Changing alignments in the Greek of southern Italy

This article investigates a peculiar pattern of subject case-marking in the Greek of southern Italy. Recent fieldwork with native speakers, coupled with the consultation of some written sources, reveals that, alongside prototypical nominative subjects, Italo-Greek also licenses accusative subjects, despite displaying a predominantly nominative-accusative alignment. Far from being random replacements within a highly attrited grammar, the distribution of these accusative subjects obeys specific structural principles, revealing similarities with historical attestations of the so-called ‘extended accusative’ in early Indo-European. On the basis of these data, Italo-Greek is argued to be undergoing a progressive shift towards an active-stative alignment, a claim supported by additional evidence from auxiliary selection, adverb agreement and sentential word order.


Introduction
Greek has been spoken as an indigenous language in southern Italy since ancient times (Falcone 1973:12-38;Horrocks 1997:304-306;Manolessou 2005:112-21;Ralli 2006:133).According to one, albeit now unpopular, view championed most notably by Rohlfs (1924;1933;1974;1977), the Greek spoken in southern Italy, henceforth Italo-Greek, is to be considered a direct descendant of the ancient (mainly Doric) Greek varieties which were imported into Magna Graecia as early as the eighth century B.C.E. with the establishment of numerous Greek colonies along the coasts of southern Italy.The opposing -and now widely accepted -view, argued most vehemently by Battisti (1924;cf. also Morosi, 1870;Parlangèli, 1953), sees the Greek of southern Italy as a more recent import dating from the Byzantine period of domination between the sixth and eleventh centuries.However, as argued by Fanciullo (1996;2001;2007), these two apparently opposing views can be reconciled if we accept that Italo-Greek is largely a Byzantine import preserving some ancient Doric features, a view further supported by Ralli (2006:134) who argues that ' [Italo-Greek] preserves some traces of an ancient Doric substratum, which could point to the continuous uninterrupted presence of Greek speakers in South Italy' (cf.also Squillaci 2017:7-9;Ralli in press).Whatever the correct view, it is clear that by the beginning of the second millennium C.E. Greek was still widely spoken as a native language in north-western Sicily, Calabria and Apulia.Indeed, as late as the fourteenth century Petrarch is reported to have advised those wishing to study Greek to go to Calabria.
In what follows we shall focus on one feature of the syntax of Italo-Greek which has to date gone unnoticed in the literature and which we believe is otherwise unattested in other modern dialects and varieties of Greek outside of Italy. 3 The phenomenon in question concerns the possibility of marking a subset of surface subjects with accusative case. 4 A careful analysis of such attestations reveals that accusative-marked subjects cannot be disregarded as random replacements within a highly attrited grammar but, rather, obey regular structural principles that underlie an ongoing progressive shift towards an active-stative syntactic alignment.It is our contestation that this change in the alignment of Italo-Greek is the result of contact with Romance where reflexes of an active-stative alignment are otherwise abundantly attested.
The article is organized as follows.After providing a brief introduction to some basic concepts in the general description of morphosyntactic alignments ( §2), we briefly consider the distribution of case-marking and formal splits in the verb system of Standard Modern Greek ( §3) and their differing characterizations in terms of alignment.This is followed by an examination of the fundamental properties and distribution of the so-called 'extended accusative' in early Indo-European ( §4), which we subsequently compare with the distribution of accusative subjects in the Italo-Greek varieties of Griko ( §5.1) and Greko ( §5.2) which are shown to follow an emerging active-stative alignment.In support of this analysis, the following sections ( § §6.1-3) review further evidence from Italo-Greek for the emergence of morphosyntactic reflexes of an active-stative alignment.The final section ( §7) summarizes the results and offers some general conclusions and remarks about the nature and role of Romance-Greek contact in shaping the grammars of Italo-Greek in southern Italy.

2
Alignments: some preliminary observations Before looking at the details of accusative subjects in the Greek of southern Italy, we must first review some basic concepts and distinctions about morphosyntactic alignments which will prove essential in our discussion of Italo-Greek below.Following a widely-accepted typological distinction (Dixon 1994:6-8; see also Comrie 1989:110-116), we can distinguish three core sentential participants labelled A and O (1a), the subject and object, respectively, of a transitive construction, and S (1b-c), the subject of an intransitive construction: (1) a. John (A) was smoking a cigarette (O).b.John (S) was smoking.c.The gun (S) was smoking.
In a number of areas of their grammars, many languages make a further distinction between two types of intransitive S(ubject): (i) an S with an agentive interpretation (1b) and hence, to all intents and purposes, identical to A(gent), bar the presence of an O(bject); and (ii) an S with an UNDERGOER interpretation (1c) and hence, to all intents and purposes, identical to O(bject), bar the presence of an A(gent).The former we may call SA and the latter SO.
To varying degrees, languages make available the means to encode these three core participants through nominal marking systems (case, adpositions), verb marking systems (agreement, auxiliaries, voice distinctions), and through sentential word order.Together these three mechanisms of argument marking variously place the three nuclear sentential participants into one of the following three typological organizations (cf.La Fauci 1997:12;Ledgeway 2012:ch. 7): (2) a.A is formally distinguished from O and, in turn, shares the same formal marking as SA/O; b. O is formally distinguished from A, and, in turn, shares the same formal marking as SA/O; c.A is formally distinguished from O, but the formal marking of S is split between A (= SA) and O (= SO); The arrangement described in (2a) is traditionally termed a nominative-accusative alignment, while the arrangement described in (2b) yields an ergative-absolutive alignment.The third and final active-/stative alignment in (2c) represents a compromise between the two preceding alignments, in that S is formally aligned in part with A and in part with O.It is doubtful, however, that the full grammatical apparatus of any language can be consistently described in terms of just one of these three alignments, although it is often possible to associate particular languages with one predominant orientation.For example, below we shall see that Italo-Greek combines an inherited nominativeaccusative orientation with an emerging active-stative orientation in certain areas of the nominal and verbal systems, as well as at the level of the sentence where we shall review evidence for an activestative orientation in the patterns of sentential word order.

