei mâine copiii, n-ai leave.IMP.2SG that come.3PLtheytomorrow children.DEF NEG=have.2SGgrijă!(Ro. 2 ) worry 'The children, they'll come tomorrow, don't worry!' (Cornilescu 2000b, p. 101) On the other hand, we have seen that western speakers only have one postverbal (v-related) subject-position, such that they are unable to structurally accommodate strings such as (46), since they lack an EPP-style preverbal subject-position within the sentential core.Our western Daco-Romanian speakers consequently rejected strings such as (46) and reported that the only possibility for ei 'they' and copiii 'the children' to co-occur after the verb would be to interpret copiii as being in apposition to ei, hence the comma intonation following the adverb mâine 'tomorrow' in (47).Thus, the structural representations of ( 46) and ( 47) can be sketched as in (48a-b), respectively.Interestingly, the two possible interpretations have actually been discussed in the literature.For instance, Philippide (1929) argues for the appositional reading of the second subject, while for Byck (1937), the structure does not contain (or no longer contains) any apposition.Nevertheless, analyses like these have never taken into account variation across speakers from different areas and, more importantly, the fact that the availability or otherwise of the double-subject construction should be regarded as epiphenomenal.Indeed, our analysis predicts that both interpretations are found across Daco-Romanian speakers in accordance with the availability or otherwise of an EPP-style preverbal subject-position: while speakers of western varieties clearly prefer the appositional reading, speakers of eastern varieties display true double-subject constructions, in which the first postverbal subject occupies a T-related position (viz.SpecTP) and the second, a v-related position (viz.SpecvP). 25

Daco-Romance
Let us now turn to the situation encountered in other Daco-Romance varieties.Similarly to the situation with the past-tense auxiliaries, the split continues to be between Aromanian and Megleno-Romanian on one hand and Istro-Romanian on the other.As discussed in Section 3.2, both Aromanian and Megleno-Romanian display internal variation with respect to V-movement, with some speakers consistently displaying higher V-movement than others in accordance with diatopic (e.g., Aromanian) and diachronic/diagenerational (e.g., Megleno-Romanian) variation.Now, in the absence of other tests such as the doublesubject construction, it is difficult to assess whether there are one or two available postverbal subject-positions.Nevertheless, our prediction is that, on a par with the case of eastern Daco-Romanian varieties, speakers who raise the verb to a T-related position (namely, Aromanian 2 and Megleno-Romanian 2 ) will also display an EPP-style preverbal SpecTP subject position, since T in these varieties comes with complementary strong V-and Dfeatures.Consequently, examples such as (49a-b) with an unmarked preverbal subject imply higher raising of the verb outside the v-domain and A-movement of the subject to SpecTP to check an EPP feature on T, thus yielding unmarked SV(O) orders.By contrast, in the case of speakers who display low V-movement below the T-area to the lower v-domain (namely, Aromanian 1 and Megleno-Romanian 1 ), where T fails to attract both the verb and the subject, the unmarked (and most frequent) word order is VS(O) (50a-b).This contrast can be seen more clearly in the near-minimal pair in (51), where the pronominal subject situated in SpecTP precedes the temporal adverb tora 'now' in Aromanian 2 (51a), in contrast to the Aromanian 1 example in (51b), where the pronominal occurs in SpecvP, and hence surfaces below the verb and the temporal adverb tora.(Papahagi 1905, p. 164) We now turn to Istro-Romanian.Given the Croatian-style option of high (C-/T-)placement of (Wackernagel) auxiliaries, examples such as (52a) in which there is no overt evidence of a topicalized/focalized reading of the subject, especially since it occurs in an embedded clause where such pragmatic readings are generally less felicitous and less frequent, can be taken to illustrate the default subject position, viz., SpecTP.This is further substantiated by examples such as (27b), where the pronominal subject voi 'you' was seen to occur between the fronted auxiliary åt , 'you have' (variously interpreted to be in C or a very high I-related position) and the temporal adverb acmo 'now', lexicalizing the specifier of TP Past/Future , squarely placing the subject therefore in SpecTP.Similar arguments are proposed by Dragomirescu and Nicolae (2021, p. 8) for strings such as (52b), where the most plausible analysis is to interpret the pronominal subject io 'I' as lexicalizing the SpecTP position, since it precedes the low manner adverb bire 'well', which, in turn, precedes the lexical participle zis 'said' situated on the edge of the v-VP domain.Now, it is not by chance that Istro-Romanian generalized this T-related subject position, since the default subject position in Croatian is also SpecTP (cf.Migdalski 2006, pp. 87-89), from where subjects can raise to a C-related position in order to prevent auxiliaries from appearing clause-initially (52c; cf. also Kovačec 1971, p. 147).Once again this syntactic reflex can be taken to illustrate a further case of PAT(tern) borrowing that comes as part and parcel of PAT(tern) replication of the syntax of Wackernagel auxiliaries.