Standard Modern Greek
The nominal system of Standard Modern Greek can unequivocally be described in terms of a nominative-accusative alignment.By way of illustration, consider the three sentences in (3a-c): (3) a. Whether the grammatical subject corresponds to the A of a transitive predicate (3a), the SA of an (intransitive) unergative predicate (3b), or the SO of an (intransitive) unaccusative predicate (3c), it invariably surfaces in the nominative.This is indicated by the nominative, masculine singular definite article o and the final inflexion -s borne by the nominal Jani-in the examples above.By contrast, the grammatical O(bject) of a transitive verb surfaces in the accusative form marked in (3a) above by the distinctive accusative form of the feminine singular definite article tin (cf.nominative form i). It follows that the nominal system of Standard Modern Greek formally contrasts A and S(A/O) (marked nominative) with O (marked accusative) to yield a canonical nominative-accusative orientation which proves totally insensitive to the semantic characterization (AGENT vs UNDERGOER) of the subject.
By contrast, the verb system is less consistent in its morphosyntactic orientation.As the examples in (3a-c) already clearly illustrate, in the active voice the verb system also operates according to a nominative-accusative alignment, in that the finite verb invariably agrees in person and number with the nominative subject (witness the final 3SG inflexion -i in all three examples above), and not with the accusative object when present.However, Greek also presents a medio-passive voice, which formally brings together intransitive UNDERGOER subjects variously drawn from the passive (4a) and unaccusative structures including some deponents, anticausatives, inherent reflexives and reflexive constructions (4b), which all share a distinct set of non-active morphological forms (cf.final 3SG inflexion in -te): (4) a.As the active-passive contrast between (3a) and (4a) reveals, the surface passive subject in the latter is underlyingly an O, hence its SO status.Analogously, the overwhelming majority of nonpassive middles are unaccusative predicates (cf.4b), whose surface subject is analysed in many current formal frameworks as a derived subject moved from or related to the verb's complement position, hence its UNDERGOER interpretation and SO status.We thus see that Standard Modern Greek combines a nominative-accusative formal distinction in the nominal system, inasmuch as all surface subjects (be they A, SA or SO) are systematically marked nominative, with a mixed alignment in the verb system: syntactically the person and number agreement of the finite verb is invariably controlled by a nominative-marked argument in accordance with a nominative-accusative alignment, but, morphologically, the finite verb predominantly displays an active-stative alignment with distinct morphological paradigms for verbs with active subjects (A/SA) on the one hand and stative subjects (SO) on the other (cf.3SG -i vs -te in (3) vs (4) above). 5Extended accusative in early Indo-European The label 'extended accusative' is traditionally used to refer to the extension of accusative case to mark the subjects of a subclass of (intransitive) verbs, a phenomenon commonly attested in many ancient Indo-European languages (Moravcsik 1978;Plank 1985) including, among others, Avestan (Lazzeroni 2002:311-313;Danesi 2014), old Persian (Kent 1946), Gothic (Delbrück 1900), early Germanic (Barðdal (2011), Ancient Greek (Lazzeroni 2013) and Latin. 6In some cases such attestations have been dismissed as cases of textual corruption, morphological conflation or anacolutha (Ledgeway 2012:329;Adams 2013:ch. XII, §6.3).Although there is no doubt some truth to some of these claims in a small number of cases, overall their number is too great and their structural distribution too regular for them to be dismissed as such.The question therefore arises as to whether those attestations which are recognised as genuine outputs of the grammars under investigation should be analysed as constructions simply inherited from a common proto-stage of Indo-European, or as independent developments that arose in individual members of the family (see, for example, Danesi 2014). 7For the sake of the present discussion, it will suffice to observe that, despite individual differences, the distribution of the extended accusative shares some common features across early Indo-European.These include: (i) the greater frequency of the extended accusative in lower-register texts; (ii) the optionality of the extended accusative, insofar as it continues to occur alongside nominative subjects in the same contexts; and (iii) the class of subjects involved, which is generally limited to inactive or involuntary intransitive subjects that exert minimal or no control over the relevant event or situation (Moravcsik 1978:254;Plank 1985).Only rarely and in later chronological periods is the extended accusative found with dynamic intransitive subjects and, even much more rarely, with transitive agentive subjects. 8 The syntactic and semantic criteria governing the extension of accusative-marking to subjects cross-linguistically are summarized in Table 1: Table 1.Extension of accusative: syntactic and semantic criteria 9 5 Cf.Mackridge (1987:96-99), Alexiadou & Anagnostopoulou (2004), Holton, Mackridge & Philippaki-Warburton (2004:ch. 6). 6Regarding the distribution of the extended accusative in Latin, see Löfstedt (1933:329-334), Norberg (1944:21-32), Gerola (1949-50), Bastardas Parera (1953:16-20), Westerbergh (1956:235f.),Herman (1966;1987:102;1995:72-75), Durante (1981:41):41), Pensado (1986), Väänänen (1982:203f.),Helltula (1987), La Fauci (1988:54f.;1997),Zamboni (1998:131f.),Pieroni (1999), Cennamo (2001a;2001b;2009;2011), Lazzeroni (2002:310-312), Rovai (2005;2007;2012;2014); Ledgeway (2011a:459-461); Adams (2013:ch. XII, §6.3); Bentley (2016:821f.). 7On the issue of the origins of other instances of non-nominative subjects in non-personal constructions across the Balkans, see also Friedman & Joseph (2018). 8See, for example, Plank (1985:290), Rovai (2005:62f.),Cennamo (2009:324-326) and Ledgeway (2012:331) on Latin.Cf., however, also Adams (2013:247-249). 9See Cennamo (2009:341) for a third pragmatic criterion, namely accusative extension to constituents originally denoting the topic.

Syntactic criterion unaccusatives → unergatives → transitives SO SA
A good case in point is represented by (late) Latin, where the extended accusative is mainly attested in low transitivity domains in the sense of Hopper and Thompson (1980:252;cf. also Sorace 2000;Rovai 2005:63), in that the appearance of the accusative reflects the underlying semantic case of the UNDERGOER subject formally aligning it with the class of O(bjects). 10It therefore typically surfaces with SO-type subjects in middle constructions with deponents (5a), anticausatives (5b), passives (5c), impersonal passives (5d), and existentials (5e), as well as in active syntax in conjunction with unaccusatives (5f) and, in particular, the verb ESSE 'be' (5g).
Semantic criterion inactive inanimate → active inanimate → active animate this.FSG.NOM be.PRS.IND.3SG'if he is unwounded, this is all part of the healing process' Cross-linguistically, instances of the extended accusative also tend to occur in varieties whose case systems are undergoing considerable weakening, a feature also readily observable in late Latin (but cf.old Persian, Danesi 2014:251, fn.70).Arguably, this 'extended' use of the accusative which increases greatly in frequency in later Latin texts can be construed as a gradual alignment shift in the nominal system, whereby non-active subjects come increasingly to be formally marked on a par with transitive objects.In particular, nominative is reserved for A/SA subjects and accusative for O(bjects) and SO subjects in accordance with an emerging active-stative alignment and, more rarely in later periods, also for SA subjects in accordance with an ergative-absolutive alignment (La Fauci 1997:57ff;Zamboni 1998:131ff;Ledgeway 2012:332;Bentley 2016:822).

5
Extended accusative in Italo-Greek

A note on case-marking in Italo-Greek
Just like Standard Modern Greek, Italo-Greek determiners and nominals show morphological casemarking for nominative, accusative and genitive-dative across three genders (masculine, feminine, neuter) and two numbers (singular vs plural).However, the morphophonological reduction of several of its nominal inflexional markers has led to many instances of apparent syncretism.Although in some cases there arise genuine instances of neutralization, in most cases the apparent syncretisms are crucially resolved by means of an additional phonosyntactic strategy whereby, following an original sandhi assimilation, an erstwhile final inflexional consonant today surfaces in the consonantal lengthening of the initial consonant of the following word.For example, when preceded by the definite article (cf.As we shall see, it is precisely the presence or otherwise of RF that will allow us in many cases to discriminate between nominative and accusative marking on many of the subjects discussed below.