(Migdalski 2006, p. 87) However, we know from our discussion above (cf.Section 3.2) that, alongside the Croatian-style grammar (viz., Istro-Romanian 2 ), speakers also have access to a Daco-Romance grammar (viz.Istro-Romanian 1 ) in which the auxiliary targets a lower position within the I-domain immediately below (Spec)TP Anterior (cf.27a).While under the Croatianstyle grammar, the unmarked word order is Aux + S SpecTP + Ptcp, a priori we would expect speakers adopting the traditional Daco-Romance grammar to adopt the word order Aux + Ptcp + S SpecvP , as illustrated in (53).While, in Croatian, the Aux + Ptcp + S word order in (53) would require a marked reading in which the post-participial subject is interpreted as a case of informational focus (cf.Migdalski 2006, p. 88), presumably occurring in a focus position of the lower left periphery, in the Istro-Romanian 1 grammar no such focus-interpretation of the postparticipial subject is required, since the subject can occur in situ in SpecvP.However, contrary to expectations, quantitatively, orders such as (53) prove extremely rare in all available corpora.Rather, all speakers, even when they employ the low V-movement grammar, typically display the Croatian-style option and generalize SpecTP as the default subject position, 26 a behaviour widely highlighted in the literature (cf.Kovačec 1971, p. 147;Dragomirescu andNicolae 2020, 2021).Consequently, rare orders such as (53) are typically rivalled and replaced by orders like the following, in which the subject, whether pronominal (54a) or lexical (54b), lexicalizes SpecTP (recall that in the Daco-Romance-style grammar, the auxiliary does not raise beyond the head of TP Anterior , and hence is situated below the EPP SpecTP preverbal subject position).( 54 (Morariu 1928, p. 133) The situation with lexical verbs is not any different, in that the low position of the verb, below VoiceP in Istro-Romanian 2 (55a) but above VoiceP in Istro-Romanian 1 but still within the lower adverb space (55b), would lead us to expect once again that SpecTP is not available, since T is inactive in both varieties.However, this prediction is not borne out, inasmuch as the subject, whether pronominal or lexical, invariably lexicalizes SpecTP.This is clear from both examples in (55), where the subject occurs to the left of the very low manner adverb bire 'well', which sits just above the first-merge position of the external argument, namely [ LAS . ..bire [ v-VP [ Spec EA]]], thereby firmly situating the subject in SpecTP (or, alternatively, in a left-peripheral topic or focus position in cases of marked word order).( 55 (Neiescu 2016, p. 51) We therefore witness in Istro-Romanian 1 , as well as in Istro-Romanian 2 in conjunction with lexical verbs, an inconsistent pairing of head and edge features across T and v, yielding a marked hybrid parametric option which variously blends facets of traditional Daco-Romance syntax with Croatian syntax: while v arguably carries both strong V and D features, respectively, correlating with low V-movement and (a certain degree of) DOM, T, inconsistently, also comes with a strong D edge feature exceptionally licensing a dedicated EPP-style preverbal subject position in SpecTP, despite the failure of T to probe the verb.While this state of affairs is unexpected, given our canonical assumptions about the consistent pairing of head and edge features across the functional heads T and v, this particular outcome is not atypical of cases of language contact in which exogenous forces, in this case the replication of a Croatian-style generalized SpecTP subject position, can produce non-linear, and otherwise apparently incongruous, results. 276.Differential Object Marking 6.1.Daco-Romanian Ledgeway (2020, sct. 4.4) observes that northern Romance varieties form a compact group, since they typically do not formally distinguish between different classes of direct object and, in particular, between animate and inanimate specific objects (56a).By contrast, southern Romance varieties formally distinguish between animate and inanimate specific direct objects (56b), realizing the former, but not the latter, with a differential marker, a case of differential object marking (DOM).( 56 Assuming DOM to be a reflex of an object shift-like operation which raises the object to SpecvP (Torrego 1998;Ledgeway 2000Ledgeway , 2020, sct. 4.4;, sct. 4.4;2022a;Peverini-Benson 2004;Andriani 2011Andriani , 2015;;López 2012;Cornilescu 2020;Tigău 2021), the relevant difference between northern and southern Romance can be simply captured in terms of the respective absence and presence of a D-feature on v to probe the object.In turn, the relevant content of this D-feature in southern Romance shows considerable cross-dialectal microvariation in relation to such properties as person, number, animacy and specificity, which variously restrict the ability of v to probe different subclasses of nominal (cf.Ledgeway 2023aLedgeway , 2023bLedgeway , 2023c)).Arguably, then, the settings for the V-movement parameter which, respectively, licenses V-to-T movement in northern Romance varieties and V-to-v movement in southern Romance varieties is also implicationally correlated with the distribution of a subject clitics and the availability of a preverbal EPP subject position (SpecTP) on the one hand and DOM on the other.In particular, the activation of T and v through a strong V-feature equally licenses a corresponding strong D-feature in their edge, in the same way that the activation of generalized V-movement to C in V2 varieties licenses a generalized edge feature satisfied through XP movement.