Accusative subjects in Griko
On a par with Standard Modern Greek (cf.§3), Griko apparently presents a core nominativeaccusative alignment, formally contrasting subjects and objects.Thus, we see in (8) that subjects of transitives (A; 8a), unergatives (SA; 8b) and unaccusatives (SO; 8c) are treated uniformly in that they are all marked nominative, in contrast to transitive O(bject)s which are systematically marked accusative (cf.us piattu in 8a): (8) a. E Marìa pleni kalà us the.FSG.NOM Maria.FSG.NOM wash.PRS.3SGwell the.MPL.ACC piattu.(Calimera) 17 Given their historical evolution (Manolessou 2005b), in the literature (Italo-)Greek verb forms in -onta/-onda such as fanta 'eaten, eating' have been variously referred to as participles (e.g.Rohlfs 1977:109f., 200f.;Mackridge 1985;Manolessou 2005a) and gerunds (e.g.Katsoyannou 1995;Holton et al. 2012).Without taking a firm position, for ease of exposition here we simply gloss them as NON-FIN.The same gloss is applied to non-finite forms in -meno/-a. 18Non-proparoxytone feminine singular nouns ending in -a are always formally ambiguous between nominative, accusative and genitive (unless marked accusative through RF).In what follows we shall gloss them as NOM-ACC-GEN only when the context clarifies that the noun is genitive.In contexts where the noun may only be either nominative or accusative, the formal ambiguity with genitive will not be indicated.However, alongside such prototypically marked arguments, viz.nominative subjects and accusative objects, our corpus also includes attestations of subjects marked with accusative case. 19 These were produced by both proficient (p.) and semi-speakers (s-s.) from all eight villages and belong to the spoken informal register of the language. 20As for their syntactic distribution, they can occur in root (9), embedded (10) and adverbial ( 11 20 By semi-speaker, we mean speakers belonging to one of the following three subcategories: (i) L1 speakers whose once full competence has been eroded as a consequence of a lack of use of the language for a more or less extended period of time; (ii) L1 speakers who have naturally acquired Griko from their families, but only partially; and (iii) local L2 speakers who have decided to learn Griko later in life, but who have never reached full native-like competence (Ledgeway, Schifano & Silvestri 2018a:13); see also Dorian (1980;1981) and, for Italo-Greek, Stamuli (2007:65-67) and Guardiano & Stavrou (2019:5-6). 21A different type of accusative subject is also found in a subset of embedded clauses, for which see Ledgeway, Schifano & Silvestri (2018c Given the highly attrited status of the language now spoken in a rapidly-shrinking speech community which is today in constant contact with the dominant neighbouring Romance varieties that lack a formal case system (viz.Salentino and (regional) Italian),24 it is tempting to disregard examples such as ( 9)-( 11) as random replacements produced by speakers whose competence has been drastically eroded (Rohlfs 1977:69;cf. also Rossi Taibbi &Carcausi 1959:LIIIf., LIX andKatsoyannou 1999 for Greko).However, a careful investigation of the syntactic distribution of accusative subjects in our corpus reveals a number of interesting affinities with historical attestations of the so-called extended accusative, suggesting that they should be interpreted as the authentic output of a changing grammar rather than performance errors of an increasingly less native grammar.
Indeed, overall instances of accusative subjects in our corpus of Griko are less controversial than many attestations from early Indo-European languages reviewed above, inasmuch as they have been systematically produced by native speakers who have also confirmed their grammaticality.Moreover, they share a number of common features with the historical instances of extended accusative reviewed above.First, Griko accusative subjects are also optional: for all the examples including an accusative subject there are speakers who produced the same sentences with a regular nominative subject (12ab).At the same time, speakers who produced accusative subjects also produced regular nominative subjects, both in the (near-)identical sentences ( 13)-( 15) and in different ones ( 16)-( 17 Second, although we are dealing with a predominantly spoken code, accusative subjects in Griko appear to belong predominantly to the spoken and most informal registers of the language.This is confirmed by a preliminary investigation of early and contemporary written sources which has brought to light some examples of accusative subjects, nearly all of which are restricted to early written records (cf.the.FPL.ACC schools.FPL.NOM-ACC 'When the schools closed, my mother would take all three of us and would take us alone by train to Switzerland to my father's, and we would stay there until the schools opened again.'Note that optionality extends to written sources too, insomuch as nominative subjects are also regularly attested, witness the following examples, where all the subjects are marked as nominative despite their occurrence with an unaccusative verb (20a-c) and the copula BE (20d): Finally, Griko accusative subjects crucially present the same syntactico-semantic restrictions outlined above for the extended accusative in early Indo-European.In particular, the extended accusative targets intransitive subjects which are relatively inactive and inert, in short UNDERGOERS.As a consequence, in our corpus accusative subjects in Griko are principally attested with middle syntax, including deponents with reflexive interpretation (21a; cf. also 10), unaccusatives (21b; cf. also 9a-d, 11, 12b,13b,14b, 15b, 16b, 17b, 18a-b), anticausatives (cf. 19b), and the copula BE (21c; cf. also 18c-d).
( 21 To this we can also add low transitivity domains such as example ( 22) involving a stative predicate with a surface subject characterized by minimal control.Indeed, in accordance with Hopper and Thompson's (1980:252)  Conversely, the vast majority of animate and/or active subjects with transitive verbs included in our corpus bear the expected nominative marking.The very few instances of accusative marking in these contexts such as (23a) were only produced by semi-speakers (cf.fn.20).This suggests that such rare examples should be interpreted either as genuine performance errors or as a separate case of reanalysis within a drastically more attrited grammar not shared by proficient native speakers (23b).3 all the classes of verb which are attested in our spoken and written corpus of Griko with an accusative subject: 28 Interestingly, in this example the referential predicative adjectival complement kkalì of the accusative subject (Diu) mèdecu is inflected nominative (cf.accusative (c)calù), showing a mixed pattern of case-marking.We also find the opposite pattern where the subject surfaces in the nominative and its predicative complement in the accusative, witness the following Greko example from the now defunct variety spoken in Roccaforte:  From the overview in Table 3 it is clear that the extension of the accusative to subjects in Griko follows a regular structural distribution targeting unaccusative syntax according to a pattern analogous in all relevant respects to that observed for early Indo-European (cf.§4).In particular, the extension of the accusative serves to draw a formal distinction on the one hand between SO (together with O) marked accusative and A and SA marked nominative on the other.We thus see the emergence of a competing active-stative alignment in the nominal domain which, although now well advanced in Griko, has not (yet) replaced the erstwhile nominative-accusative alignment with SO subjects still optionally occurring in the nominative.Indeed, in some cases nominative marking is still obligatory today.More specifically, while the extension of the accusative can target nominals which are high in the animacy scale (Silverstein 1976;cf. also Lazzeroni 2002:309;Rovai 2005:64) such as proper nouns and kinship terms, it is never found with pronouns.This undoubtedly reflects the fact that case distinctions are typically most robustly retained with pronouns (Spencer 2009:195), as evidenced by all modern Romance varieties (with the exception of Romanian) where case distinctions have been lost on full DPs but retained to differing degrees in pronouns (Blake 2004:178f.;Sornicola 2011;Dragomirescu & Nicolae 2016:913-916).Revealing in this respect is the example in (14b), repeated here as ( 24), where we see that the first-person plural subject imì occurs in the nominative (cf.accusative (e)mà(s)), but its accompanying (appositional) nominal modifier antròpu (cf.nominative àntropo) occurs in the distinctive accusative form.

Accusative subjects in Greko
Having ascertained above the presence of accusative subjects in the Italo-Greek variety of Griko spoken in Salento, it is instructive as a point of comparison to consider now Greko spoken in southern Calabria.The results of our fieldwork in southern Calabria show a situation very similar to that reviewed above for Griko.Indeed, already in an article from (1999), Katsoyannou had noted a small number of instances of accusative subjects in her data from Gallicianò collected in 1984 (cf. also Rossi Taibbi & Carcausi 1959:LIIIf., LIX;Rohlfs 1977:69), some examples of which are reported in (25).3125) and (26b) reveals an inescapable structural regularity to the extension of the accusative in that it invariably targets unaccusative syntax (namely, SO subjects). 32 Further substantial confirmation of this emergent active-stative pattern also comes from a consideration of late nineteenth-and early twentieth-century written texts.Once again the instances of accusative subjects are quite numerous in collections of originally orally-recounted tales and stories, witness the following examples taken from Rossi Taibbi & Caracausi (1959) As these illustrative examples reveal, accusative subjects consistently occur with core unaccusatives, including verbs of motion and position.However, we also find once again, and indeed more frequently in these late nineteenth-and early twentieth-century texts, nominative subjects in the same contexts, as the following representative unaccusative examples demonstrate. 32Significantly, Greko is also reported by Katsoyannou (1999:243f.) to employ the accusative form of nominals in asyntactic uses (e.g.lists, citation forms), a feature also reported for the extended accusative in late Latin (see Ledgeway 2012:304f.). 33Observe the RF effect produced by the masculine singular indefinite article ena (< enan) in this example.( 28 32a-b) with coordinated subjects in a context of low transitivity (negated modal): as with the Griko example in (14b, 24), example (32b) shows that accusative-marking of subjects extends to nouns, but not to pronouns which must obligatorily occur in the nominative.Similar evidence can also be found in another corpus of contemporary data (cf.Stamuli 2007), where again accusative subjects are attested with unaccusative verbs (33), the copula BE (34), and with an involuntary subject (cf.'the scabies in 35) exerting no control over the event: (33) a. 'irtɛ miaŋ ɣi'nɛka.(Gallicianò, Stamuli 2007:126) 35 For final -n on the feminine articles in (28a) and (28b), see fn. 27. 36In Greko there are two distinct forms for 'dog', namely o sciddho (m.) and to sciddhì (n.) (M.O.Squillaci, p.c.).Although some speakers mix these two forms, this particular informant in (31) consistently uses the masculine form throughout all the interviews, hence the selection of to in this example can only be interpreted as accusative marking, as also confirmed by the fact that it licenses RF here.