Now, if V-to-T movement is bundled with a strong D-feature in the TP-edge, hence licensing the so-called double-subject construction in eastern Daco-Romanian varieties, then lower V-movement in western Daco-Romanian varieties is predicted to license a strong D-feature in the vP-edge and, hence, DOM.Deliberately oversimplifying, 28 Daco-Romanian varieties typically display DOM with animate referents (cf.also ALR (VI) 1969, map 1642, 1704), marking, for example, (personal) pronouns (57a), proper names (57b) and definite DPs (57c) with the preposition p(r)ă/pe 'on'.(Marin 2023, p. 310) Nevertheless, a series of tendencies and preferences in the distribution of DOM may potentially be explained through differences in V-movement (cf.Ledgeway 2020, pp. 59f.).For example, standard Romanian grammars usually discuss the case of 'optional contexts' (cf. Pană Dindelegan 2013, pp. 133f.), in which the absence of DOM does not lead to ungrammaticality, including when the direct object is realized by an indefinite, interrogative or floating quantifier.When examined in greater detail, however, such cases of apparent optionality turn out to be epiphenomenal, inasmuch as speakers who display lower Vmovement and, crucially, interpret the double-subject construction as an appositional structure employ DOM with a much higher frequency than speakers who have higher V-movement.In short, speakers of western varieties clearly prefer-indeed require-DOM when the direct object is introduced by an indefinite (58a), interrogative (58b) or floating (58c) quantifier.( 58 (Marin 2023, p. 449) By contrast, eastern varieties, which display higher V-movement, do not generally employ the marker pă/pe in these same so-called optional contexts, witness the absence of DOM in (59a-c).( 59 In conclusion, the parallel assumption that the activation of T and v automatically comes with a strong D edge feature not only predicts the differential marking of subjects and objects in northern and southern Romance varieties, respectively, but also provides a natural explanation for significant differences between eastern and western Daco-Romanian varieties.In particular, we now have a principled explanation for the more consistent distribution of DOM in western Daco-Romanian than in eastern Daco-Romanian, its distribution falling out as a concomitant of the setting of the T/v parameter.However, while we witness in western varieties a complementary and consistent distribution of head and edge features on T and v, in that head and edge features on T are weak but strong on v, in turn, correlating with unmarked VSO order and a robust distribution of DOM, in eastern varieties T arguably carries both strong V and D features (manifested in high V-movement and unmarked SVO order), but, inconsistently, v also comes with a strong D feature that exceptionally licenses a less consistent distribution of DOM.Indeed, this marked setting of the DOM subparameter in eastern Romanian, in contrast to its unmarked setting in western Romanian which combines a positive edge feature on v with a corresponding positive head feature on the same, finds independent support in recent proposals by Cornilescu (2020, pp. 127-29) and Tigău (2021).They argue that DOM-ed and unmarked objects exhibit the same distribution in Romanian, such that DOM-ed objects (= KPs) necessarily scramble to the vP periphery, while this is merely an option for unmarked objects (= DPs), which may either scramble or remain in situ.Presumably, such optionality underlies the more marked nature of the distribution of the D edge feature seen in eastern Daco-Romanian varieties (viz., Romanian 2 ), where the relevant nominals may be licensed in situ on a par with non-DOM grammars, in which v is specified [-D].In western Daco-Romanian varieties (viz.Romanian 1 ), by contrast, the unmarked [+D] specification on v aligns with a more consistent licensing of the relevant nominals which invariably undergo scrambling and surface with DOM.
Indeed, this view finds direct empirical support in the following word order contrasts.In answer to the question in (60), a thetic response involving an animate subject and inanimate object, in which none of the arguments is presupposed or topicalized, invariably gives rise to unmarked VSO word order (60a), both among western speakers and, optionally, among eastern speakers who also allow SVO word order in such contexts.Interestingly, however, if the object is also animate and hence a potential candidate for DOM, the unmarked order found among eastern speakers remains VSO (60b), but is replaced by unmarked VOS among western speakers (60c).( 60 The contrast between the respective unmarked eastern and western orders in (60b-c) transparently betrays the availability or otherwise of object shift with DOM-ed objects in each variety.In eastern varieties, the object pe Maria remains in situ, as in the case of inanimate unmarked objects such as icoana in (60a), yielding VSO order in which the subject Ion occurs in its base SpecvP position, and hence to the left of the object.In western varieties, by contrast, the object pe Maria is probed by a D-feature on v causing it to raise to an outer specifier position immediately above the subject in SpecvP, thereby placing it to the left of Ion and yielding the derived VOS order.
Discussing Aromanian, Capidan (1925, p. 203) remarks that 'true Aromanian does not employ pre', concluding that the presence of DOM attested in available Aromanian corpora (e.g., Papahagi 1905 collection of Aromanian fairy tales) is nothing more than an artificial phenomenon possibly influenced by standard Daco-Romanian norms (Capidan 1932, p. 530).Nonetheless, a few lines later, Capidan himself admits that other interpretations cannot be completely dismissed, given that instances of DOM are found in eighteenthcentury Aromanian (61), leading him to conclude that DOM is a genuine feature of some, but crucially not all, Aromanian varieties.According to him, DOM is found in southern Aromanian dialects, but is completely absent in northern Aromanian and the varieties spoken in Albania.
(62) Io va o nt , ap la ureacl'a ndreapta pri I AUX.FUT.1SGher= sting.1SGat ear.DEF right(DEF) DOM Mus , ata-Loclui.(ARo. 1 , Samarina) Mus , ata-Loclui 'I will sting Mus , ata-Loclui on her right ear.' (Papahagi 1905, p. 196) (63) U m'esku pri nve'astă tu-ać'a o'ară.(Aro 1 ., Kranea/Turya) her= cherish.1SGDOM wife in=that instant 'Then I cherish my wife.' (Sobolev 2008, p. 117) Although the presence of DOM may seem fortuitous, given that there were Romanian schools between 1884 and 1944 in Aromanian villages in (Greek) Macedonia (cf.Rosetti 1968, p. 535), the retention of DOM in this area must be explained on structural grounds directly related to the T/v parameter.In particular, we have seen that these southern varieties display a lower level of V-movement licensed through a strong V-feature on v, a parameter setting which, in turn, correlates with the generalization of a single perfective auxiliary HAVE (unlike northern Aromanian varieties which display an emerging HAVE/BE active-stative auxiliary split in compound tenses thanks to a higher movement of the verb to the T-domain) and the licensing of DOM on account of a strong D feature on v. 29 In a similar fashion, modern low V-movement varieties of Aromanian spoken in Northern Macedonia, mainly in the Ohrid-Struga region (predominantly inhabited by speakers of the southern Gramostian variety), though not in the Krushevo-Bitola area, also display productive DOM with pronouns (64a) and definite common nouns (64b), 30 a syntactic feature also found in Macedonian dialects spoken in the same area (Tomić 2006, p. 184;Bužarovska 2017, p. 67;Makarova and Winistörfer 2020;Hill and Mardale 2021, p. 31 There is significantly less variation in Megleno-Romanian varieties, in that DOM now appears to be extinct (cf.also Hill and Mardale 2021, p. 32).Nevertheless, although Capidan (1925, pp. 203f.)maintains that Megleno-Romanian lacks DOM, his own corpus does display some, albeit infrequent, examples of DOM.Although the reduced number of attestations in his corpus makes it difficult to arrive at any firm generalizations, it appears that speakers with lower V-movement (for example, those from Huma/Uma) are more likely to use DOM (65) than speakers from other areas who display higher V-movement.
(65) S , i au muri s , i pri tsea mul'ari.(MRo., Huma/Uma) and her= kill.PRT.3SGtoo DOM that woman 'And he killed that woman too.' (Capidan 1925, p. 116) Alongside DOM proper, it must be noted that, on a par with their respective contact languages, both Aromanian (66a) and Megleno-Romanian (66b) also make use of clitic doubling to mark topical direct objects (Hill and Mardale 2021, pp. 29-33; for Aromanian, see also the article by Savoia and Baldi in this special issue).In this respect, language contact can be seen as one of the main causes for the loss of pe/pi-marking, given that both Greek (67a) and Macedonian (67b) mark (animate and inanimate) topical direct objects through clitic doubling.Thus, in the history of Aromanian and Megleno-Romanian, what seems to have changed are the features involved in differential object marking, whether realized through the preposition pe/pi or clitic doubling: while older stages arguably favoured Romance-style DOM highlighting the centrality of specificity and animacy, more recent stages have seen a gradual increase in the frequency of Balkan-style clitic doubling that singles out direct objects in accordance with their topicality.(Tomić 2006, p. 252) Turning finally to Istro-Romanian, DOM has been claimed to be either entirely absent (Hill and Mardale 2021, p. 34) or only marginally present (Geană 2020) in this sub-branch of Daco-Romance.From a diachronic perspective, DOM seems to be optionally employed in recordings from 1928-1933 (68a;cf. Traian Cantemir's Texte istroromâne and Leca Morariu's Lu Frat , i Nos , tri), 32 but becomes extremely rare, if present at all, in recordings from 1982-1996 (68b;cf. Richard Sârbu and Vasile Frăt , ilă's Dialectul istroromân), before disappearing entirely in recordings from 2009 (68c; cf.Preservation of the Vlaški and Žejanski language project (VlaŽej)).However, unlike Aromanian and Megleno-Romanian, in which Romance-style DOM is being gradually replaced by clitic doubling where topicality is now regarded as the main trigger for differential object marking, Istro-Romanian shows no such tendencies.From a structural point of view, Istro-Romanian is expected to display Romance-style DOM, inasmuch as lexical verbs do not target high positions within the IP such that v is predicted to license a strong D-feature in its vP-edge.However, while lower V-movement strongly favours the use of DOM, Istro-Romanian DPs have progressively been losing their ability to overtly encode definiteness/specificity, which is similar, though not identical, to the situation found in Croatian. 33For example, under the pressure of Croatian feminine nouns, which end in -a (69a), Istro-Romanian feminine nouns, e.g., cåsa 'house.DEF' (69b), also began to license an indefinite reading, namely 'house'.From a structural perspective, the Istro-Romanian D-head containing both the noun and definite suffix was reanalysed as an N-head (68c) underspecified for definiteness. (69 [ NP [ N' cåsa]]] (IRo., after reanalysis) house '(a/the) house' Given that DOM is intrinsically linked to definiteness/specificity as formalized in various referential hierarchies (cf.Aissen 2003; Witzlack-Makaraevich and Seržant 2018, pp.5-7), a direct consequence of the loss of transparent encoding of definiteness/specificity on Istro-Romanian nominals is manifested in the gradual loss of DOM.This development presumably first affected the licensing of DOM with DPs containing non-specific, indefinite specific, and definite nominals (cf.Bossong 1991) which were not intrinsically definite or specific.This stage is already well attested in Cantemir's and Morariu's corpora, where such referents overwhelmingly occur without the pre marker.Subsequently, DOM would have no longer served to distinguish between specific and non-specific referents, eventually leading to its progressive loss, even with proper names and personal pronouns which are intrinsically specific.Thus, it is not by chance that in Sârbu and Frăt , ilă's (1998) Dialectul istroromân there are only a handful of (perhaps dubious) attestations of DOM, all of which involve personal pronouns, the last class of nominals to lose DOM.Indeed, by the time of the first recordings of the Preservation of the Vlaški and Žejanski Language project initiated in 2007, DOM had completely disappeared, a development which is matched by a concomitant reanalysis of erstwhile definite nouns/adjectives which are now generally, though not exclusively, interpreted as being underspecified for definiteness.