Interim conclusions
In summary, we have observed how within the nominal system the Italo-Greek varieties Griko and Greko present increasing evidence for a progressive shift from a traditional nominative-accusative alignment, in which an extended nominative marks all surface subjects (A, SA, SO) in contrast to the accusative restricted to marking O(bjects), towards an active-stative alignment in which the accusative is extended beyond O(bject) nominals to now include SO subjects thereby restricting nominative-marking to just A and SA subjects.However, the emergence of the so-called extended accusative in Italo-Greek represents just one of several surface reflexes of an original Romance active-stative alignment which, in a process of partial replication, has progressively been extended and adapted in the native grammars of Italo-Greek speakers.In the following sections we shall consider some further evidence for this hypothesis from the verbal and sentential domains where other reflexes of a Romance active-stative syntactic alignment have transparently been replicated in the local Greek varieties, confirming that Italo-Greek is undergoing a partial alignment shift.

6
Other reflexes of an active-stative alignment

Auxiliary selection
Beyond accusative subjects, the effects of an active-stative alignment are also clearly observable in the patterns of perfective auxiliary selection.Historically, all Romance varieties, and still many today (cf.Bentley 2016:824), exhibit an alternation in the selection of the auxiliaries HAVE and BE in conjunction with the past participle in the formation of various compound verb forms. 37In Italian, for example, auxiliary HAVE (= avere) is selected in conjunction with transitives/unergatives (36a), whereas unaccusatives (36b), including the passive (36c), select auxiliary BE (= essere).
(  (cf. Rohlfs 1977:198;Gemma Italia & Lambroyorgu 2001:109f.;Tommasi 2001:188;Baldissera 2013:42) report the sole use of auxiliary HAVE in conjunction with the invariable non-finite form in -onta (cf.footnote 17 above) for the formation of the pluperfect, regardless of the thematic structure of the lexical verb.Thus, in (37) we witness the systematic use of íxa 'had' with both the transitive/unergative gráfsonta 'written' and the unaccusative értonta 'come'.Although most of the data in our oral corpus of Griko comply with this picture (cf.38a-b), some speakers occasionally show signs of an active-stative split of the type exemplified in (36), selecting HAVE with unergative/transitive verbs (39a) and BE with deponent verbs with an UNDERGOER subject (39b) in accordance with an A/SA vs SO alignment: Given these facts, it is highly plausible to interpret the novel differential selection of the auxiliaries observed in (39a-b) as part of a larger Romance active-stative alignment which is influencing the morphosyntax of Griko.While it might be objected that in the relevant pluperfect paradigm the local Romance dialects show the generalization of a single auxiliary (cf.41a-b), and furthermore the opposite auxiliary to that traditionally selected in Griko (cf.38a-b), the relevance of the more frequent present perfect paradigm (cf.40a-b) must not be forgotten, nor the influence of (regional) Italian on these Greek speakers, a genuine part of their linguistic repertoire, which, following the pattern in (36a-c), consistently marks the active-stative auxiliary split also in the pluperfect (42).( 42 We note finally that in the Greek of southern Calabria, by contrast, the sole auxiliary consistently employed in the pluperfect is BE (43a), a pattern which is extended to the local Romance dialects of the area (43b) which do not show an active-stative split in the perfective auxiliary (Schifano, Silvestri & Squillaci 2016;Squillaci 2017: §2.7;Remberger 2018).In this domain of the grammar, the overt reflexes of an active-stative alignment are therefore more advanced in Griko than in Greko.

Adverb agreement
Similar conclusions to those seen for auxiliary selection in the verbal domain can be drawn from the sentential domain in relation to the phenomenon of adverb agreement.As demonstrated in detail in Ledgeway (2011b;2012:ch. 7;2017) and Ledgeway and Silvestri (2016), dialects of southern Italy show a formal syncretism in the categories of adjective and adverb, with adverbial functions systematically expressed by the category of adjective.Yet, in contrast to most languages that conflate the functions of adjective and adverb into a single formal class (cf.Romanian, German) in which the adverb assumes an invariable (e.g., default masculine singular nominative) form, in the dialects of southern Italy the adjective in adverbial function may show overt agreement for gender and/or number.Such agreement is not, however, unconstrained but, rather, follows regular and structurally predictable principles which in most dialects can be formalized in terms of an active-stative split.In the unergative examples in (44), the adjectival adverb invariably occurs in its default masculine singular form irrespective of the number and gender features of the (implied) subject, allowing us to conclude that the SA subject is unable to control the agreement features of the adverb.In the respective transitive and unaccusative examples in ( 45) and ( 46), by contrast, the adjectival adverb now shows full agreement with the O(bject) in the former case and with the SO subject in the latter case.The relevant agreement patterns can thus be readily framed in terms of a canonical active-stative alignment, inasmuch as there obtains a split between those participants (A, SA) which do not license adjectival adverb agreement and those (O, SO) which do.
As already noted in the literature (Rohlfs 1977:135f.),Italo-Greek exhibits both non-inflecting deadjectival adverbs in -a, as in (47), and adjectival adverbs showing agreement as in ( 48 The data from written sources and our corpus reveal a similar picture for Griko.In particular, many of our speakers produced, alongside non-agreeing forms of the adjectival adverb, agreeing forms in conjunction with full DP objects (49a), including obligatory agreeing forms with clitic O(bjects) (49b) according to a pattern also found in southern Italo-Romance (cf.Ledgeway 2011a; 2017), as well as with unaccusative SO subjects (50a-b) (cf.also 50c from Palumbo 1971).Crucially, though, none of our speakers accepted agreeing forms in conjunction with unergative SA subjects (51a-b), which do not show agreement in written sources either (51c).( 49 The examples in ( 52)-( 53) show agreement of the adjectival adverb which is variously controlled either by an O(bject) (cf.52a-c) or by an SO unaccusative subject (cf.53).However, once again we observe that such agreement is optional, witness the use of the non-agreeing adverbial form kalá in (52d-e), and indeed impossible with unergative SA subjects (54).
In summary, the evidence reviewed in this section reveals an additional reflex of an emerging, though not yet fully stabilized, active-stative alignment in the distribution of adjectival adverb agreement which proves sensitive to the A/SA vs O/SO split.