Postverbal Negators
A final significant consequence of the difference in the ability of T to probe the verb that we now explore is the prediction that it makes about the distribution of different negation strategies.As is well known, 34 in the dialects of central, southern and northeastern Italy, Ligurian, eastern Romansh, Catalan, European Portuguese, Spanish and Romanian, sentential negation is marked by a simple reflex of Latin preverbal NON 'not', so-called Stage I of Jespersen's Cycle (70a).Elsewhere, negation is either at Stage II, as in many northern Italian dialects, standard (written/formal) French, and Gascon, where negation is expressed discontinuously by both a preverbal and postverbal negator (70b), or at Stage III, as in many north-western Italian dialects, western/central Romansh, spoken French, Occitan, Aragonese, and northern Catalan dialects, where negation is expressed by a single postverbal negator (70c).In light of these distributions, Ledgeway and Schifano (2023) propose an original connection between each of these three stages and the extent of V-movement.In particular, varieties with simple preverbal negation (Stage I) may display either low V-movement (e.g., Cosentino) or high V-movement (e.g., some northern Italian dialects such as most Ligurian and Venetan varieties), whereas in varieties with non-emphatic postverbal negators (Stages II-III), the verb must raise to a high position (e.g., French and most northern Italian dialects).The relevant empirical generalization then is as in ( 71), inasmuch as we see that the presence of a non-emphatic postverbal negator is not compatible with low V-movement.Rather, a postverbal negator can only be licensed if the verb raises through the projection hosting the negator (for a more detailed technical implication of this basic idea, see Ledgeway and Schifano 2023, sct. 6).It follows that high(er) V-movement is essential for the licensing of Stages II-III of Jespersen's Cycle, and that only in the modern dialects of northern Romance, where we have independently established that T probes V, does postverbal negation obtain.By contrast, in southern Romance where V-movement is low, all varieties are necessarily at Stage I.The inescapable conclusion therefore is that Jespersen's Cycle is inextricably tied to the relative height of V-movement.Significant in this respect is Daco-Romanian, which, as repeatedly noted above, offers us evidence of both low (western) and high (eastern) Vmovement grammars, but which in both cases uniformly displays Stage I negation (72a-b), highlighting how high V-movement is a necessary but not an automatic trigger for Stage II or III negation.
Bas , că zâs , e că nu-i pasă.(Ro 1 ., Checea, Timis , , Ban.) b.Spune intent , ionat că nu-i pasă.(Ro. 2 , Cumpăna, CT , , Dobr.) say.3SG deliberately say.3SG that NEG=DAT.3=matter.3SG'He deliberately says that he doesn't care.'However, an important synchronic prediction of our approach to the distribution of Jespersen's Cycle is that, if the empirical generalization in ( 71) is correct, signs of grammaticalization of new postverbal sentential negators should only be detected in varieties which exhibit high V-movement.A very revealing test case is represented by colloquial varieties of eastern Daco-Romanian, where our investigations have revealed a productive use of n-words employed as sole (increasingly non-emphatic) negators, suggesting an ongoing shift towards Stage III negation (cf.discussion of northern Italo-Romance in Ledgeway and Schifano 2023, sct. 5.2.2) As predicted, examples like those in (73a-f) are only found in high V-movement varieties where the placement of the verb above the LAS may license new postverbal negators.Western Daco-Romanian speakers, by contrast, categorically reject all such examples, which simply do not form part of their grammars, which fail to show any signs of grammaticalizing n-words, a conclusion entirely in line with our observation that all western varieties display low V-movement.In short, our original independent observations about differences in V-movement in western and eastern Daco-Romanian turn out once again to have more general explanatory power, in this case providing a principled account for the restriction of incipient cases of grammaticalization of postverbal negators to Daco-Romanian varieties with high V-movement. 35 In the absence of relevant data, we leave it to future research to determine whether similar differences in negation strategies can be identified in the other sub-branches of Daco-Romance in accordance with their particular setting of the V-parameter.

Conclusions
We have reviewed robust parametric evidence both from within Daco-Romanian and across Daco-Romance to support some major differences in the height of V-movement in accordance with diatopic (e.g., Daco-Romanian, Aromanian), inter-speaker and diagenerational (e.g., Megleno-Romance) and language-endogenous/-exogenous (e.g., Istro-Romanian) factors.At the same time, we have observed how parameters do not operate in isolation, inasmuch as the setting of one parameter entails significant knock-on effects elsewhere in the grammar.Such is the case with T, which probes the verb in eastern Daco-Romanian, northern Aromanian, optionally in 'traditional' Megleno-Romanian 1 , and auxiliaries in Croatian-style Istro-Romanian 2 , a parametric setting which, in turn, accounts for V-Adv orders; (incipient) active-stative perfective auxiliary selection and the tendential loss or weakening of the preterite (though not in Aromanian and Megleno-Romanian); a dedicated preverbal subject position licensing unmarked SVO; the (near) absence or heavily reduced distribution of DOM; and incipient Stage III negation.By contrast, in western Daco-Romanian, southern Aromanian, 'modern' Megleno-Romanian 2 , and in conjunction with lexical verbs in traditional Daco-Romance-style Istro-Romanian 1 , all these same options are largely absent, since T fails to attract the verb, which is probed instead by v.This parametric setting explains the presence in these varieties of Adv-V orders; the generalization of a single perfective auxiliary in accordance with a nominative-accusative alignment, and, with the exception of Istro-Romanian, the retention of the preterite; the absence of a dedicated preverbal subject-position, yielding unmarked VSO, though not in Istro-Romanian, which, under the exogenous influence of Croatian, has generalized the preverbal subject-position; a robust and productive distribution of DOM, except in Istro-Romanian, where contact-induced changes in the marking of definiteness and specificity have arguably led to its attrition; and robust Stage I negation.These facts are summarized in Table 2. Daco-Romanian and the other Daco-Romance varieties, including written texts (e.g., plays, fairy tales), annotated collections of oral dialogues, linguistic atlases, grammars and individual linguistic studies.
6 Indeed, one anonymous reviewer also notes that in their dataset from the Fier area in Albania, a northern Aromanian variety, verbs invariably precede adverbs, as shown by the following examples.Evidence such as (i) below might suggest that lexical verbs actually target a position in the lower portion of the higher adverb space, witness the position of the finite verb avemo to the left of the lower temporal adverb "vet" 'already'(Spec TP Anterior ) and, in particular, to the left of acmo 'now' (SpecTP Past/Future ).However, the example involves the verb HAVE, which even in its copular possessive uses is probably more appropriately considered a functional predicate, which, as we shall see below, is a class which independently targets higher positions than lexical verbs.For Croatian, see Radanović-Kocić (1988), Franks and Progovac (1994), Rivero (1994Rivero ( , 1997)), Wilder and Ćavar (1994), Schütze (1994), Dimitrova-Vulchanova (1995), Halpern (1995), Progovac (1996), King (1996), Tomić (1996), Embick and Izvorski (1997), Franks (1998Franks ( , 2000)), Bošković (2001), andRoberts (2010).9 This account potentially explains contradictory data found in the literature.For example, Giusti and Zegrean (2015, p. 126) observe that the Istro-Romanian auxiliary cannot occur in clause-initial position, while Dragomirescu and Nicolae (2018, pp. 96f.) highlight an example where the auxiliary is found in clause-initial position.Significant in this respect is Dragomirescu and Nicolae's (2018, p. 97)  Here we gloss over the precise details of the derivation of the inverted structure and, in particular, how the participle comes to precede the auxiliary.In the corresponding southern Slavonic construction, the inverted structure is typically explained by positing independent movement of the lexical verb over the auxiliary to SpecTP or the C-domain (Migdalski 2006, pp. 62-81).According to Rivero's (1992Rivero's ( , 1994) ) classic analysis across a variety of Balkan varieties, such structures involve long head movement of the lexical verb to C over the auxiliary.It is not clear, however, that the Megleno-Romanian inverted construction can be readily equated with that found in southern Slavonic since, for example, clitics and negation occur immediately before the participle+Aux cluster in the former, but not in the latter, in which clitics occur between the participle and the auxiliary.
12 When the auxiliary verb exceptionally raises to T in the south, as can happen in irrealis contexts, active-stative effects may persist, as demonstrated in Ledgeway (2022b) for Ibero-Romance.It is not by chance that western Daco-Romanian employs fi 'be' with transitive participles.When the fieldwork for ALR (VI), 1997 was carried out, parts of western Romania, predominantly from Banat, still spoke Serbian.As a result, some speakers used Serbian ia sam pevo (lit.'I be.1SG sing.PTCP') to render Daco-Romanian am cântat (lit.'have.1SGsing.PTCP') 'I sang/have sung'.14 Crucially, when testing the grammaticality of this type of construction the periphrasis was always used in conjunction with temporal adverbials such as de trei zile 'three days ago' and de ieri 'since yesterday', which force a resultative reading.Finally, it must be noted that structures containing fi 'be' and a past participle can also be found, albeit extremely rarely, in the west.However, unlike in the east, in such uses the participle assumes a purely adjectival nature (cf.also Iordan 1973, p. 405;Coteanu 1982, p. 169;Avram 1994, pp. 494, 506), as shown by the possibility in (i) of coordinating it with other adjectives.