Sentential word order
One final piece of evidence in favour of an ongoing shift towards an active-stative alignment comes from sentential word order.With the exception of some modern Gallo-Romance varieties, Romance languages have broadly converged towards an unmarked SVO word order.However, this SVO order masks in most modern varieties an active-stative alignment where S and O are to be understood more broadly as A/SA and O/SO, respectively (cf.Bentley 2006:364-368;Ledgeway 2012:334f.).This explains why in the unmarked case (answering the question: What happened?)transitive (55a) and unergative (55b) subjects occur preverbally, whereas unaccusative subjects (55c) occur in a postverbal position corresponding to that occupied by the complement in transitive constructions (cf.la finestra in 55a): Now, in contrast to Standard Modern Greek where sentential word order is notoriously very liberal (Philippaki-Warburton 1985;Mackridge 1987:234-239;Tsimpli 1990;Horrocks 1994;Holton, Mackridge & Philippaki-Warburton 2004:229-232;Roussou & Tsimpli 2004;Anagnostopoulou 2013:13, 20-22), the word order of Italo-Greek is considerably more constrained, excluding, for example, VSO orders in root clauses.Rather, on a par with what has just been seen for Italian in (55), the neutral word order of Griko follows an unmistakable active-stative split.Consequently, transitive (56a) and unergative (56b) subjects occur in preverbal position in the unmarked case, whereas unaccusatives subjects (57a) occur in postverbal position on a par with transitive objects (57b).( 56 Once again, evidence from word order points to an active-stative orientation at the level of the sentence to parallel the identical alignment pattern examined above for adjectival adverbs, as well as those in the verbal domain (auxiliary selection) and nominal domain (restricted nominative for A/SA and extended accusative for O/SO).