(i) Fata e căzută pe scări s , i plină de sânge.(Râmnicu Vâlcea, Vâlcea, Olt.) girl.DEF be.3SG fall.PTCP.FSG on stairs and full.FSG of blood 'The girl is fallen (i.e., lying) on the stairs and completely covered in blood.' 15 There are other explanations in the literature that have been put forward to explain the loss of the preterite in the east and the standard.For example, S , iadbei (1930) and Frâncu (1967Frâncu ( , 1984Frâncu ( , 2009) ) see the gradual replacement of the preterite with the analytic past as a result of homonymies in the relevant verb paradigms, e.g., ORo.preterite cântămu 'we sang' vs present cântămu 'we sing'.Now, even if true-but cf.Sp. preterite cantamos 'we sang' and present cantamos 'we sing', and, similarly, BrPt.preterite cantamos 'we sang' and present cantamos 'we sing', where the preterite has not been lost despite some formal homophony with the present-it must be noted that, while in the east the tendency was to lose these forms altogether, in the west (initially in Oltenia and then in Banat and Cris , ana) there arose innovative preterite forms containing the -ră-suffix, e.g., ORo.cântămu 'we sang' > ModRo.cântarăm 'we sang/we have sung '. 16 The preterite is also used to refer to past events without present relevance in Maramures , (e.g., in Transcarpathia, Ukraine), as exemplified in (i).name.DEF his 'His name was Victor.' (Marin et al. [1866(Marin et al. [ ] 2000, p. 81) , p. 81) 17 There also seems to be idiolectal and nanoparametric (viz., lexical-based) variation among speakers from Cris , ana.For example, our informants from S , us , tiu (Bihor, Cris , ana) showed a clear preference for the use of the preterite with (more functional) verbs such as fi 'be', veni 'come', and merge 'go'.However, as shown in (i), other (more lexical) verbs can also be found in the preterite, albeit more rarely.This development, described by Squartini and Bertinetto (2000, p. 418) as 'a peculiar form of aoristic drift', has been shown to be a relatively recent development (cf. Pană-Boroianu 1982).More specifically, while the form per se undoubtedly has a Latin etymon, the semantic development and hodiernal specialization of the form mirror those found in some neighbouring Balkan (Slavonic) varieties such as Serbian and (Štokavian) Croatian (Babić 1992, p. 265;Lindstedt 1994, pp. 36f.).By way of illustration, both the Oltenian example in (i.a) and the (Štokavian) Croatian in (i.b) can only be interpreted as referring to events that took place in the immediate past, but not for more distant events, witness the ungrammaticality of the use of the preterite form in the Oltenian example (ii) for non-hodiernal events.In this respect, it is important to note that at the beginning of the nineteenth century south Oltenian regions were populated by a significant number of 'Serbians' (e.g., 1242 families in 1831), an umbrella term for Slavswho had fled from the Ottoman Empire.Indeed, the same territories continue to be populated by Slavonic peoples even today: Macedonians represent the overwhelming majority in Brebeni Sârbi (Olt county), while Serbian is still spoken in Băiles , ti (Dolj county).

21
The interpretation of auxiliary fi 'be' in conjunction with unaccusatives as an emerging syntactic tendency rather than an old Daco-Romance relic is supported by both empirical and theoretical arguments.As Geană (2017, p. 213) notes, 'there is no consistent use of the compound past with the auxiliary fi 'be' (if any)'.Indeed, if fi 'be' had been preserved in Istro-Romanian from earlier stages of the language displaying an active-stative split, then there is no straightforward explanation for the inconsistent use of the auxiliary in modern Istro-Romanian.If anything, the influence of Croatian, which exclusively displays BE in the analytic past periphrasis, should have enhanced the use of fi 'be' with unaccusative verbs.Rather, the use of auxiliary BE in conjunction with unaccusatives must represent an innovation arising from an imperfect replication of Croatian-style higher placement of the auxiliary which feeds unaccusative reflexes in the auxiliary system in accordance with (33).

22
As argued in detail in Ledgeway (2020, sct. 3.3;2022a, sct. 2.3), in northern Romance varieties more generally, which also show high V-movement, the strong D-feature on T is manifested in the grammaticalization of a dedicated preverbal SpecTP subject position, leading to a reversal in the pro-drop parameter in French and some (northern) Occitan varieties, which is supplemented in northern Italian dialects, Raeto-Romance (including Ladin) and some northern Occitan varieties by the overt spell-out of the strong D-feature of the subject through a (partial/complete) system of subject clitics.This explains the contrast between the availability of unmarked SVO word order in the north and VSO order in the south, where T is inactive and lacks an EPP-style D-feature that probes the subject.

23
This lower v-related subject position is compatible with both a (narrow) focus reading, as in (45c), as well as a topicalized reading, as in (i).We leave open here the question of whether such interpretations are licensed in situ in SpecvP or whether they involve raising of the subject to a focus or topic position in the lower left periphery (cf.Belletti 2004) Indeed it is not by chance that the verb in the double-subject construction typically carries an irrealis interpretation (e.g., future, subjunctive) since there is independent evidence across Romance that irrealis verb forms raise to an even higher position than in realis uses (cf.Ledgeway 2009aLedgeway , 2013Ledgeway , 2015Ledgeway , 2020Ledgeway , pp. 38-40, 2022Ledgeway , 2023d;;D'Alessandro and Ledgeway 2010;Ledgeway and Lombardi 2014;Schifano 2018, pp. 42-51, 96-113, 237f.;Groothuis 2020, pp. 199f.).For example, of the thirty examples of the double-subject construction reported in Cornilescu (2000b), there are twenty-four examples of the 'present' with future interpretation, one example of the subjunctive involving V-to-C movement, and one example of a gerund with an irrealis reading; the remaining four examples involve what is arguably a distinct construction labelled by Cornilescu as an example of 'standard familiar Romanian' (cf.i), in which there is a gender mismatch between the postverbal pronominal subject (always in the default masculine singular) and the following lexical subject, and the verb has a realis interpretation (e.g., present, past).From this, it is clear that licensing of the double-subject construction requires an exceptional movement of the verb to an irrealis modal position situated within the highest portion of the sentential core, or even the C-domain.Orthogonal to our argument about the availability of one or more subject positions within the sentential core is the manner in which genuine double-subject constructions are actually derived.Nevertheless, one possible analysis, given the order of pronoun + lexical DP, is to argue for a 'big DP' analysis (cf.object clitic-doubling structures) in which the pronominal part of the underlying big DP floats off to the SpecTP position, stranding the lexical DP in situ.For further details, see Cornilescu (2000b). 26 Note that this tendency to generalize SpecTP is not sensitive to a preverbal AGENT vs postverbal UNDERGOER active-stative distinction, witness the example in (i), where the UNDERGOER subject typically occurs in SpecTP.