Conclusion
Above we have reviewed considerable evidence from the nominal, verbal and sentential domains of Griko and Greko which highlight an ongoing shift from an original nominative-accusative alignment towards an active-stative alignment.In the nominal domain we have seen how this alignment shift results in a redistribution of nominative and accusative case-marking according to underlying semantic roles, rather than surface syntactic relations.Accordingly, active subjects (A/SA) of transitives and unergatives are case-marked nominative, whereas stative subjects (SO) of unaccusatives are increasingly marked accusative on a par with canonical O(bjects). 40In this way, 40 It is also striking that, in contrast to Griko, all examples of accusative subjects in Greko noted in Rossi Taibbi and Caracausi (1959), as well as those in Katsoyannou (1999) and in our corpus (with the exceptions of the relative example in (26b), the coordination example in (32b), and the example with copula BE in 34), occur in the typical postverbal O(bject) position; whenever inactive subjects (SO) occur in preverbal position, the unmarked position of active subjects (A/SA), they invariably surface in the nominative.Particularly revealing in this respect is the minimal pair reported by Katsoyannou (1999:242)  We leave it to future work to establish to what extent accusative-marking of inactive subjects, at least in Greko, is also structurally tied to their surface position.If our interpretation of the facts is correct, then this would suggest that Greko represents a more conservative variety than Griko, inasmuch as accusative-marking has not yet been (fully) extended to the preverbal position as in Griko.Presumably, this tendency also explains the sole example in our Greko written corpus of the otherwise exceptional accusative-marking of a transitive subject (cf.ii) ostensibly determined by its postverbal the Italo-Greek nominal system comes to mirror the formal split already visible in the verb system where, on a par with Standard Modern Greek (cf.§3), the morphological paradigms of the active and non-active (viz.medio-passive) largely correlate with the distribution of A/SA and SO subjects, respectively, as schematized in Table 4 for the present and past imperfective of plen-'wash' (cf.Rohlfs 1977:110-113, 199f.):Table 4. Correlations between active and non-active morphology and case-marking A/SA [+Nom] <---------------------------->SO [+Nom/Acc] Active Non-active Present Imperfective Griko Greko Griko Greko plen-o plen-i(s) plen-i plèn-ome plèn-ete plènune plen-o plen-i(s) plen-i plèn-ome plèn-ite plèn-usi plèn-ome plèn-ese(s) plèn-ete plen-òmesta plen-èsesta plèn-utte plèn-ome plèn-ese plèn-ete plen-òmesta plèn-este plèn-onde Past Imperfective èplen-a èplen-e(s) èplen-e plèn-amo plèn-ato plèn-ane èplen-a èplen-e(s) èplen-e eplèn-ame eplèn-ete eplèn-asi plèn-amo plèn-aso plèn-ato plen-àmosto plen-àsosto plèn-atto eplèn-ommo eplèn-esso eplèn-eto eplen-òmesta eplèn-este eplèn-ondo It is thus legitimate to ask why in other varieties of Greek such as Standard Modern Greek a similar active-stative alignment has not arisen in the nominal case system.One possible answer would be to invoke endogenous factors present in Italo-Greek, but not in other varieties of Greek.This is essentially the line taken by Katsoyannou (1999:239f.) in her analysis of accusative subjects in Greko, which she interprets as the surface effect of a case system in an irreparable state of collapse in a highly endangered language which is rapidly being abandoned by a bilingual community with greater native competence in a Romance variety without a case system.Yet, Katsoyannou's view represents a misconception of the Italo-Greek case system which, despite some apparent superficial neutralizations (cf.Table 2), still constitutes a robust system with a high functional load, consistently with Dimmendal's (1998:87) claim based on Dorian's (1978:608) original observation that 'an obsolescent language often dies "with its morphological boots on"' (on the reduction of the morphological structure of the case system in Italo-Greek, see also Guardiano & Stavrou 2019).This is clearly demonstrated by the representative Italo-Greek nominal paradigms with accompanying definite article in In particular, we see that in masculine and neuter nouns the core distinction between nominative and accusative is neutralized.With neuter forms this is unsurprising in that nominative and accusative are syncretic in the neuter in other Greek varieties too (and in Indo-European more generally; cf. also Table 5), but this has never led to a generalization of accusative-marking of subjects in these varieties.However, the neutralization witnessed in masculine indefinites in Griko where, for example, underlying nominative and accusative forms such as èna(n) liko and ènan liko can both surface indiscriminately as èna lliko, could a priori be argued to provide the original impetus for a progressive, but still optional, extension of accusative-marking to the subject relation. 42Tempting though this superficial morphophonological explanation might appear (cf.also foonote 27), it must be immediately dismissed since it incorrectly predicts an indiscriminate extension of accusativemarking to all surface subjects.Yet, we have seen that accusative-marking of subjects is specifically 42 Prevocalic contexts where, for apparently euphonic reasons, non-etymological -n surfaces most robustly on the nominative indefinite article (i.a), including in Greko (i.b;cf. Rossi Taibbi & Caracausi 1959:LVIII), also give rise to (apparent) cases of surface neutralization of nominative and accusative in masculine (and of course neuter) noun phrases (cf. Rohlfs 1977:69).Instead, we argue that the emergence of accusative subjects in Italo-Greek is due to exogenous factors and, in particular, to language contact with Romance.This immediately explains why the extended accusative is only found in those Greek varieties that have been in contact with Romance, but not, for example, in Standard Modern Greek.Moreover, although Griko and Greko are not, and never have been, in contact with one another (Profili 1983;Katsoyannou 1995;Manolessou 2005;Squillaci 2017:2), they have both independently developed the extended accusative precisely because they have both individually been in intense contact for centuries with Romance varieties where the evidence for an active-stative alignment is robustly attested in various areas of the grammar (for an overview, see Ledgeway 2012:ch.7).As a consequence, the speakers of Italo-Greek are also native speakers of local Romance varieties, and in most cases more natively competent in Romance than Greek, such that after many centuries of Greek influencing local Romance varieties, a case of socalled spirito greco, materia romanza 'Greek spirit, Romance material' (cf.Ledgeway 2006;Ledgeway, Schifano & Silvestri in press), their local Greek varieties today often display many Romance features, a case of spirito romanzo, materia greca 'Romance spirit, Greek material' (cf.Ledgeway 2013;Ledgeway, Schifano & Silvestri 2018b).It is therefore our contention that the emergence of the so-called extended accusative in Italo-Greek represents just one of several surface reflexes of an original Romance active-stative alignment which, in a process of partial replication, has progressively been extended and adapted in the native grammars of Italo-Greek speakers.It is for this reason that we have been at pains to show above that the extension the accusative should not be considered an isolated phenomenon within the grammars of Italo-Greek, but must, rather, be interpreted as part of a larger gradual and ongoing shift towards an active-stative alignment which surfaces in various areas of the nominal, verbal and sentential domains.
Within this context, it is significant to note that, while the surface reflexes of this active-stative alignment observed in the verbal (auxiliary selection) and sentential (adjectival adverb agreement, subject placement) domains of Italo-Greek find an immediate structural parallel in Romance, ultimately the result of a process of PAT(tern) replication (Matras & Sakel 2007;cf. also Heine & Kuteva 2006), accusative-marking of stative subjects represents a Greek innovation since the relevant Romance contact varieties do not have a (nominal) case system.What we therefore see is an expansion of a Romance alignment PAT(tern) which, once embedded in the replicating Greek varieties through the increasing establishment of active-stative-driven auxiliary splits, adjectival adverb agreement and differential subject placement, is further reinforced by the extension of the alignment to new areas of the grammar using Greek MAT(erial) amenable to this same split.At the same time, we must not underestimate the complementary role of the Italo-Greek verb system where the inherited formal opposition between active and non-active verb forms (cf.Table 4) readily maps onto the semantico-syntactic distribution of nominative and accusative subjects, respectively, whilst further strengthening the emerging active-stative patterns in the auxiliary system, adjectival adverb agreement and subject placement. 43We are therefore led to conclude that the role of contact-induced change in the emergence of accusative-marking of subjects is only indirect (cf.Willis 2017: §26.3): the motivation for the change clearly requires a language-internal, endogenous account in terms of spontaneous innovation (namely, expansion of active-stative syntax to the nominal domain), but the original catalyst for the introduction of the syntactic alignment PAT(ern) that it extends is the result of language-external, exogenous factors, namely contact with Romance.
In conclusion, our discussion of Italo-Greek and Romance alignments has shown how, at least on the surface, the grammars of these two linguistic groups are in many key respects converging, to the extent that the observed structural parallels are far too striking for them to be dismissed as accidental 43 Relevant here is Guardiano et al.'s (2016) Resistance Principle according to which syntactic change under horizontal pressure only takes place if surface evidence that makes such a change structurally possible is already independently available in the language.
or the output of heavily attrited grammars.Rather, they must be considered the result of centuriesold intense structural contact between Greek and Romance, ultimately to be placed towards the upper end of the five-point scale of contact intensity proposed by Thomason & Kaufman (1988).Indeed, while it is well known that traditionally the direction of such contact has consistently involved the transfer and extension of original Greek structural features into the surrounding Romance varieties (cf.Ledgeway 2013), large-scale linguistic shifts among recent generations of the southern Italian Greek-speaking communities towards Romance have resulted in a reversal of the direction of contact.Consequently, today we see many examples of transfer of Romance structural features into Italo-Greek.In this respect, the ongoing emergence of an active-stative alignment in the syntax of the nominal, verbal and sentential domains of Italo-Greek represents a prime example of Romance-Greek contact and, in particular, highlights how the role of language contact may genuinely prove pervasive insofar as it is even able to trigger a shift in alignment, arguably involving a change of a macroparametric order (cf.Sheehan 2014).
limited to stative subjects (SO), incontrovertibly showing that what lays behind the extension of the accusative is of a structural nature replicating a distribution independently observed in early Indo-European.