29
Apart from the north-south divide regarding the presence or otherwise of DOM, Aromanian represents an interesting case from a diachronic perspective.Simplifying somewhat, the Aromanian spoken in Albania, argued Capidan (1932, p. 530), completely lacked DOM, whereas Manzini and Savoia (2018, pp. 167-69) remark that, in the variety spoken in Diviakë (Albania), 'DOM with highly ranked objects is possible, but not enforced'.Although an in-depth analysis of V-movement in the Aromanian varieties spoken in Albania is required in order to identify the exact triggers of this development, it would seem that in the last 100 years there has been a (re)emergence of DOM.As expected, the first constituents to be marked with pe/pi are highly definite (personal) pronouns (i).(Manzini and Savoia 2018, p. 168) 30 An ongoing tendency in these territories is the optional use of pi with definite inanimate nouns to signal topicality, which is similar to what is found in Macedonian (Bužarovska 2017, pp. 78f.).31 DOM, viz., na-marking, seems to be a recent phenomenon in Macedonian, given its absence in Daniel's late eighteenthcentury Tetraglosson (cf.Nichev 1997).Interestingly, nineteenth-century Aromanian texts from the Struga region also lack DOM (Bužarovska 2017, p. 71), a conclusion that overlaps with Capidan's (1932, p. 530) observations regarding northern Aromanian varieties.On this note, Tošev (1970) argues that Aromanian refugees from Moschopolis/Moscopole (Albania) brought this syntactic feature to the Struga area (though cf.Friedman 2000 for a different explanation that relies on the effects of standard Daco-Romanian being taught in schools).32 Such optionality is probably only apparent, inasmuch as the distribution of DOM appears to be determined by diatopic factors, with virtually all the examples of DOM attested in Cantemir's and Morariu's corpora coming from the southern Šušnjevica, Noselo/Nova Vas and Sucodru/Jesenovik varieties; cf.Cantemir (1959, pp. 39, 60, 93, 105) and Morariu (1928, pp. 5, 7, 38, 68, 69).

33
While in older stages of Croatian, the definite/indefinite distinction was marked through adjectival forms, with long forms marking definite DPs and short forms indefinite DPs (cf.Aljović 2000, pp. 28f.;Sussex and Cubberley 2006, p. 266), in modern Croatian the system is slowly falling into disuse, with speakers lacking sharp grammaticality judgements in relation to long and short adjectival forms (Velnić 2015, pp. 29f.).In a similar development, some prenominal Istro-Romanian adjectives bearing the definite article can convey the same meaning as prenominal adjectives without the definite article (cf. also Kovačec 1984, p. 567;Geană 2019, p. 65).For example, the DP in (i.a) mårle såd was translated by Sârbu and Frăt , ilă (1998, p. 270) as an indefinite (cf.måre 'big'), despite the formal presence of the definite article -le.By the same token, in the recordings from the Preservation of the Vlaški and Žejanski Language project, begun in 2007, the DPs in (i.b) were translated as indefinite despite being formally marked as definite.See Zanuttini (1997), Parry (1997), Manzini and Savoia (2005, III, pp. 127-55), Poletto (2008, 2016a, 2016b), Garzonio andPoletto (2009, 2018), and the various Romance-specific chapters in Willis et al. (2013). 35 Although this requires further research, it is entirely plausible that this difference between western and eastern varieties of Daco-Romanian may also explain the fact that the preverbal negator nu from Lat. NON, which typically produces clitic reflexes in preverbal position across Romance (cf.Fr. ne + V vs constituent negator non (pas)), is frequently stressed in western varieties (i.a) but not in eastern varieties, where it typically surfaces as an unstressed clitic (i.b).This can be clearly seen from the atlases; witness the realizations of nu mi-a păsat 'I didn't care' in (ALR (VII) 1972(VII) , map 2003) ) where the negator predominantly occurs in stressed form in the west but as an atonic clitic in the east.In particular, it is well-known that high V-movement varieties of northern Romance that have developed Stage II-III negation display considerable weakening of the original preverbal negator, ultimately leading to its complete loss in Stage III varieties.Given the incipient tendencies towards the grammaticalization of postverbal negators in eastern Romanian varieties observed in (73a-f), it is not therefore surprising to find some prosodic weakening of the preverbal negator in the east, but, significantly, not in western varieties where, given their Stage I status licensed by a low V-movement grammar, the negator can in fact occur in stressed form, and frequently does.
or BE (32b), or, alternatively, by a person-based system, as in (32c), which generally contrasts BE in the first and secondperson with HAVE in the third person.

Table 2 .
V-parameter and its consequences across Daco-Romance.
. have.1SGspeak.PTCP today with Mihai but me.ACC= ring.3SG he when arrive.3SG home 'I haven't spoken today with Mihai, but he'll call me when he gets home.' NEG huge dog came in front of him' (VlaŽej, Grandfathers' interesting stories)