Table 2
) clauses, and both in pre-and postverbal positions (cf.9a vs 9b): 21 MSG.ACC priest.MSG.ACC go.IPFV.PST.3SGgo.around.NON-FIN house.NSG spiti. 22(Martano, p.) house.NSG 'The priest used to go around visiting all the houses.' die.PFV.PST.3SGthe.MSG.ACC sciddho.(Sternatia,p.)dog.MSG.NOM-ACC 'I'll go and fetch my brothers to tell them how their dog died.'19Neutersubjects are not taken into account here as nominative and accusative are syncretic in the neuter gender; indeed in glossing neuter forms below we do not indicate case, unless genitive.Note, furthermore, that we do not distinguish in what follows between (abstract) Case and (morphological) case, but simply use the spelling 'case' throughout.
FSG.ACC daughter.FSG.ACC the.MSG.GEN Petro.MSG.GEN run.PFV.PST.3SGDAT= seem.NON-ACT.IPFV.PST.3SG that at.the.NSG morning.NSG i Mmarìa epplìnite the.FSG.ACC Maria.FSG.ACC NEG=clean.NON-ACT.IPFV.PST.3SGkalà. 23(Corigliano, s-s.) well 'It seemed to me that Maria didn't used wash properly in the morning.'PFV.PST.3PLat.home tell.PFV.PST.3PLIRR.PRT go.SBJV.3PLto.the ospitali.(Corigliano, s-s.) hospital.MSG.ACC 'When Pietro and Rita arrived home, they were told to go to the hospital.' ). 22 Note that in examples such as (9a) we gloss (and place in bold) the masculine singular article o as distinctively accusative, and not as ambiguously nominative-accusative (cf.nominative o vs accusative (t)o(n)), since it produces RF on the following nominal.Clearly, what is relevant in such examples is the determiner's ability to license RF and not just its surface form.(11)Motte to Ppetro ce ti Rrita when the.MSG.ACC Petro.MSG.ACC and the.FSG.ACC Rita.FSG.ACC ). SBJV.3SG the.MPL.ACC nails.MPL.ACC IRR.PRT fix.SBJV.3SGthe.FSG.ACC Morosi 1870)of originally orally-recounted tales and stories (18) and contemporary informal texts such as those exemplified in (19) taken from a selection of personal testimonies about life in the past published in the local magazine I Spitta ({HYPERLINK "http://www.rizegrike.com/spitta.php"}):26 MPL.ACC be.PRS.3PLgood.MPL.NOM the.MSG.NOM other.MSG.NOM-ACC e mmu piace poddhì. 28(Calimera, p.) NEG= me.DAT= like.PRS.3SG much 'Two doctors are good, as for the other I don't like him very much.' interpretation of 'low transitivity', we note that example (22) involves just one participant, an A low in potency, and denotes a non-action (viz.state) which is atelic, nonpunctual and negated.IRR.PRT eat.SBJV.3SGwith= us.ACC 'Maria doesn't know if she can come and eat with us.'

Table 3 .
All attestations of accusative subject in Griko Unsurprisingly, many local Romance dialects of Salento also display a robust active-stative split in auxiliary selection, at least in the present perfect where once again HAVE surfaces with transitives/unergative (40a) and BE with unaccusatives (40b), though not in the pluperfect where most Salentino dialects generalize BE across all verb classes (41).
By way of example, consider the following Romance examples from Salento.
) a. E Marìa èpline us the.FSG.NOM Maria.FSG.NOM clean.PFV.PST.3SGthe.MPL.ACC MPL.ACC good.MPL.ACC the.FPL.ACC windows.FPL.NOM-ACC FSG.NOM mother.FSG.NOM old.FSG.NOM-ACC the.FSG.GEN Maria.FSG.GEN This same active-stative distribution of adjectival adverb agreement is also in evidence in Greko, witness the following representative examples taken from both our fieldwork and written sources.
in (i.a-b) produced by the same speaker, where the pre-and postverbal positions correlate with nominative and accusative case-marking, respectively (we assume, despite appearances, that the meteorological predicate 'ekamen (tin) kia'ria is not a true transitive verb, but a compound unaccusative with cognate surface object and a stative (viz.SO) subject xri'sto):

Table 5 .
Italo-Greek definite nominal paradigms NPL make.PRS.3PLthe.FPL.ACC women.FPL.NOM-ACC 'What type of cakes do the women make?'However, it is true that nominals introduced by the indefinite article do introduce some limited ambiguity into the system, as Table6illustrates(cf. Rohlfs 1977:68f.).